Improving margins is often painful, but the levers are known. You can tell a company to reduce SaaS spend by 20%, and there is a reasonably clear path to execution, especially now with AI. You can tell a company to reduce personnel expenses by 20%, and it can just freeze new hires, letting natural attrition do the work. To use the cringe term du jour, improving margins is a legible challenge for management teams.

Conversely, you cannot tell a company to “grow more” and expect the same probability distribution of success. New products may fail. New customer segments may not convert. New geographies may disappoint. New channels may look promising and then stop scaling. Even when growth initiatives work, the feedback loops are slower and attribution is messier.

By the way, growth is also more valuable. Bessemer Venture Partners’ Rule of X says something that most software investors intuitively understand: growth is worth (much) more than profitability. For late-stage cloud (~SaaS) companies, Bessemer argues that revenue growth should be weighted roughly 2 to 3x more than free cash flow margin when assessing valuation.

That makes sense. Growth is open-ended. You can grow 20% per year for a looooong time. A company can keep expanding into new customers, new products, new use cases, new geographies, new channels, and new price points for a very long time. Some companies, like the hyper scalers, have even shown that growth begets more growth, a phenomenon W. Brian Arthur, from the Santa Fé Institute, dubbed “increasing returns”.

Profitability improvement, on the other hand, is more bounded. You just can’t reduce expenses and costs by 20% per year for a very long time. You can freeze headcount, even make a small rif. Cut vendors. Renegotiate contracts. Reduce waste. But eventually you run into a floor. Accounting physics says expenses cannot go below zero. And well before that, cost-cutting starts damaging the business - cuts cut muscle and reach the bones, harming future prospects for the company.

This is the important asymmetry: Growth is more valuable, but cost-cutting is more legible.

AI makes this legibility asymmetry much more dangerous for SaaS. Replacing a SaaS product with an AI-generated copy is a legible management project. You know the vendor-to-be-replaced. You know the contract value. You know the workflows. You know the users, and they’re much more forgiving than customers. You know which features need to be replaced. You know what “good enough” looks like. You know what the savings will be if the project works.

Aside from legibility, there’s also the incentives picture. Public company management teams are mostly compensated for short term results, even when they get stock, because stocks move up, mostly, when a company beats the Street’s earnings/cash flow estimates. The perfect story is the “beat and raise” story. Management teams are incentivized to post guidance they can beat, beat it, and raise guidance just enough to beat it again next quarter.

Given the legibility issue, the sure-fire way to get more earnings/cash flow next quarter is to cut costs and expenses. This is why the “SaaS apocalypse” may be less about software suddenly becoming useless and more about software spending becoming a highly valued target for cutting, for lack of better options (and/or lack of management competence).

Some SaaS products will be extremely hard to replace, of course. ERPs, regulated workflows, highly integrated platforms (ERPs?), mission-critical infrastructure, and products with strong network effects will be more protected. But a lot of software is not that. A lot of software is workflow glue, reporting, internal tooling, lightweight automation, dashboards, approval flows, data movement, content generation, support operations, and administrative process.

Growth is the more valuable, high risk, hard to pull lever. Cost-cutting is the less valuable, lower risk, legible lever. And when executive teams are under pressure to deliver, the legible lever often wins.

---------- Forwarded message ----------
From: Francisco Homem de Mello <franciscomello@XXXX.com>
Date: Monday, June 2, 2014 at 17:00 GMT-3
Subject: On Ned Beauman’s Glow
To: <editorial@granta.com>

To the editors,

Ned Beauman’s Glow left me flummoxed. Actually, the first twelve pages left me flummoxed, so much that I couldn’t go on reading. Or was it startled, in a very bad way?

I purchased the book only because it read, on the upper right hand corner of the cover, “A Granta Best Young British Novelist 2013”. I thought to myself if Granta said so, it must be good. Why not help a new novelist? “Ned Beauman has reinvented the international conspiracy thriller” is not bad either. Should I believe the back of a book? I am a believer. And, everybody loves conspiracies.

So there was I sitting in a not-so-comfortable chair in Madrid Barajas Airport, waiting for my turn to board a flying tin can, when I looked into my Relay plastic bag and decided to tackle Glow before John Le Carré’s A Delicate Truth. An eight-hour flight to Brazil awaited me and I was eager for company.

On the first page of the book Ned was describing a launderette rave party in London when I stumbled upon a – very – unfortunate analogy tucked into a description of a girl our protagonist had sighted on the dance floor: “and she has one of those faces where the entire bone structure seems to ramify from the cheekbones in such a way that the result looks like a 3D computer graphic from the eighties because it’s composed of such an economical number of sharp, flat planes, except that the angles are confused here by strands of long black hair escaping from where she’s pinned the rest of it up at the back of her head;” I tried to think about what it meant to have lines ramifying from the cheekbones, but my imagination goes only so far, and thinking about the economical sharp flat planes which are so 80s didn’t help me either.

My first contact with serious writing came when I did a “Inside the writer’s mind” course at Gotham Writers. One of the things I remember from that freezing February Friday was that “first and last pages are the writers’ Achilles’ heel”, where one feels compelled to be extravagant in an effort to impress readers. Damn I can’t remember the name of my teacher/actor/performer/writer. So I forgave Ned for that awful handful of lines. And Knopf. And Inkwell. And Lutyens & Rubinstein. People must be underwater these literary bull market days with so much editing work that they must have missed this.

But until the third page I was soaked with four more drug and techno-related hideous analogies: “The sound system isn’t even that loud but the room’s so small that the treble pushes at the sides like a fat toddler stuffed into a car seat.” Isn’t it the bass that makes things move because of air displacement? “Raf had thought of ecstasy as a substance so synthetic it was almost a pure abstraction”; Jesus! Since when is syntheticity related in any way is abstraction? Is a mathematical theorem synthetic? “To find out that ecstasy – like cocaine, like opium, like marijuana – comes from a plant that grew in the ground is to find out that angels have belly buttons.” Oh, how did you find this out? Can you introduce me to some angel? I am dying to see an angelical belly button.

Finally, in distress, I read “he started to feel a change, but so weakly that he wasn’t even sure, like when you go into a room and you think you can feel a cold draught but no windows are open and it might be just your imagination.” Oh god, take any sharp objects away from me or I will kill myself. From what he wrote I would guess he felt a draught in a room with wide open windows: had he not taken potent new drugs after all

But wasn’t Ned Beauman a Top British Novelist? Hadn’t Granta, the literary Bible, said so? Maybe I was the stupid one, who just didn’t know the latest trend in creative fiction was to ignore all the time-tested style rules.

But the descriptive analogy horror show – and my willingness to plow through – wasn’t over yet. “When the bus finally arrives, its windows are bright like a goods vehicle hauling not flowers to market but bulk protons.” Shit, bulk protons? Has anyone, ever, saw a few protons, as opposed to a lot of them? Doesn’t a light beam carry gazillions of protons?

Not sure. But hey,“and about two dozen soiled earplugs scattered under the bed like the droppings of a hamster that eats only packaging foam.” You know, this was a very specific looking bunch of hamster shit. Sooooooo earplug-looking I couldn’t touch my Beats by Dr. Dre for-a-week.

It wasn’t before I read the first useful analogy of the book - when the author described how the main character suffered from a rare syndrome - that I took the courage to close it and give up. I had entered the Iberia plane, sit on my minuscule coach seat, and the thing was getting ready to take off. “Whatever the cause, the effect is that each morning he slips one more hour out of sync with the rest of the world, as if he’s taking a short westbound flight every day of his life without ever leaving London.” Finally something I could even understand. I was myself going to fly west that afternoon, “eating” five hours out of ten and making up five hours from thin air in the process. Is it really that difficult to describe things without making a mess? To use a couple of interesting analogies, but not overuse them? Neither John Le Carré nor I deserved that torture.

Cheers,

Kiko

In the essay, PG describes Chesky's agony after following Silicon Valley “conventional wisdom”:

“He followed this advice and the results were disastrous. So he had to figure out a better way on his own, which he did partly by studying how Steve Jobs ran Apple.” That conventional wisdom is what PG calls “manager mode”: "hire good people and give them room to do their jobs."

PG then goes on to describe what Chesky incorporated after realizing he was on the wrong track: “founder mode”. The essay doesn't quite describe what it looks like. He points out that it has to do with bold-contrarian moves: “there are things founders can do that managers can't, and not doing them feels wrong to founders, because it is”. In other Chesky talks that came after that one he mentions centralizing some decisions (i.e., around product releases), going into the weeds of things, and reports being inspired by how things were done by Steve Jobs at Apple. [1]

Nobody tried to define founder mode and manager mode well. What we have are highly partial views of these “two systems of management”.

This resonated with many founders. But it also confused them. Nobody tried to define founder mode and manager mode well. What we have are highly partial views of these “two systems of management”, to borrow from Kahneman (Managing, Fast and Slow?).

That's why I wrote this article. I hope it helps.

Management by absence

PG describes manager mode as hiring great people and giving them room to do their jobs. But that's not bad advice. You do want to hire good people and give them space to work. So what I think PG is actually describing as bad is another thing: (i) hiring “great” people and (ii) delegating everything on their domains to them and (iii) never looking back. That's management by absence. That's not good.

The second job of a startup CEO

Curiously, an essay from YC had really resonated with me many years ago, and I think it rightly captures a good chunk of what “good manager mode” is. It's called The Second Job of a Startup CEO, and describes how after finding product market fit (the first job), founders' jobs must change to creating a company to explore the product-market fit created in the first place.

In the essay, which was written by Ali Rowghani, then YC partner and head of its growth investing efforts (YC Continuity), the second job of a startup CEO is to:

1. Hire a leadership team and make sure it works well together

2. Create purpose and alignment

3. Nurture/reinforce the company's culture

Is this bad? No, it's an AMAZING starting point. I think it captures a significant part of what I think “manager mode” should be. All CEOs should do that, and doing it right doesn't mean managing by absence.

Anyway, I'd add a few things to Rowghani's job description to make more whole. I'll call this more complete job description's proper application “systems mode”.

Systems mode

I think “good manager” mode is also specifically about creating management systems and a culture that encourages managers to care both about how things are done and the results the work is causing. Peter Drucker made a bit of a disservice when he articulated management by objectives as the better way to manage executives, because objectives are usually outputs, or results. But many people articulated this balance well.

Vicente Falconi, a Japanese Total Quality Management scholar and consultant, does a great job at teaching how management should care about both inputs and outputs: in a very loose definition edited for simplicity, inputs are measures of how things are done (i.e., the process), and outputs are the actual results of the process.

is it enough to, as a CEO, be a systems designer, and then let the system do its thing? No. Systems have entropy, and even if they hadn't, it's impossible to design human systems perfectly. That's where founder mode comes in, or, better, “right founder mode”.

Jeff Bezos talks a bit about inputs and outputs as well in his 2009 annual letter to shareholders as CEO of Amazon, saying that you can't actually manage outputs: you have to manage (managing meaning assessing and improving) inputs, and if you do it well, the results will come. He says “Senior leaders that are new to Amazon are often surprised by how little time we spend discussing actual financial results or debating projected financial outputs. To be clear, we take these financial outputs seriously, but we believe that focusing our energy on the controllable inputs to our business is the most effective way to maximize financial outputs over time”.

Bill Walsh, the most famous coach of the San Francisco 49ers, puts the same concept into a very catchy phrase: “the score takes care of itself”.

But is it enough to, as a CEO, be a systems designer, and then let the system do its thing? No. Systems have entropy, and even if they hadn't, it's impossible to design human systems perfectly. That's where founder mode comes in, or, better, “right founder mode”.

Founder mode done right: roll-up your sleeves/go see for yourself

Now, the way people interpreted founder mode is also not fair. Operating on founder mode doesn't mean going tyrannical: not listening to your people, making all decisions yourself, refusing to delegate, overlooking hiring a great management team.

Founder is all about selectively doing two things: rolling up your sleeves to do work yourself, and not relying on second (or third or fourth) hand observations of what's going on in your company.

Rolling up your sleeves and doing work yourself is about not losing touch with how the work is being done. Tobi Lutke, for example, reports coding himself whenever he can (which is very infrequently, of course, because he has a company to run). It's the concept of having managers be “player coaches”, as opposed to “coaches only”. I think it's also about setting the example for excellence and commitment.

Now let's talk about going and seeing for yourself, and not relying only on second-hand reports.

The Japanese Total Quality Manangent/Lean people had two interesting concepts [2] that I think can serve as interesting mental models for founders: the first is called “gemba walks” where management (defined as the people with managerial knowledge more than managerial power) would be expected to walk around where the work happens (the gemba, usually the factory floor); the second is called Genchi Genbutsu, which management literature translated as “go and see”, which is more pointed than a gemba walk: you go and see for yourself, firsthand, the problem that's causing the company trouble.

Rolling up your sleeves and doing work yourself is about not losing touch with how the work is being done… It's the concept of having managers be “player coaches”, as opposed to “coaches only”. I think it's also about setting the example for excellence and commitment.

Well, gemba walks and Genchi Genbutsu are two amazing ways for a founder to check on how the system is working (checking the inputs, auditing the process), and helping fix what's going wrong with the system. Elon Musk's Genchi Genbutsu sprints are notorious: many sources including himself describe how during some of the tensest moments of building Tesla's factory in the East Bay he would work himself on getting it running, even sleeping on the factory floor. Elon Musk's extremeness aside, I think Genchi Genbutsu's a good part of what “good founder mode” looks like. It's also a great thing to be built into the culture of the company. Having managers at all levels be expected to go see for themselves and to expect having their managers also go see stuff for themselves and not rely on their reports alone.

But it's not all there is to good founder mode: It's also about the founder being more active in areas where he's the expert. If the founder has differentiated knowledge about some domain within the company (this could be product, or customer needs, or whatever), he has the duty of being more active there, because his differentiated knowledge will produce competitive advantages. Now, of course, that rests on the founder's assessment of his own skills being right.

A better mental model: both, or, in-and-out

Now we have a better grasp of what (good) manager mode and (good) founder mode mean. But which should you pick?

In my humble opinion, the answer is yes.

Founders should definitely build their companies and view them as systems. That means hiring the best possible executives, articulating and enforcing an effective culture, putting together management systems (such as OKRs when well implemented) and processes, and giving their executives room to work. But they should also dive deep when they think something is wrong, or in areas where they think they have a differentiated insight that must be reflected in how the company does things, or even doing regular “gemba walks” when nothing's problematic. A founder that's great at product should be around the major product decisions. A founder that's great at sales should be around the major sales decisions (e.g., the creation of the company's sales processes).

At all times, a great founder should be operating on (good) manager mode. Thinking in systems, building the company, hiring great people, managing inputs and outputs, designing and reinforcing the culture.

But he should add, on top of (good) manager mode, (good) founder mode, at times:

• In all company defining decisions (life or death, success or failure, etc., which are often the decisions PG says nobody else can take)

• In important-but-not-company-defining decisions in areas where he has domain knowledge or skills that are superior to his executives’

• In important areas/decisions where the founder has a spider-sense that things aren't being done right/optimally

• In his regular gemba walks around the company (and it's Slack channels lol)

I think the mental model is both, stacking founder mode on top of manager mode at specific moments and areas of the company.

Notes

[1] Another prominent founder mode type is Elon Musk. He goes notoriously deep in things he deems important.

[2] It's CRAZY how much good stuff there is to be learned from old lean/TQM books. Go read for yourself.

Just as the steam engine sparked the Industrial Revolution, AI threatens to profoundly reshape our economic and social landscapes, albeit far faster.

Recent breakthroughs in AI have transformed AGI from theory to imminent reality. Sam Altman recently remarked, "We are now confident we know how to build AGI as traditionally understood. By 2025, we may witness the first AI agents joining the workforce and materially transforming company outputs."

Companies will spend less on labor. Some of that money will go to compute (data centers, model providers, software applications) and some of that money will go to bottom-lines.

This assertion demands immediate attention. AGI—and eventually ASI—will fundamentally alter our economies, social structures, and the nature of work itself.

I. Understanding AGI and ASI: Defining Our Future

AGI has many definitions. Some say it refers to autonomous systems capable of performing at a level similar to a competent human's. Other's, like those early at OpenAI, defined it as systems that can do any economically valuable task - digitally at first, like a human on a computer; physical afterwards, like a robot in a warehouse - at human performance or better.

Even though AGIs aren't Einsteins, their potency lies in their digital nature: AI systems can be instantly replicated, merged, evolved, and scaled in ways humans simply cannot. Imagine companies duplicating their most capable employees millions of times, flawlessly transferring knowledge, and deploying specialist teams across countless projects simultaneously. It's what Nick Bostrom called the “collective superintelligence”. Which means given a sufficient number of AGI's, the collective can become an ASI by brute force.

ASI, on the other hand, represents intelligence superior to even humanity's greatest minds. Dario Amodei describes ASI as initially comparable to a Nobel laureate but quickly becoming exponentially smarter. Though the exact timeline is uncertain, experts believe progression from AGI to ASI and beyond could occur rapidly, given scale, a scenario dubbed the "fast takeoff", or "singularity”, based on the fact that models will get so good they can speed up the process of developing models in the first place, eventually self-improving autonomously - a goal actively pursued by leading AI labs.

II. The Rapid Path to AGI: Progress and Indicators

The emergence of AGI won’t be marked by singular discrete events but rather by gradual and continuous progress. Systems close to AGI, such as OpenAI's o3 reasoning model—a follow-up to the initial o1 model leveraging test-time compute - are here, and the pace of model releases has accelerated. OpenAI safety researcher Stephen McAleer underscores this urgency, stating recently, "Many researchers at frontier labs seriously consider short timelines to AGI, yet few outside these labs discuss it sufficiently."

These advancements aren’t merely technical feats; they signal a fundamental transformation in organizational productivity. Companies hesitant about adopting AI will struggle to compete against rivals benefiting from AI-driven efficiencies.

III. Work Automation: A New Economic Paradigm

Reasoning models will power agentic software, capable of independent planning, decision-making, and adaptive action. These technologies will radically reshape both white-collar (agentic software) and blue-collar (robotics) work. Tim Urban of Wait But Why anticipates that within two decades, humanoid robots and autonomous drones performing everyday tasks will seem as ordinary as smartphones today.

The first wave of work automation is already happening in customer support. Klarna, a European fintech company, claims that AI manages two-thirds of its support tickets, generating a $40 million profit boost.

But this trend extends to other types of work, such as software development (Devin, Replit Agent, Cursor Agent), pre-sales automation (Apollo, Clay, PatagonAI), legal services (Harvey, Enter, Manor, Igual, Lexter.ai), and many other sectors.

IV. Dramatic Health Improvements and Extended Lifespans

As AGI systems become increasingly intelligent and capable of independent scientific inquiry, their impact on healthcare and longevity research will be transformative. Imagine replicating entire research teams with a few commands, with thousands of the world’s best scientists simultaneously tackling diverse medical challenges. Dario Amodei envisions AI functioning as virtual biologists, revolutionizing biological research by accelerating experimentation and innovation exponentially.

As AGI systems become increasingly intelligent and capable of independent scientific inquiry, their impact on healthcare and longevity research will be transformative. We will devise cures to most diseases, dramatically reducing early deaths, and maybe even revolutionizing longevity. We will all live to be 120.

V. Economic Implication #1: From Wages to Compute

Automation will gradually but long-term drastically reduce company spending on human labor, redirecting some of the saved resources towards technology providers (cloud computing, specialized AI chips, software agents, which I'm bundling into the term “compute”). Not all of it will be redirected, tho. The difference will improve bottom-lines.

For example, in the past, a company had to hire a designer (freelance or full time) or a firm that itself hired a designer (freelance or full time) to create an ad. But in its latest release, OpenAI has a model that does pretty darn good ads (and mind you, this is the worst these ads will ever be, because the models are getting better at a super fast pace):

Will all designers be replaced for now? Of course not. But you can clearly see that there will be dislocation in money from wages to compute. And the same is already starting to happen to different extents in customer support, coding, driving, and increasingly to many other areas of the economy.

Labor dislocations due to AGI (and much longer lifespans) may lead to double-whammy spending pressure for all governments, and thus a constant struggle against ever-increasing deficits from ever increasing welfare payments.

If we extrapolate this migration to physical robots, corporate financial statements will change structurally: opex from human wages will decline and some of these savings will migrate towards capital expenditures for purchasing robots, temporarily, at least, boosting profitability (until we get RaaS: robot-as-a-service).

This productivity revolution poses a paradox: reducing human wages simultaneously shrinks consumer purchasing power, potentially threatening economic stability unless new economic frameworks emerge. That leads us to economic implication #2.

VI. Economic Implication #2: Less Disposable Income, More Leisure (and Expanded Welfare)

Mass automation will initially decrease aggregate wages, and thus wage incomes, reducing aggregate disposable income, yet increasing leisure time. This shift might temporarily boost sectors like entertainment, education, and travel. However, without income redistribution, lowered consumer demand could dampen economic growth, with “no one left to buy stuff”.

Governments will need to proactively mitigate unemployment through mechanisms like Universal Basic Income (UBI), as traditional unemployment schemes may not cope with widespread structural displacement. And given that we'll probably live much longer (see IV. Dramatic Health Improvements and Extended Lifespans), these welfare payments will have to go on for much longer.

This may lead to spending pressure for all governments, and thus a constant struggle against ever-increasing deficits from ever increasing welfare payments.

VII. Aside - Economics of Abundance: Rethinking Monetary Policy?

AI-driven abundance could fundamentally alter traditional economic constraints, particularly inflation driven by government spending. Historically, increased spending risks inflation when production capacity is limited and marginal costs are significant. However, AI technologies drastically reduce marginal costs, approaching near-zero for digital and automated products. In such a scenario, traditional monetary policy designed to manage scarcity may become obsolete.

Efficiency from automation is going to be highly deflationary. Would that make room to accommodate printing of money without moderation?

Instead, governments could shift towards managing abundance. Economic growth might become predominantly demand-driven, with AI systems dynamically responding to increased consumer purchasing power. Rising government expenditures, including UBI, would act less as inflationary threats and more as signals instructing AI-driven economies to ramp up production immediately. Thus, monetary policy might evolve from managing inflation and scarcity toward managing equitable distribution and incentivizing sustainable, abundant production.

VIII. The Challenge(s) Ahead

People, companies, and governments need to actively consider what's coming, and prepare for it. What happens when everybody lives to 120 and beyond? When most diseases are cured? When most jobs are automated? These are very hard questions that deal with a scenario that may seem far away but IS NOT.

Every early stage VC wants to be a top-quartile performer. Empirically, top-quartile performers have had at least one company in their vintages that has returned the size of the fund in proceeds. That means that on selling its stake in the company, the fund gets back, from this one sale, as much as it raised from investors.

So if the fund has raised $300m from investors, it has to get back $300m from the sale of its top company or companies (each). And if the fund ends up owning, at the moment of the sale, 15% of the company, the equity value of the company has to be $300m / 15% = $2b. Two billion dollars is the valuation a $300m fund is looking for at exit.

Of course, that math depends heavily on (i) stake at sale (post-dilution) and (ii) fund size. And stake at sale depends on iii. stake pre-dilution.

For a Series A investment, funds model something from 25% to 50% dilution, meaning they buy, for example, a 20% stake in the company and end up, at exit, owning just 10-15%, because other, later-stage investors will have bought in and diluted their ownership stake. (This ignores subsequent investments, e.g. pro-ratas, which make the math a bit more complicated but that do not fundamentally nor directionally change the VC equation). With more competition amongst VCs for top startups, it is not that common for a fund to buy a 20% stake in a startup anymore. So at-sale stakes can easily run below 10% if no subsequent investments are made.

The other key variable is fund size. VCs are incentivized to grow their assets under management, because they make a management fee (something around 2%) on these assets. That means the average VC fund has been growing in size. Et ceteris paribus, a $100m fund means a fund-returner valued at $670m, whereas a $500m fund means a fund-returner valued at $3.3bn. Benchmark Capital has been really disciplined and kept its funds at around $450m raised. Andreessen Horowitz has a $600m fund only focused on American Dynamism, and $42b in assets under management. $500m at a 10% stake means a $5b valuation at exit.

The underlying point I'm trying to make is that your startup has to at least have the potential to be a fund returner, and that certainly means at the very least, for a small fund, unicorn valuation. If not, the math just doesn't add up for the VC, and you won't be able to raise the money needed to play the game.

So next time you're thinking about your startup, or pitching VCs (the preparing for which is a great way to think about your startup in a structured way), ask yourself whether what you have is a fund returner.

Thanks to the YC Brazil community for feedback on this.

Founding and running a startup [1] is kind of a game or discipline of its own when compared to founding and running other types of companies (e.g., consulting businesses [2], restaurants, travel agencies, industries, etc.), just like running 100m sprints at the Olympics is a game or discipline of its own and very different from the 1000m, the 4000m, and of course the Marathon.

The game of startups [3]

So, what's really the game of startups?

The game of startups is not just the game of starting new companies. Startups are special kinds of new companies designed to grow really fast. The game of startups is the game of building the biggest companies in the shortest time possible [4]. Empirically, the fastest-growing businesses you can build tend to heavily involve some form of technology.

How much growth is high-performance growth?

If the game of startups is the game of building the biggest companies in the shortest amount of time, understanding the game of startups means understanding the levels of growth that allow for front-of-the-pack performance.

Back in the day, we used to talk, at least for b2b startups, of the T2D3 model, which meant the fastest growing cos would, after reaching one million in ARR, triple, then triple, then double, then double, then double their revenues, from $1m to $3m to $9m to $18m to $36m to $64m in 5 years after the first milestone.

Nowadays, T2D3 is not enough. AI startups are growing at even faster rates, even if some of the revenue is curiosity revenue [6]. We are seeing companies reach $10m to $20m at neck breaking speeds.

Does it mean all startups in the game will follow that trajectory? No. But it means that's the benchmark against which you'll be compared, especially by VCs, which brings us to…

Startups and venture capital

Venture capital is tightly associated with the game of startups but is not a requirement per se. One could, theoretically, play the startup game without it. It's just that in order to win, chances are that one will move faster if one uses the help of venture capital. Or one will be forced to use venture capital by the competition. It helps to think about Olympic swimming: fancy wetsuits are not a requirement per se of doing the freestyle 50m, but just about every successful swimmer uses them to some extent, just like just about every successful startup uses venture capital to some extent.

The game changes as the company grows

One specificity about the greater game of building and running companies is that the player playing the game usually changes modalities as the journey progresses.

From founding to pre-IPO, for example, the game may be indeed to build as big a business as possible quickly; from pre-IPO to IPO and beyond, the focus of the game changes slightly to building the most valuable business, even if that means sacrificing growth a bit; as time progresses, a founder may switch gears and play the game of trying to make the company endure the test of time. We could even argue that the best founders are the ones that correctly time these changes and make them successfully (e.g., founders of Airbnb, Facebook, Stripe, etc.)

Seeking the best advice

Seeking (and giving) advice that's good requires one to understand what game or discipline they are seeking advice for, and as important, what disciplines the people you ask is qualified to give advice on.

Notes:

[1] For a great defintion of what a startup is, read startup = growth, by PG. TL;DR, a startup is a company that's designed to become very big very fast.

[2] Whatever that means.

[3] Let's just recursively define "startups'' as the types of companies one tries to build when one plays the sport of startups.

[4] By the way, the game of startups is not really to build the biggest companies in the shortest possible time, but to build the most valuable companies in the shortest possible time. It's just that in the early stages of a company, size [5] is probably the best proxy for value.

[5] We could also discuss what "size" means. Usually, it's revenue, but sometimes some other thing will be measured, such as the number of users a service such as a social network has, if the people playing the game think that users are a reliable indication of future revenue and then value.

[6] Curiosity revenue is revenue from customers trying out a product without really a high-quality commitment to using it “in production”. Its implication is that it is followed by high-churn.

It got me thinking.

The first thing that came to me is that ~all startups fail (i.e., go out of business) because of a lack of product-market fit. When there is real product-market fit, all problems are solvable, because maintaining strong growth tends to feel *more* like chasing a boulder down the hill; maintaining growth with weak PMF (or lack thereof), on the other hand, feels *more* like pushing said boulder up the hill.

If we take “fail” in a more latu sensu, why so many startups fail to be huge success stories, such as unicorns, decacorns, etc., when they've survived that first stage with some form of PMF? Probably because they picked a small market, or a big market with terrible dynamics (e.g., airlines).

In any way, where that thinking got me to is that so much of the success of a startup is a function of the product/market that founders decide to tackle. And to get back to the initial tweet, there's not a lot of content out there on how to pick a good market.

Moreover, most content about the earliest stages of the startup journey tend to be somewhat misleading.

Most founding stories, on the one hand, tend to romanticize how founders first get to pick their ideas (and that's understandable, bc these stories are geared toward current/prospective customers and employees, not other founders). David Velez, founder of Nubank, tells the story of how he decided to work on Nubank after being stuck in an armored revolving door when struggling to open his checking account in a bank branch. That moment might have had an impact on the trajectory, but truth be told, David was a highly sophisticated founder, keenly aware of the ingredients of Silicon Valley-type successes. He had worked both at General Atlantic, a famed tech growth investing firm, and Sequoia, which needs no intro. He had scoured Latam in search of venture-scale businesses to invest in on behalf of Sequoia, frequently in the company of Doug Leone, the legendary Sequoia managing partner. If there is one founder that was well-informed on what it took to make it big, it was him. That explains why he picked an industry that was incredibly large in Brazil - banking - with an oligopoly of a few tech-backwards banks that were stupidly profitable. Or why he picked an industry where customers were being treated poorly, and had to jump through whoops to go over internal limitations of these oligopolists, such as their inability to open up a checking account without having the poor customer go over to a bank branch. Or why he picked an industry regulated by an vanguardist Central Bank that was giving lot's of indications that it wanted more competition. In other words, David thought A LOT about what and where to start up.

Most founding advice, on the other hand, tend to downplay the role of picking a good starting point, or being very judicious about where to pivot to. Paul Graham, for example, says about finding out How to Do Great Work:

The first step is to decide what to work on. The work you choose needs to have three qualities: it has to be something you have a natural aptitude for, that you have a deep interest in, and that offers scope to do great work.

In practice you don't have to worry much about the third criterion. Ambitious people are if anything already too conservative about it. So all you need to do is find something you have an aptitude for and great interest in.

Why, if David Velez had taken this to heart, he would probably have started some b2b software company to help PE or VC investors make better decisions, or some other b2b SaaS product to help entrepreneurs like his father better manage their businesses. But he didn't, and was very judicious about what and where to start. I call this top-down thinking.

Truth is, most entrepreneurs are terrible at picking high-potential businesses to start with. When they do, it's often by chance. They tend to pick where to work on based on path-dependency from decisions made in the past (like what college to go to, or which industries/sectors to start their careers on) or from hunches. I, for one (and I mean it as a *bad* example), started Qulture.Rocks because I had been interested in the people management practices of the 3G guys, which lead me to write a book about them, which led me to try to code in software what they did, and the rest is history. Most of my fellow YC batch mates similarly founded their companies out of random paths, with very little consideration to top-down thinking.

Now, is top-down thinking alone enough? Of course not. But it is INCREDIBLY important.

So what does a good product/market combination look like out there?

Sequoia Capital has this little-known web page I first came in contact with circa 2012 called Elements of Enduring Companies. Here's a screenshot, bc god knows if they'll delete it any day.

Let's build a list of traits for markets/products that can produce era-defining companies:

• Large market of customers

• who have strong, frequent pain

Large market is the first and foremost. First, we can broadly define a market as a collection of customers who have a similar job to be done. A large market, per that definition, is one where there are a lot of customers (lots of people, lots of companies) with a similar job to be done your startup is going to solve.

But lots of customers aren't enough. The job to be done you are going to solve has to be the source of lots of pain, and that pain has to be frequent. Solving once-in-a-lifetime intense pain is nice, but solving frequently-experienced intense pain is much better.

PG and Michael Seibel have an interesting heuristic that turns the problem upside down: a hair on fire problem is a hair on fire problem when even a half baked solution is better than nothing in the eyes of customers. Therefore, putting an MVP out there in front of a market is a good litmus test of whether the pain is big enough.

But strong pain is *still* not enough.

The customers in this market have to ideally have money to spend on your solution. I recently heard an interview with the founder of Duolingo. He was talking about how Android app users make up the vast majority of their user base, but iOS app users make up the vast majority of their revenue. That's because Apple phone owners are richer, have more money to spend, and spend more money. In b2b, selling to the CFO (rich, bc is the owner of the budget) is better than selling to the analyst (poor, bc has no budget at all). Selling to the CFO (rich) is better than selling to the CHRO (poor, for many reasons that are not the scope of this post).

We looked a lot at the demand side of the market question, but the supply side of the market question is also super important. (From the demand side, a market is a collection of customers with similar jobs to be done. From the supply side, a market is a collection of companies trying to solve these jobs to be done.) There are markets where all the conditions above are true, but companies in it are still not Apples or Amazons or Googles or even Nubanks, for that matter. Airlines are the prototypical example. Bc of many factors, all profits in the airline industries get, to use Michael Porter's words, competed away. The best framework to analyze a market from the supply-side is the Five Forces Model, from Porter. The stronger the forces acting upon companies in a market, the less profitable they tend to be. The forces are the bargaining power of suppliers, the bargaining power of customers, the intensity of rivalry amongst competitors, the threat of new entrants and the threat of substitute products. I won't get in depth about each, you should read it up, but in the airlines market, some of these forces are so strong that profits are erased and everybody loses money.

So our list is, I think, now complete:

On the demand side:

• large market

• with rich customers

• facing intense, frequent pain

On the supply side:

• favorable competitive dynamics (as articulated by the 5 Forces)

There are other angles we could explore, but I think these are already plenty.

Original tweet: https://x.com/kiko_himself/status/1865047702137933825

In this article I'd like to discuss some concepts from Total Quality Control (or TQC or Total Quality Management or TQM) that I think are very well applicable to the world of b2b software sales. Some of these concepts are in fact applicable to all areas of management, but in this article I'll focus on sales, a great interest of mine since I've founded and became the self-appointed CEO of Qulture.Rocks almost 9 years ago.

The collision of two seemingly very different worlds

Total Quality Control is a management philosophy that blossomed in Japan in the wake of WWII. American management experts such as W. Edwards Deming were sent there by the US Government in an effort to help the Japanese rebuild after the war (here, here).

The Japanese learned and perfected what the yankees had to teach, so much so that Japanese firms began beating their American competitors on their own turf, the U.S. market. One of the most notable markets where this beating took place was semiconductors, the then-nascent industry based out of the San Francisco Bay area. Intel and other American memory makers had a hard time keeping up with their eastern counterparts like Hitachi, Fujitsu and Nippon Electric. (Another famous market was automobiles, with the rise of Toyota and later other makes.)

As a response to that threat, many American firms adopted TQC.

Intel itself became a major TQC case. It implemented it among myriad other initiatives including abandoning the production and sale of memory chips in favor of that of microprocessors. TQC became the company's backbone. Other key TQC practices adopted by Intel were quality circles and TQC's goal management framework, Hoshin Kanri, named internally "iMBO," or Intel Management by Objectives (more about the origins of Hoshin Kanri, MBO, OKRs and the sort here).

TQC in a nutshell

TQC is based on a holistic view of the firm. Quality means addressing the needs of "customers," and total quality means addressing the needs of all of the firm's stakeholders: customers, of course, but also employees, shareholders, and society at large.

Controlling quality starts with the definition of lagging quality indicators and leading quality indicators. Controlling means keeping such indicators within acceptable ranges or, when needed, scientifically working to improve them, in an ongoing cycle called PDCA, or Plan, Do, Check, Act.

This logic doesn't stop at the company's top-level, foremost stakeholder quality indicators (be them shareholder-centered (EVA, share price performance, EBITDA, net income, etc.), customer-centered (NPS, churn, retention, avg. ticket, etc.), employee-centered (turnover, eNPS, etc.) or society-centered (ESG stuff mainly)); top-level indicators must be unfolded throughout the organization, so that every important indicator is controlled, and so that every person within the company works towards maintaining or improving a subset of them.

The logic of how these indicators (you can call them metrics, KPIs, whatever) are unfolded is the logic of the business and the metrics themselves. For example, the board may decide that the main shareholder-centered indicator is going to be net income. Net income is composed of sales, costs, expenses and taxes, as well as other non-operating revenues or expenditures. In b2b SaaS, sales can be further broken down into sales from existing customers and sales from new customers. Sales from new customers can then be further broken down into number of deals closed, average ticket per deal, and date of closing. These numbers can be also broken down by customer segment, territory and by the organizational structure of the team (VP A, VP B, VP C, and then further down until you reach the reps themselves). This is of course not an exhaustive unfolding of sales, but it should give you a good idea of what we're talking about.

Anyway, since this is not supposed to be a treatise about TQC, I'll stop the intro here and dive into the concepts.

The importance of standardization

A core tenet of TQC is standardizing how things are done in the company. Standardization has two aspects: on the one hand, it means executing the latest version of a process in the same (best) way every time. On the other hand, getting consistent results out of the process.

For example, if an AE follows the sales process correctly and consistently, she should hit quota (say, plus or minus 10%). Also, all AEs that a) are equally skilled/experienced, b) sell the same products and c) to the same customers should follow the same process and get very similar results.

Why, you ask, is standardization important? The answer is simple: if the latest (and best) process isn't being followed, we can't make tweaks to said process in an organized manner so as to get better results in a predictable, consistent way.

We'll be throwing darts blindfolded.

If the process is being followed, we can make experimental changes to parts of it and check if they improved results or not. If not, we can go back to the original process, or try something else, knowing that going back to the standard process takes results back to where they were originally expected to be.

Standardization is a core principle of TQC because TQC is all about creating a predictable, consistent system for improving results. And the only way to predictably and consistently improve results is to do it in an organized, "scientific", if you will, way.

Two important consequences follow from these two two aspects: the first is that if the AE doesn't follow the process, that's her fault and she should be held accountable. The second is that if the AE does follow the process (again, correctly and consistently), bad results are not her fault, and she should not be accountable for them.

These two consequences take us to our next concept, which is the relationship between authority and accountability.

Authority and accountability

This concept is pretty simple: Employees should only be held accountable for results if they have authority over the processes that generate said results. It follows, logically, that employees should not be held accountable for results if they aren't allowed to create and change the processes that generate said results (nor be expected to create and change processes if they don't have the experience and skills to do so effectively).

This is critical.

We've said before that if an AE follows the sales process correctly and consistently, she should hit quota (say, plus or minus 10%). If she does follow the process and doesn't get the expected results, she shouldn't be held accountable. The actual person who has authority over the process should be held accountable.

In the next topic, "the general job ladder", will tie this principle to how organizations are structured.

The general job ladder

TQC experts (e.g., Nemoto, 1987; Falconi Campos, XXXX) have created an amazing framework that divides work at a company in a vertical progression with four buckets of jobs with decreasingly complex functions as they relate to TQC. The four buckets are:

• Executive jobs,

• Middle-management jobs,

• Supervisory jobs, and

• Operational jobs.

The four buckets can be illustratively superposed over an "org chart" pyramid, as show in the following drawing.

Executive jobs

At the top of the org, we have the executive jobs.

Executive jobs are in great part composed of translation functions: getting a high-level strategy from the board (where the CEO also usually sits) and translating, with the CEO, that high-level strategy into top-level organizational goals. Then, translating top-level organizational goals into more cross-functional goals (think NPS) and more functional goals (think "1 month retention for product A" or "% avg. quota attainment"). (I already outlined how goals unfold or cascade throughout the top of the org in the TQC in a nutshell section.)

Middle-management jobs

Next come middle-management jobs.

Contrary to popular lore, it's where very interesting work - arguably the most interesting work from a TQC standpoint - happens.

Middle managers own the processes that are executed by supervisory and operational functions. Since they have direct authority over the creation and change of these processes, they also have responsibility over the results these processes produce. They receive ambitious goals from the executive functions and are accountable to tweak processes (or even create new ones) so as to achieve them.

Supervisory jobs

Next, come supervisory functions, who are very close to who's actually executing most processes within an organization.

There are two key supervisory functions: the first is training the people who execute the processes. The second is regularly auditing how correctly and consistently the processes are being executed (ironically, both training and auditing are processes themselves). In doing so, supervisors might get external help from sales enablement and/or sales ops if they need and depending on the completeness of the sales organization.

Supervisors don't own - and shouldn't own - processes. Supervisors should be promoted into their functions after a) having been performed well in operational jobs and b) having a nack for leadership and people management. They don't have the necessary experience to have authority over processes, and therefore should have limited responsibility/accountability over results. They must be held accountable for process execution, and only that.

Supervisors should also gather suggestions made by those holding operational jobs and take them to middle managers, so they might be incorporated into processes.

Now, of course, if someone performing a middle-management function is redesigning a process in hopes of improving its results, they will confer with those in charge of supervisory functions who are close to the action and can provide valuable input and feedback. There is no point in a middle-manager trying out a change in the sales script if AEs for some reason can't follow the changes.

Operational jobs

Last, but not least, come operational functions. Operational functions are those comprised of executing established processes, period. They are what is usually equated to "actually doing the work". People holding operational jobs should also be asked to contribute ideas to improve their work. These are not big process changes, but small tweaks. For example, an AE may suggest a new way to close calls that produces clearer next steps, or that others read this or that article. In traditional TQC, operational people participate in Quality Circles, or QCs.

Every corporate career starts in performing operational functions. Nonetheless, operational work is not the same thing as blue-collar work. This is a common misconception. At the base of the office pyramid, white-collar workers also start performing operational functions.

As AEs do their jobs well - following processes perfectly - they might get promoted from junior to senior. That should bring about increasing productivity (e.g., more MRR per quarter). They can then go tackle more complex sales processes (e.g., from mid-market to enterprise) or tackle a supervisory job (e.g., managing other AEs).

Jobs at different org. maturities

Mature sales orgs should have roughly four hierarchical levels: salespeople (e.g. AEs, SDRs, BDRs, or even support analysts in enablement and ops), sales supervisors (who oversee a number of salespeople), sales managers (who oversee a number of supervisors) and sales executives, who oversee a small number of managers. Some of the actual job titles - the names - may vary from org to org: supervisors might be called managers; managers might be called managers or directors (or even VPs at very large orgs); executives might be called VPs or chief officers.

Some professionals at smaller orgs might accumulate supervisory, middle-management and even executive functions: a nascent sales org at a pre-series A startup might be composed of AEs and a VP Sales, with the VP sales doing all the non-operational work. Between such small nascent orgs and mature orgs are a large number of different setups.

Putting TQC and sales together: implications

You might have intuited a few consequences of these concepts as they apply to b2b software sales organizations. If not, I'll help you out, for it's really useful to look at a sales org through the lens of TQC, standardization, authority x responsibility and the general job progression.

• Again, for the sake of repetition: AEs should be accountable for following the process correctly and consistently; Supervisors should be accountable for training AEs on following the process (correctly) and audit process adherence (consistency); management should be accountable for results; executives should be accountable for the impact results have on strategy and company success.

• If an AE follows the sales process correctly and consistently, she should hit her quota (say, plus or minus 10%). Also, all AEs that a) are equally skilled/experienced, b) sell the same products and c) to the same customers should follow the same process and get the same results.

• Since AEs have no authority over the sales process they follow, it should be obvious that we should be very careful in measuring AE performance by the results they achieve: the core performance management metric for AEs should be their individual adherence to the sales process. AEs should follow the processes established by middle-managers, and get the predicted results out of them, i.e., AEs that follow the process perfectly should hit their quotas consistently over time, and all AEs should be generally hitting their quotas.

• Conversely, if an AE is following the process (as measured by her supervisor and/or enablement) but not getting results, managers - not supervisors - should be held accountable, because it's their process. (Exceptions should be due to bad hires (which shouldn't happen if recruiters and intverviewers are following their processes properly) or localized anomalies, such as territories being affected by some external, unpredictable factor.)

• As we saw, the core function of supervisors or frontline managers is to a) train AEs on their jobs (i.e., to train them so they are able to follow the processes perfectly) and audit AE adherence to said processes. In doing so, supervisors might get external help from enablement if they don't have the time necessary to perform all their functions.

• The core performance management metric for a supervisor should be the aggregate adherence of their direct reports - AEs - to the sales process. They should at the one hand, at a high level, monitor activity KPIs that indicate adherence to the process, such as the number of calls, demos, meetings made, proposals sent, etc., and on the other hand, at a lower level, listen in on a sample of calls, demos and meetings and read a sample of emails, proposals etc. to ensure adherence.

An important final observation (not wanting to get recursive, but getting a bit recursive nonetheless): supervisors should follow established processes in doing their work. There should be a standard way to sample work, to set shadowing frequencies, and so on.

Anyway, Zuck ended up buying Instagram, and that put him at ease re: The Twitter Threat, which was probably non-zero going forward, but still there running in his mind's background. To make his psychological life easier, Twitter was moribund for a long time, shipping features at a turtleish pace and showing no competitive aggressiveness. "Let it be," thought the cyborg.

With Elon Musk now at the driver's seat, Twitter is making aggressive moves (for example, it now allows for the posting of long-form video, which is an attack into YouTube's garden), Mark is most definitely more afraid of The Twitter Threat, a potential entering proper into photos and short-form video.

Enter the move he made with Threads. Starting up a Twitter "clone" is actually just a boring part of the picture. The most interesting part is the fact that Threads is going to be "open". As per the company's post introducing the product:

"We’re working on Threads soon being compatible with the open, interoperable social networks that we believe can shape the future of the internet"

and

"Soon, we are planning to make Threads compatible with ActivityPub, the open social networking protocol established by the World Wide Web Consortium (W3C), the body responsible for the open standards that power the modern web. This would make Threads interoperable with other apps that also support the ActivityPub protocol, such as Mastodon and WordPress – allowing new types of connections that are simply not possible on most social apps today. Other platforms including Tumblr have shared plans to support the ActivityPub protocol in the future."

This is really interesting. There is a long history of technology companies using open-sourcing-as-a-tool-to-commoditization as a competitive tool, both to kill potential competitive threats or to boost product complements.

An early mention of the theme happens in Hal Varian's amazing Information Rules:

Traditional rules of competitive strategy focus on competitors, suppliers, and customers. In the information economy, companies selling complementary components, or complementors, are equally important. When you are selling one component of a system, you can’t compete if you’re not compatible with the rest of the system. Many of our strategic principles are specifically designed to help companies selling one component of an information system.

The dependence of information technology on systems means that firms must focus not only on their competitors but also on their collaborators. Forming alliances, cultivating partners, and ensuring compatibility (or lack of compatibility!) are critical business decisions…The history of the Microsoft-Intel partnership is a classic example. Microsoft focused almost exclusively on software, while Intel focused almost exclusively on hardware. They each made numerous strategic alliances and acquisitions that built on their strengths. The key for each company has been to commoditize complementary products without eroding the value of its own core strengths. For example, Intel has entered new product spaces such as chipsets and motherboards to improve the performance of these components and thereby stimulate demand for its core product: microprocessors. Intel has helped to create a highly competitive industry in component parts such as video cards, sound cards, and hard drives as well as in the assembly and distribution of personal computers.

Microsoft has its following of independent software vendors (ISVs), and both companies have extensive licensing programs with original equipment manufacturers (OEMs). And they each have each other, an extraordinarily productive, if necessarily tense, marriage. It’s in the interest of each company to create multiple sources for its partner’s piece of the system but to prevent the emergence of a strong rival for its own piece. This tension arises over and over again in the information technology sector; Microsoft and Intel are merely the most visible, and profitable, example of the complex dynamics that arise in assembling information systems.

Joel Spolsky was another one to tackle the subject:

"A complement is a product that you usually buy together with another product. Gas and cars are complements. Computer hardware is a classic complement of computer operating systems. And babysitters are a complement of dinner at fine restaurants... All else being equal, demand for a product increases when the prices of its complements decrease... When computers become cheaper, more people buy them, and they all need operating systems, so demand for operating systems goes up, which means the price of operating systems can go up... In general, a company’s strategic interest is going to be to get the price of their complements as low as possible. The lowest theoretically sustainable price would be the “commodity price” — the price that arises when you have a bunch of competitors offering indistinguishable goods."

A key quote in this paragraph I glued together is "The lowest theoretically sustainable price would be the ‘commodity price’ — the price that arises when you have a bunch of competitors offering indistinguishable goods." That's why companies tend to try to commoditize their complements (in Spolsky's case by making the complement open-source) in order to make sell more of its core product.

Another potential move that's very related is to commoditize a piece of your value system upstream or downstream (for simplicity purposes, customers or suppliers) so as to make them less individually bargaining-powerful, thus being able to capture more of the value system's [1] profit pool.

I took this little detour to give some color on how I think Meta's launching of Threads with an "open" flavor is related to this whole idea maze of open/commoditization/competition.

By boosting ActivityPub adoption (making Threads compatible) Meta seems to be trying to weaken Twitter and its closed protocol, so as to reduce the chances that Twitter might become a potential future competitor at Meta's most profitable turf, which is Instagram (feed, reels and stories, which faces an already steep bunch of competitors, esp. TikTok). By making Threads open to Mastodon and BlueSky content (and vice versa), Threads is boosting the strength and attractiveness of the trifecta's joint network as an alternative to Twitter's very strong one. If it succeeds, it might even push Twitter to also adopt the open standard. And the more the content is interoperable between networks, the less relevant the "client" becomes. [2]

Notes:

[1] I use Michael Porter's "value system" x "value chain" terminology instead of the more usual use of the terms, which ignores the term "value system" and just uses "value chain" when describing the processes/players/steps upstream and downstream of a company. For Porter, "value chain" is, for simplicity's sake, the inner workings of how a company creates value, from the walls in, not out.

[2] Reminds me of the Tesla x Edison fight for standardizing AC x DC.

We are still in the earliest innings of understanding how these new technologies will affect enterprise software. Nonetheless, I'd like to take a stab and make some predictions and provoke a few avenues of further investigation.

I'm especially interested in looking at enterprise software market's competitive dynamics as it relates to the ongoing fight between incumbents and new entrants. We can dive deeper through the following angles: a) understanding how these said advancements will affect the disruption dynamics that are very present in the enterprise software space and b) understanding how access to data will affect who wins a given battle between incumbents and new-entrants. I also propose a few questions that may help us assess the potential effects of AI in a company's future prospects.

The Never-Ending Cycle of Disruptive Innovation in Enterprise Software

Enterprise software applications, from the bedrock (CRMs, ERPs, HCM suites such as those made by Salesforce, Oracle, et al.,) up to the more mission-oriented systems (such as those made by early Hubspot, Segment, Resend) are subject to Never-Ending Cycle of Disruptive Innovation in Enterprise Software, a textbook application of Clayton Christensen's Disruptive Innovation theory.

In most markets, customers' abilities to absorb (i.e., use and value to the point of paying more for) product improvements grows at a slower pace than the product improves [1]. This causes segment after segment of customers (from least needy to most needy) to at some point become over-served.

As I wrote in the above-linked article [2]:

Enterprise software is an incredibly fertile ground for disruptive innovation. Let’s take the example of Salesforce to illustrate that assertion.

Salesforce’s CRM offering goes much further than what a “simple” CRM is expected to do: aside from storing prospective customer information and allowing for a pipeline of deals to be managed, it has all sorts of bells and whistles, such as industry-specific reports, complex workflow automation, AI [3], an app store, and LOTS of flexibility that cater to all different sophisticated customer needs; it costs a lot and takes long to be implemented. And Salesforce keeps adding incremental sustaining innovations to the product every year and causing a bigger and bigger portion of the market to become over-served.

In Christensen's theory, his over-serving happens because incumbents tend to focus on the needs of their highest-end customers - i.e., the customer segments that are willing to pay the most - when prioritizing their roadmaps. The highest-enders not only pay more but also demand more: they are the most demanding in their needs for product improvements, generally along the lines of more and more bending of the software to their specific idiosyncrasies. That inevitably creates more complexity and thus makes the software harder to use - and eventually more expensive - for all the lower-than-the-highest-end customers, who'd be increasingly happy, if we were to follow along the highest-to-lowest-sorted list of customers, with a simpler, easier to implement, cheaper solution if only one were available.

In sum, we could think of it all like this: With every new feature Salesforce introduces (i.e., a product “improvement” through the eyes of, e.g., Nike, which is happy to pay more for those “improvements”), a new cohort of customers (e.g. some hypothetical John Doe & Sons agricultural implement retailer from Iowa) falls over into the over served bucket, because it gets incrementally frustrated with the pricing and/or complexity of the product, and thus becomes fertile ground for Pipedrive's entrance into the market.

Prediction I: LLMs will delay the onset of disruption, favoring incumbents

So how, if at all, will evolving AI infrastructure (LLMs becoming ubiquitous, available via easy to use APIs, etc. etc.) affect this cycle?

A new layer of AI capabilities (e.g., Microsoft's Copilot in case of Microsoft Office customers) is a sustaining innovation for incumbent enterprise software offerings; that is, one which improves the product in a dimension important to said company's most demanding customers.

These dimensions are things like flexibility of customization, convenience of adoption, ease of use, etc.

My tentative thesis is that AI will add a new vector of product characteristics aligned with the current vectors in place, so as to in practice bump up every customer segment's innovation adoption curve and thus delay the onset of "overservedness" along this line (the highest-to-lowest-end), therefore delaying the inevitable onset of disruption-by-disruptive-innovation by potential new entrants. AI could slow the cycle of disruption by introducing a new dose of functionality | convenience | customization that's relevant to these customers. No large feat of imagination is necessary to post that AI could help products gain new functionality.

If that's the case, products that were about to be disrupted could gain new leeway. From another angle, segments of customers that were about to become over-served could take longer to become so.

This could mean AI is potentially disproportionately advantageous to incumbents as opposed to new entrants. But not all incumbents benefit equally.

Data seems to be key

Another interesting angle to look at this problem is data. Many argue that AI models are going to become commoditized. If that's so, the logical conclusion is that those who hold differentiated data corpora will be the able to leverage baseline models and offer AI functionality that's differentiated.

For an enterprise (customer) the most differentiated data corpus is its own (sitting on Workday, Salesforce, Workday, Oracle, SAP, Hubspot, etc.). Therefore, those same business applications holding proprietary company data are best positioned to benefit.

Oracle's ERP, for example, has a company's history of financial data, and most probably its competitors'. It could build on top of open-source LLMs by training them on company-specific data, therefore being able to offer highly unique AI solutions to its customers. No rocket science here.

They could, thus, offer AI capabilities that are hard for anyone else to replicate. The question then becomes “will AI capabilities become key for solving customers’ jobs-to-be-done?” That's still an open question.

Other types of data are also important, although a bit less. One type is industry-specific data (for example, that held by Bloomberg). There're players that house both proprietary and industry-specific data, such as Workday's Peakon product - and Qulture.Rocks’ climate/engagement surveys product - which allow customers to measure their own historic employee engagement data while also comparing said data to that of an anonymized group of relevant industry peers’).

Delayed onset - but onset nonetheless

Even though I think there's a significant chance that AI will delay the onset of disruption for most enterprise software incumbents because, in part, data, I still think the Never-Ending Cycle will go on for the foreseeable future.

On the shorter term, the most over-served customers are still very unhappy, and will jump to simpler, cheaper solutions if they become available, even if their AI capabilities are still nascent.

On the longer term, it will most likely happen at scale when new entrants, especially startups, figure out an AI first way to leverage these new capabilities. A parallel is ads [4]: when the first newspapers got online, they served ads on a webpage just like they served ads in print: banners that would be placed on top of or besides articles. Then mobile happened, and Facebook figured out an entirely new way to place ads in front of customers: along the news feed. A new paradigm that was optimized for the new reality of mobile. The same will eventually happen to AI and enterprise software: today we're layering AI on top of an 'old' software paradigm. What will the AI-first enterprise software paradigm be? Conversational?

A framework to evaluate the impact of AI

Building on top of the work of Sills, B. et al (2023) [5], I've given their framework my own spin. For every positive-effect question, add the stated number of points to the company's scorecard. Alternatively, for every negative-effect question, subtract the stated number of points. Add the points in the end for an overall AI-impact score, the greater the more positively affected the company should be.

Positive-effect questions:

• Does the company's products benefit from network effects, especially in relation to data? (+ 4 points)

• Does the company's products house an important corpus of customers' own data? (+ 2 points)

• Does the company's products house an important corpus of customer's industry-specific data? (+ 2 points)

• Can AI capabilities expand the customer's TAM, or accelerate TAM penetration? (+ 1 point)

• Do the use cases and end-customer bases of the company’s offerings have more stringent requirements in terms of data governance and compliance? (+ 1 point)

Negative-effect questions:

• Does AI have the potential to reduce the company’s total addressable market via commoditization? (- 4 points)

• Does AI have the potential to reduce the company’s total addressable market via seat count compression? (- 3 points)

Notes

[1] According to Christensen, product and service improvements come from three core dimensions: functionality, reliability, and ease of use (ease of use being itself composed of convenience, customization and price).

[2] The core of the article can be summarized by the following sonnet, written by our friend ChatGPT:

In enterprise domains, the giants stand tall,

Their products refined, answering the call.

Sustaining innovations fuel their might,

While in their shadow, disruptors take flight.

A simpler, cheaper product takes the stage,

For overshot customers, it's all the rage.

The giants, they scoff, ignore this new threat,

In high-end markets, their sights are firmly set.

But the disruptor, it does not stand still,

With steady improvements, it starts to fill,

The market gaps left by the giants' retreat,

Its growth in power is no small feat.

"Rationality" the giants' doom does seal,

And thus, disruption turns the enterprise wheel.

[3] At least when this was written, Salesforce's AI offering was very simple, so I don't think it counts as “AI” in the sense of current advancements in LLMs, etc. etc.

[4] This analogy is still very poorly structured, but serves to illustrate the point.

[5] “Navigating the next frontier of enterprise software – AI Primer”  by Brad Sills and team at Bank of America Securities.

The full transcript (and a link to the episode) can be found here. The quotes were lightly edited for clarity and the most important point highlighted in bold by me, with no change in meaning.

On the Ukraine war

On what happened in the Ukraine war up to date:

... when the Russians decided to invade, they had strategic objectives: to collapse the Zelensky government; capture the capital of Kyiv, to do that relatively quickly; advance from the Russian border all the way to Dnipro River, do that in a short amount of time, four to six weeks, perhaps; and then also to cut off Ukraine’s access to the sea, to the Sea of Azov, by securing Kherson and Odessa... the invasion started on the 24th of February—in very short order, Russia fell short of their strategic objectives, really within about a month or so. And the Russians could not militarily achieve what they set out to do.

And then at the end of March, maybe beginning of April, Putin readjusts and resets his strategic objectives. And then he says, "I’m just going to limit my objectives to the southern border, the southern provinces of Ukraine, and to consolidate power in Donbas, secure the Crimea," and so on. So he takes his entire military, shifts them further to the south and to the east. And then he launches a second set of what I would call operational objectives—and he failed there as well.

So you’ve got essentially a situation along a frontline that extends [from] Washington, D.C. to Atlanta... It’s quite a ways. And that front line hasn’t changed hands. It’s essentially been stalemated. And then the Ukrainians asked us for help to build up their force so that they had the capability of conducting offensive operations with combined arms maneuver with heavy forces, mechanized armor and infantry. We’ve done that

Ukraine is now more prepared than ever to either defend against Russia or attack it:

What I will say is that over the last several months, the Ukrainians have asked us for assistance, military assistance, to help them, to train, to man, to equip their forces. Specifically about nine brigades worth of combined arms, armor, and infantry type forces. Also, there’s some light infantry, ranger type units, that we helped train. And I say we—I mean NATO, all of the European partners.

So I would tell you that the Ukrainians right now have a capability to attack and they also have a capability to defend that's significantly enhanced from what they were just a year ago for conventional operations. So they can either do attack or defense. So I don’t want to suggest that they may or may not conduct an offensive operation in the coming weeks. That’ll be up to them. They’ve got a significant amount of planning and coordination and all of that to do, if they were to do an offensive operation. But they’re prepared to do offense or defense.

On prognostics for the war:

How does it end? Let’s just say for the sake of argument that there is an offensive... I think that it’s fair to say, if there were an offensive, that there’s a possibility of a variety of outcomes. Clearly one of those outcomes could achieve significant success and collapse the Russian frontline across the board... Then there’s a possibility of partial success; There’s a possibility of limited success; There’s a possibility of no success. Then the opposite is true—maybe the Ukrainians are going to do a defensive operation, and the Russians would have a great challenge mounting an offensive operation.

I do think, though, that the probability of either side achieving their political objectives—war is about politics through the sole use of military means—I think that’s going to be very difficult, very challenging. And frankly, I don’t think the probability of that is likely in this year.

And I think that rational folks, as part of the Russian decision-making process, will conclude, I believe over either months or a year or two, that the cost exceeds the benefit,  and it’ll be time to do something, at least from a negotiating standpoint. I don’t know when Putin will be ready to do that—but at a certain point, if he’s rational, he needs to do that... They’re going to have to figure out [what they need to do to realize they need to negotiate] because they’re not going to win.

Escalation (towards direct (and potentially nuclear) conflict):

I think that it’s in everyone’s interests not to escalate. Russia does not want a war with NATO or the United States, and NATO and the United States don’t want a war with Russia. So it’s in everyone’s interests in that regard, and Ukraine certainly doesn’t want that scale of war in its territory. So it’s in everyone’s interests not to escalate. Having said that, the possibility of escalation is very real. Wars are highly emotional; there’s a tremendous amount of fear, there’s pride, there’s interest, as Thucydides would tell us.

On the military consequences of a tri-polar world:

What we have to be conscious of, and careful of, is not to drive China and Russia close together in a military sense.

There’s going to be relations between countries, so competition’s not the issue here. The issue is conflict and war. So we want to make sure that Russia and China don’t form some sort of geostrategic, political, military alliance against the United States. We’ve seen some economic assistance; not strong in terms of the military piece of this. Whatever exercises they do are small, relatively inconsequential. In terms of military support and lethal support to Russia, nothing really significant yet. The Russians have asked, for sure; they’re asking a lot of countries for ammunition and so on...

President Xi, I would argue that he—very, very tough guy, hard guy, consummate realist. Very ruthless, very ruthless; but they’re very realist in the sense that they are keenly aware of cost, benefit, and risk, and they too do not want outright armed conflict with the United States. They recognize—the Chinese do—how powerful the United States is. Despite what people may say out there, the Chinese are fully aware of how powerful the United States is. And so they’re not looking for that kind of armed conflict either. They want to achieve their national objectives, but they want to actually do it without armed conflict.

On China and a potential attack on Taiwan:

An invasion of Taiwan by China is not going to look like an invasion, necessarily, of Ukraine by Russia. The fundamentals are different, in the sense that—just the terrain and the weather, it’s obvious. You’ve got a landlocked country of Ukraine, with a land border with Russia.

In order to attack Taiwan, they would have to mount an amphibious invasion combined with paratroopers and air assault, rotary wing helicopters, missiles, all the prep fires that would go into that; they’d have to isolate beachheads and then have to have the amphibious lift in order to do that; and cross basically a hundred miles of water, which is challenging in and of itself. Then they’d have to ensure that the subsurface of the water was secure, as well, from submarine attack. They’d have to clear mines, clear beaches, they’d have to go in and essentially attack and seize an urban area that’s about three and a half million people, in a country that’s very mountainous and lends itself to the defense.

Think about Normandy. At Normandy, the United States and Britain put about 120,000 troops ashore—plus they dropped in three airborne divisions the night before, and they put about 120,000 troops on the beach, I think before noon or by the afternoon; and then there were follow-on troops in the days afterward. That military that landed at Normandy had already done the invasions of North Africa, had done the invasion of Sicily, had done the amphibious operations in Italy, and they had the benefit of the lessons learned of, I don’t know, probably a hundred or so amphibious operations in the Pacific during World War II. And they were led by experienced, seasoned leaders that were hitting the beach, et cetera.

We need to deter armed conflict. And how do you deter? We know through history that the way to deter is to have a very, very strong, capable, multi-domain military, and ensure that your opponent knows that you have that capability, that that capability is overwhelming, knows that you have the will to use it, and you’ve communicated that to them. So what we need to do is make sure the United States military is not only just a little bit better, but a lot better, it’s overwhelmingly better than the Chinese military—to make sure that they know it and that we have the will to use it in the event of a crisis.

I kind of fall back to, you know, the old saying from Teddy Roosevelt’s time, right, Which is, you know, speak softly, carry a big stick, that sort of thing.

I think we should develop our military to such a—modernize our military to such a degree that it is overwhelmingly obvious to the Chinese that they cannot defeat it.

On contemporary warfare and the fact that we're at a pivotal moment

I think we’re in a pivot point in terms of what I’ve referred to in the past as the character of war. You’ve got the nature of war and you’ve got the character of war. The nature of war, arguably, is immutable. War is politics. That war involves fear and friction, uncertainty, chance—that’s the realm of the nature of war. And as long as humans are involved in war, then I would say those fundamentals about the nature of war—probably true. So the nature of war arguably doesn’t change. But the character of war changes frequently... it refers to the tactics, the techniques, the procedures, the organization, the weapons, et cetera. And the character of war changes often; every time you get a software upgrade, technically the character of war has changed somehow.

The character of war only changes fundamentally once in a while. Think the development of the wheel, and then all of a sudden you’ve got chariots. Think of putting a bit in a horse’s mouth, and now you have the development of cavalry. Think about putting lands and grooves inside a metal tube and you go from a musket to a rifle. The biggest fundamental change that is commonly cited historically is between World War I and War War II, where you get the introduction of three technologies—the airplane; mechanization, the wheeled and tracked vehicles; and then those are linked together through wireless communications, through radio.

I would argue that in today’s world, we are undergoing the most fundamental change in the character of war ever in recorded history, and it’s primarily being driven by technology. So what are some of those technologies? Well, first of all, you’ve got precision munitions and you’ve got ubiquitous sensors. So we can conduct long-range precision fires with greater accuracy, at greater range, than at any time in human history, period.

You’ve also got fundamental change happening in the ability to see. So anyone who wears a Fitbit or GPS watch or runs around an iPhone—that’s a sensor. You know, it’s a means of communication for most people, and some track your health, I guess. But for other people it might be a sensor, a tracker. So we have an ability to sense and see the environment, and to pick up signals because there’s so much electronic signals in the environment. We have the ability—not just we, Russians, Chinese, et cetera—have the ability to see and sense that environment like never before. You can go on Google Earth today and get mapped data and see satellite imagery that was only available to the world’s most advanced militaries as late as, like, five years ago, 10 years ago. So our ability to sense the environment is incredible.

So the ability to see, and the ability to shoot, and shoot at range with accuracy, never before like it is today—just those two fundamentals in and of themselves augur a change in the fundamental character of war.

You could potentially—potentially—see artificial intelligence and quantum computing combined with robotics, become a dominant factor in the conduct of war. Combine that with the domains of cyber and space. There’s a lot of things happening undersea, and there’s about 20 other technologies that I won’t go over. But you’ve got this convergence of technologies that is driving, fundamentally driving, significant change in civil society, in human relationship to work, for example, our relationship to each other.

And there’s zero doubt in my mind that that’s going to have a huge impact on the conduct of military operations in the future. And just like in the past, the country that optimizes those technologies for the conduct of warfare—that country is going to have a decisive advantage, at least at the beginning and the opening shots of the next war. I want that country to be the United States.

Many of you guys asked me for a podcast or audio file to listen to instead of a video, so thankfully Substack offers this podcast feature to which I uploaded the audio of my talk. I'm also offering the slideshow below :)

Link to slides

A (rough) transcript of “A genealogy of network effects”

“I thought about doing something a bit different today for unlocking potential, which is to record a video. I didn't have time to write something good enough to publish. So I thought I'd reproduce this talk I gave at Qulture.Rocks. We're trying to do more of these tech talks where people teach something they know to the rest of the team, and I did one such talk a couple of weeks ago about moats and Network effects and so on and so forth, which was an open-ended talk,  done off the cuff. So I thought about trimming it here and there and sharing. 

I'm using this new tool called Loom, which just raised a big round from Sequoia. It's nothing groundbreaking. We've been using something different called Soapbox from a company called Wistia since about two years ago. It's honestly better than Loom. But, somehow, this Loom product got much publicity after the cool kids from the Valley (meaning Sequoia but a bunch of angels like the founders of Instagram and Front) invested in it. 

Anyway, the first time I did this talk I named it "moats, network effects, and positive feedback Loops." This current version is more like a "Genealogy of network effects." One thing I do, when I try to understand something new (or deepen my understanding of something not so new to me) is to try to roughly figure out where it stands within a greater a corpus of knowledge. I think genealogy would be this tree structure that helps me understand this.

What got me to study moats and network effects

I've been really interested in studying more about moats, so as to figure out what makes an amazing company successful in the long run. I did a lot of Michael Porter rereading and then stumbled upon the works of this guy called Pat Dorsey's, which I've shared with you in the past (he's a value investor, of the Warren Buffet type). I also went back to Peter Thiel's Zero to One book, which touches on this tangential concept of monopolies, and how they arise and have seen more and more stuff about moats generally.

And network effects are a special type of moat that I think are pretty interesting to look at. And I was reading about network effects when I came across this article from NFX (a firm which is supposedly the expert in network effects-powered businesses). They're a venture fund and they tried to do this network effects map that in order to catalog all sorts of network effects.

Since reading about McKinsey's way of thinking and presenting their consulting work I became obsessed with this concept of MECE, which means Mutually Exclusive and Collectively Exhaustive, just a beautiful way of saying you're doing these categorizations in a way that doesn't mess with your mind [1]. Each category is independent of each other, just like a bunch of not-overlapping Venn diagrams (the categories are mutually exclusive). Also, the sum of all categories (all the Venn diagrams) is exhaustive. 

The NFX article was helpful but could be much more helpful. It was very confusing. Just a weird Venn diagram with circles that seemed to be randomly thrown at a slide. So that just boiled my stomach. It didn't look MECE at all, and therefore it didn't really give me a framework with which to understand network effects.

To get up to speed, I brought this Robert Metcalfe quote that says, "In network theory," which is what preceded network effects, "The value of a system grows at approximately the square of the number of users of the system." This squared relationship between the size of a network and its value was named Metcalfe's law and it's kind of a corollary or something that's really used whenever people talk about network effects. Generally, if we think about Facebook, the idea is that the value of Facebook grows at the square the number of users of it, whatever that means, which I just can't grasp. You just get a sense of this very strong relationship between the size of a network and its value. I'm not a numbers person.

Why are Network effects important? 

A company is generally worth the present value of its future cash flows. That's Finance 101, or maybe Valuation 101. It's pretty hard to grasp the impact of cash flows when these same cash flows - meaning the positive ones - are very much in the future, as they are with startups. 

What determines the size and the direction of these future cash flows is the quality of the business, which comes from, in great part, how defensible the business is. What's generally accepted is that there is no economic profit, meaning a company's not going to make more money than it "should" because its competitors are going to eat away that excessive return. They're going to just bite and bite until there's nothing left.

What's understood as an economic profit is just basically the company making more money than it should. It means making more money than its risk-adjusted cost of capital, which is how much when investors generally debt or equity or whatever are willing to lend money to it for. The interesting thing is these economic profits, they may be possible or they may be bigger, the bigger these moats are. 

What is a moat? 

Moats as a concept became popular, I'd risk, because of  Warren Buffett and Charlie Munger. 

The literal definition of a moat is this lake that encircles a castle. In medieval times, the lake - the width of the lake and the depth of the lake, and maybe the presence of alligators or piranhas or something - would deter invaders from penetrating the castle. That's the idea of a moat. 

The moat is your protection against these competitive invaders that want to eat away your economic profit. When we see Warren buffet talking about it, is he likes to chase businesses that have big moats, of course. The whole idea of what Warren buffet does is buying a big moat, wide moat, strong moat businesses, but he also looks for management that wants to widen, to strengthen, the moat every year. He likes the idea of deliberately widening and deepening the moat as sort of this strategic imperative of companies. 

As we said, a moat is basically defense against competition. So what do we mean by competition?

When we talk about competition, we mean competition for profits. I think here Michael Porter helps a lot to understand who competes away our profits or who are the players that are trying to eat away our profits. His five forces model is the best I think to think about this. What he basically says is that there are five forces determining your economic profits. These are existing competitors who are currently in their market, new potential competitors who may join your market if they think it's attractive, your suppliers and your customers who are also bargaining with you and charging you more or paying you less and so on and so forth. Also the threat of substitutes, which is generally a degree of separation more. If we think about current competitors are these ... I think there's no clear line that really separates what's a competitor from what's a substitute.

If we look at Clayton Christensen's 'Jobs to be Done' framework, he's got the Innovator's Solution, which is this book here [author showing the book]. He's basically telling you that everybody that gets a job done competes for a market or, better, can be considered competitors. You may be doing, for example, your performance reviews in a system like Qulture.Rocks but you may as well do it in Excel or Access. These guys are like competitors, but we'd call them substitutes. These five groups or constituencies are our competing away your profits and fighting for it.

New entrants, they enter your market either because you have very low barriers to entry or not enter your market if you have high barriers to entry. This is a good way of thinking about how defensible. I'm not going to get into five forces or porter here but if you want, it's pretty easy to find stuff online about it. New entrants may enter your market as a function of how big are your barriers to entry. Your current competitors may compete more fiercely or less fiercely for your pool of profits. There are different market dynamics that make them more fierce or less fierce. For example, if you have assets that are highly specific, you might be incentivized to compete more fiercely because you can't do anything else with your assets.

Also, if there's a 'winner takes all' dynamics in your market or even the perception of a 'winner takes all' dimension, which sometimes perceptions and realities are not the same. I think I've argued before on the newsletter that I think the 'winner takes all' dynamics of right sharing is overblown. They will compete too fiercely. Another dynamic that drives up the internal competition is marginal costs. The smaller they are, the bigger than the incentive of players cutting prices for example because it just doesn't matter. Customer and suppliers, they also may take away your profits, the bigger their bargaining power against you or with you. So again, moats are a defense against competition.

What's interesting is that there are different types of moats.

A taxonomy of moats

As we talked about a bit earlier, Peter Thiel doesn't call them moats but he talks about these factors that enable monopoly-like businesses. I've talked before about Pat Dorsey who worked on this theme when he was at Morningstar and then he founded his own asset management company, like a long-only equities management company. He talks about four types. He talks about economies of scale. He talks about intangibles which are brand, eventually technology and also some regulation power or monopoly that you have. So the government is granting you a monopoly or a stronger position. He talks about switching costs and he also talks about Network effects, which are our focus here.

But I was reading this book by Hal Varian, who's the chief economist at Google. He has a very old book, oldie but goodie, called Information Rules. I think I cited it before. He doesn't speak quite exactly about moats, but he speaks about what makes strong competitive positions in the information economy. He basically has these two broad categories. One is locking, which is basically switching costs, how hard it is for me to switch products, and the other one is positive feedback loops. He talks about two subtypes of positive feedback loops, demand-side and supply-side economies of scale. I like this example much more because I think a brand is a fickle moat. I think regulation is an unpredictable moat. I think tech, it's hard to find tech that is really something else. I do think tech sometimes spills over to Network effects, which are a sort of positive feedback. IMHO, it is a better model.

So I like this and I was wanting to work with this taxonomy of moats.

Economies of scale are pretty simple. There is a type of supply-side economies of scale. They are the type of supply-side economies of scale. The more you produce, the cheaper your unit costs are. Most manufacturing businesses have that dynamic, also businesses that have high operational leverage, which is high fixed costs. Once they start producing more volume, unit costs fall abrupt thusly. For example, one sticker. I can produce one sticker for R$10. I can produce a thousand stickers for 20 cents. That is a sort of moat because it deters ... like if a competitor comes, he's probably going to start producing, I don't know, one or 10 stickers.

His cost is going to be very high. My cost is going to be very low and I can even drive him out because I might cut prices or whatever. I have this big protection against my new entrants especially. Switching costs prevent my customers from switching out of my product. There's a very good example, which is ERP software. It's so damning to install into and to implement, that people don't ever want to move out of it. These have very high switching costs and they give the supplier a lot of bargaining power because the supplier may want to raise prices by 10% of 20% a year and still don't weed out the customer base. For example, how easy it is for a customer to leave AWS or Heroku. You have deep integrations for your system.

You're going to think like a thousand times or you're going to think a lot before moving out. These businesses tend to have high bargaining powers with us and especially if these lines, like the best businesses, have very high switching costs, but they're a small line on my P&L. I'm even more incentivized to ignore a price hike. Then we get to the demand-side economies of scale. Well actually, supply-side economies of scale are also positive feedback. Maybe we should just ignore this slide. Let's talk a little bit about this. Hal Varian, he spoke about positive feedback, which Jim Collins calls the 'flywheel effect'. Hal described these two blocks, which is supply-side and demand-side economies of scale.

We already spoke about the supply-side, which is cost advantages attributable to scale. Demand-side economies of scale are what generally people call Network effects. This is pretty general. A network effect is when a product is worth more, maybe to each additional user, the joints, the more users this system has. This is a more working definition of Network effects but I think it's similar to Metcalfe's definition, which we showed you at the start of this. Hal also speaks about this. He says, "Whether real or virtual," so he's saying like either Facebook or maybe a telephone line network. Networks have a fundamental economic characteristic. The value of connecting to it depends on the number of other people already connected to it.

The value of me joining a social network depends on the number of people that are already in. He says this fundamental value proposition goes under many names; network effects, network externalities and demand-side economies of scale. The quintessential question that describes what a Network effect is, how valuable would new Instagram be if none of your friends were there? You have no reason to join whereas if everybody is in there, you have all the reasons to join because people are there. You'll have friends to connect and friends to stalk and so on. I think a good thing is to think about a better taxonomy of these positive feedback loops.

Hal Varian speaks about supply-side and demand-side economies of scale. On the demand side, I think we can generally have these two categories which are, you have demand-side economies of scale that have a user as a primary driver of value whereas you have other types of these demand-side economies of scale where users are not the primary driver of value.

We're going to look at some of these types and I'm going to show you how and why and examples they're not the same. If we look at WhatsApp for example, the value of using WhatsApp or Telegram or Symphony is greater the more users are in and users are the primary lever of the value of entering this network. I want to use WhatsApp if, and only if, other people are inside WhatsApp.

It’s users as a primary value driver. Also, it doesn't matter to me if there are only users. Everybody that joins this network are users and only users, whereas if we look at Facebook, it's a different thing. I'm going to join Facebook because of other users and I have more incentives to join Facebook the more users it has, at least users that are relevant to me.

These network effects are almost geographically or social network limited. I'm going to join Facebook if people that are relevant to me are inside. There's also a different dynamic going on, which is the more users on this left side of the platform there are, the more advertisers there will be on the right side of the platform. This is not a network effect per se, but it is a positive feedback loop because the more users, the more advertisers, and the more advertisers, the more money Facebook makes. There are interesting specificities here, which is the fact that maybe in some ways, the more advertisers there are on the right side, maybe the worse this experience becomes to the users on the left side.

Also maybe, the more advertisers there are on the right side, advertisers don't want the competition. There may be even negative Network effects on the advertiser side. Ad costs go up and you may crowd out advertisers. Google is even different.

When I use Google, I don't care if other people that are relevant to me use Google. Nonetheless, the quality of the product becomes better, the more people use it. That's what I call this Network effect where the user is not the primary driver. When users join, the quality gets up but the driver of Network effects or demand-side economies of scale, it's actually the quality of the model and not the people. That's the only difference. In the other sense, there are also two sides of the platform and so on and so forth.

A part of it works just like Facebook. On eBay, it's actually buyers on the left side and sellers on the right side. I don't care if there are other buyers on my side. I just want to find the products I'm looking for. The more people on one side, the more people on the other side.

There's not a lot of a relationship between the more people on the right side, the more people on the right side. It's across. Maybe the more people on the right side, the more people on the left side, and the more people on the left side, the more people on the right side. This is a positive feedback loop, but it's quite different.

Quora for a change is like Google. I joined Quora to get interesting content and to get my questions answered. I don't really care about the social aspect of Quora. We may argue that there is a social aspect like status-seeking, but in some ways, I want answers.

I want content. The quality and the network effect is more driven by quality, which is therefore driven by users, but not users directly. Users are primary drivers. This is to give you a sense of what I think are primary versus non-primary drivers. I think the type of network Metcalfe was thinking about was a network that has users as the primary drivers of value for the network. I don't think he was thinking about platforms, like these two-sided platforms or these multi-sided platforms.

We're really trying to break ground here and figure out a better taxonomy of this thing because when people coined the Network effect term, there wasn't really anything of the sort of platforms or maybe something. So thinking about network effects as this whole positive feedback group, I think it's not helpful. I'm trying to give you a better way to think about this.

When Jim Collins talks about flywheels, he may be talking about positive feedback in how variants taxonomy or he may be talking about demand-side economies of scale, which are like your Network effects. When he talks about Amazon for example as having a very strong flywheel, it's a mix between both because now ... Amazon has the “selling their own goods” part of their eCommerce operation.

There's this clear supply-side economy of scale dynamics but they're also operating in a marketplace and on that side, there are demand-side economies of scale in terms of a two-sided platform where the more buyers, the more sellers, and the more sellers, the more buyers. There's a mix between the two. Reforge, which I wrongly typed, regorge calls these macro goal growth loops. I think they go on just too far with this. I don't like their definition. Jim Collins, a flywheel is a company with positive feedback. Amazon's fly was very normal or whatever example we showed. The more customers, the lower the prices, the lower the prices, the more customers come in. The more customers come in, the lower the prices and so on and so forth, until they become a trillion-dollar company.

These positive feedbacks or flywheels, I think they have different flavors and people want to say every company is a flywheel. I heard this outrageous example the other day, which is, "The more product I sell, the more revenue I make, and the more revenue I make, the more salespeople I hire." This is stupid. Every company is like this. I don't think that's a very strong form of a positive feedback loop. A strong form of a positive feedback group is a winner takes all kind of a dynamic. I think Windows was a good example because everybody's going to develop for the winner. I don't think it is a winner takes all like a perfect winner takes all dynamic. For example, I use Mac and a lot of people use Windows. Some people use Android, some people use the iPhone. I think there's maybe room for a duopoly or maybe an oligopoly. Surely not for a lot of competitors. It may be not a winner takes all dynamic, but you must get up to a certain scale or critical mass before you're something, before you pass the surf and get to calmer seas.

Going back to moats, I think a good way to think about this is like three types of moats: Regulation, brand, which is a weak one, and positive feedback. Within positive feedback, there are supply-side economies of scale, which are what we think about when we think about “economies of scale”, and demand-side economies of scale, which are, roughly, what we think about when we think about network effects.

I really hope I have helped you better understand moats and network effects. It helped me a lot, anyway. I'm trying to be a Richard Feynmann devotee here in trying to explain stuff as a way to understand stuff. I'm sure my mom or my grandma wouldn't understand a single second of what I talked about. I have a lot of progress in the making. But I think my understanding got better by trying to explain this to Q.Players and now to you guys.

Thank you so much.”

Notes

[1] If you want to learn more about MECE, there's an interesting book called "The Pyramid Principle" by a lady called Barbara Minto, which I think is pretty interesting as a handbook of how to organize your thinking and how to write. I think MECE comes from Minto. 

Bank run

Imagine your company has some money in a checking account, in a bank, and you want to withdraw that money to pay for something, such as payroll. But, figuratively, when you get to the bank you see a long line of people waiting to take their money out too. You might start to worry - why are so many people in line to take their money out? Is there something wrong with the bank? Should I be doing the same?

The thing with western banks is, if enough people start to worry and decide to withdraw their money, there's no money for everybody. And when customers start to withdraw their money in tandem, trying not to be the ones left with the bag, it's what's called a "run on the bank." Everyone rushes to take their money out and the bank goes under.

Statistically, only a small number of customers might request their money back at any given day and in future days. Statistically, it follows, the bank could use some of that money in a number of ways, including making loans to other people or businesses and purchasing securities.

Mismatching mental models and regulation

There is a fundamental mismatch between most people's mental model of what a bank checking account is (a "safe box in a vault”) and the actual workings of the product (a "loan to the bank that you can call back instantly"). When one deposits money in a bank, one's actually lending said money to the bank. Only, it's an (unsecured) lending facility that pays zero interest and that must be repaid instantly shall the customer request it. Similar, but also radically different.

But customers don't think that way. Customers think they've placed their money in a digital safe box, where the money is kept until they ask for them. Customers who deposit money in a checking account don't assess credit risk the way investors who buy a bond from the same issuer do.

(Many critics to the FED + FDIC's actions (guaranteeing all deposits regardless of size) think checking account depositors should bear the loss of a bank's failure, as if they were investors in its bonds. I don't agree with this view.)

This mental model mismatch seems to be a good reason to reassess how the product works (i.e., regulations around it).

Reserves

Under Federal Reserve rules, banks operate by using a system called fractional reserve banking. For every dollar the bank gets in deposits, it has to keep a fraction of that dollar deposited with the central bank, i.e., the FED in SVB's case, in part so it can meet the demands of customers who want to withdraw their money.

Reserve levels, i.e., the percentage that has to be kept with the FED, are set according to two main variables. The first is the FED's perception of what's, statistically, the most that can be withdrawn by customers in a short time frame; the second is the FED's predisposition to ease or tighten monetary conditions [1].

Depositor's Dilemma

The prisoner's dilemma is a scenario in game theory where two individuals are arrested for allegedly committing a crime together. They are held in such way as to not be able to communicate with each other. The figurative prosecutor makes an offer to each prisoner: if you confess and your partner remains silent (i.e., doesn't confess), you'll get a lighter sentence, let's say 1 year, while your partner will get a heavier sentence, let's say 5 years. If both of you confess, both get a moderately heavy sentence, let's say 2.5 years. Finally, if both of you remain silent, both of you get the lighter 1 year sentence.

It follows that each prisoner must decide whether to confess (hoping the other partner keeps quiet) or remain silent (again hoping the other partner keeps quiet). But one can only hope, without knowing for sure what the other partner will do. The dilemma arises because each prisoner is individually better off by confessing, regardless of what the other does, so as to "guarantee" either the lightest or the moderate sentence. However, the logic applies to both prisoners, so both prisoners will tend to confess, and then receive the worst outcome of 5 years in prison, showing how individual rationality can lead to a worse outcome for both when they are unable to cooperate or communicate with each other.

The prisoner's dilemma has been often used as a metaphor for the dilemma faced by depositors in a potential bank run (let's call it the Depositor's Dilemma): In a bank run situation, each customer of a bank is rationally compelled to withdraw their money as fast as possible. If one doesn't and other customers do, there'll be no money left for everybody.

That's probably the similarity where the confusion comes from. But it's a pretty imperfect comparison [3].

Going forward

As a result of the failure of SVB (and Signature) there's the risk that other regional banks are affected with similar runs on their deposits. The risk of such contagion is surely contributing to First Republic Bank's shares dive into the abyss:

Investors probably think other regional banks may i) hold losing positions in fixed income [4] or ii) face huge deposit withdrawals in the coming days.

One huge problem that needs to be solved by U.S. regulators is that these movements tend to become vicious cycles: shares and bonds plummet, depositors become scared, withdrawals skyrocket, shares and bonds fall even further, and so on and so forth, a cycle which took a matter of days in the 30's - i.e., was already really fast - and has been further compressed by the internet, smartphones, and Twitter.

Inflation and interest rates

An interesting second order effect of this whole crisis is the position it's left the FED in, or, whether the FED will change the way it has been dealing with inflation.

The path until now has been to tighten monetary conditions, by raising the FED funds rate and refraining from buying government bonds from banks (quantitative tightening). But tighter monetary conditions dry up liquidity and may make the lives of regional banks even tougher. On the other hand, loosening monetary conditions (and helping the banking system remain stable) could make taming inflation even harder than it already is - rates haven't fallen significantly even on the back of one of the steepest rate hike cycles ever. The market seems to believe that the FED will indeed change course, as seen by the following graph, of the yields of 5-year treasury bonds:

It will be certainly fun to watch how these events unfold.

Notes

[1] The FED has a number of tools at its disposal in pursuit of its mandates of inflation control, economic growth - subordinate to inflation being under control - and financial system stability. The main one, of course, is the so called FED funds rate, the rate at which the FED loans money to banks on an overnight basis, which in practice serves as a floor for interest rates charged in an economy; the second is purchasing and selling government bonds from banks (more about this tool here and everything else on this article here); the third is setting reserves on cash balances, commonly referred to as deposits. If the reserves are higher, a smaller fraction of marginal deposits can be lent out. Conversely, if the reserves are lower, a higher fraction of marginal deposits can be lent out [2].

[2] We're not going into the fact that money lent out can be then deposited at another bank - or even the same bank - and, aside from reserves, be lent out a second time, and on, and on, and on. The effect of reserve requirements on this virtuous or vicious cycle is called "velocity."

[3] For example, there's a timing component in the Depositor's Dilemma that's absent in the Prisoner's. In a bank run, you're not better off only by withdrawing your money: you have to do so before the bank runs out of it, which means doing so before other customers.

[4] The whole problem with SVB started with the risky decision by the bank's management to invest a huge portion of their cash in very long-dated (10 year) U.S. treasuries and mortgage-backed securities (MBS). Management found a bit of yield in those bonds (it purchased the treasuries at a yield-to-maturity of 1.5%) when FED funds were at 0% or thereabout. But FED funds (and the whole rest of the yield curve) climbed fast as the FED sought to fight inflation, and the value of those treasuries took a crazy dive [5]. This loss wasn't being accounted for because the bank used this crazy accounting gimmick to not mark the bonds to market - it placed the bonds in accounts meant for securities that are going to be held until their maturity (i.e., for ~10 years). BUT, and there's always a but, the bank changed its mind regarding holding to maturity and decided it needed to sell the bonds, and by selling them, they sealed the loss, and had to announce it to investors.

[5] The same logic applies to MBS, which were, for different reasons, the culprits of the 2008 financial crisis.

After a long hiatus, about which I pretend to talk about later, I wanted to write about a topic I've been thinking about a lot: performance, more specifically at work.

Every entrepreneur on earth wants her company to perform at its best, or at its fullest potential. In order to make that happen, each and every person in the company has to perform at their best.

But what is performance? How can we influence performance? What are its components and determinants?

That's what I want to discuss in this essay. I hope you come out on the other side with a better mental model of what is performance and how to drive it.

Performance = NPV of behavior(s)

Let's start with a working definition of what performance actually means.

The title above is not quite true but points to the truth. A couple of folks (Motowidlo and Kell, 2013) defined performance as “the total expected value to the organization of the discrete behavioral episodes that an individual carries out over a standard period of time.” [1a] I've tried to dumb it down a bit and got to the following: Performance is the value of the behaviors of an employee in a given period of time. Let's break down the definition.

First, we talk about behaviors. It's stuff like:

• writing lines of code, or a post for social media, or a 6-pager memo,

• calling on a customer or prospect,

• designing a UI,

• building a financial model (or tweaking one that was built by somebody else),

• running a webinar,

• contributing an idea in a brainstorming session,

• making a decision, etc.

In other words, behaviors are the actual work.

If we're going to evaluate somebody's performance in a given period, we have to look at all work done by someone during that period, and the work will be a collection of individual behaviors (discrete episodes).

Second, they talk about value. Here, it helps to think that each relevant behavior we take at work has a net present value. Borrowing on finance, it produces a tilt in the company's future cash flows, even if tiny. We hope, of course, that the value is positive, and not negative. But people work for organizations because they are able to offer value to the organization, ideally more value than they cost in terms of salaries (even tough this is most probably sometimes not possible).

If we analyze all these relevant behaviors for a given period - let's say a year - we can theoretically add up their NPVs, compare said NPVs with the NPVs of previous periods, compare the NPVs of different employees, and so on (we'll get more into performance measurement - e.g., assessments, reviews - in a bit.

Results: Abstracting performance

We frequently confuse the meanings of performance and results, but we shouldn't. Results are abstractions. And abstractions are created to make our lives easier, like shortcuts.

One example of such abstraction we frequently call a “result”: some measured improvement in a metric/KPI like “sales for product A” or “net income for year 2023.”

Results can be attributed (with its limitations, as we'll see) to some individual, as in “Erik is bringing in solid results. He sold 100 dollars in new MRR last quarter.” We can even compare actual results with expected or desired results (as in goals/quotas and goal/quota attainment), and say that for somebody as experienced as Erik-the-salesperson, this quarter's sales of 100 dollars are about right, and last quarter's sales of 90 meant some level of “underperformance.”

But just like the map is not the terrain, results are not performance per se. If Erik sells 100 dollars of new MRR, the amount sold is not his performance. We might tentatively say the present value of this result is his performance. But it's still not it. Erik's actual performance is the present value of all his behaviors at work. His sales - the result - is a proxy.

We have to be very aware of that proxy status and its potential limitations.

Some limitations of the results-performance proxy

There are two problems with blindly equating (specifically in, but not limited to, the case of salespeople) sales dollars with performance.

The first problem is that several other factors outside of Erik’s control may have contributed to this quarter's number being 100 (and not 50 or 120):

• A warmer-than-average lead may have been referred to Erik by the CEO and converted into a 20 dollar deal.

• The Government may have pushed a bill that created regulation in a territory/industry covered by Erik in a such a way as to make the product he sells required by regulation.

• Erik may have caught COVID-19, and therefore spent two weeks at home, unable to work because of the symptoms.

The point here is that affirming Erik's 100-dollar production was strictly attributable to Erik is probably not true

The second problem is that many factors may influence how much value the 100 dollar sales figure actually created for the company. In other words, to presume that the 100 dollar figure is a good measure of how much value was added to the company is frequently dangerous. For example,

• Erik may have - knowingly or not - sold a deal that is very likely to churn in the short-run

• The majority of the deals sold by Erik may be into a low-growth industry, which is very unlikely to bring upsells throughout the lifetime of the contract

• The average pricing practiced by Erik may just be off - too cheap, eating margins, or too expensive, causing reputational risk.

The point here is: the 100-dollar figure produced by two different salespeople may mean a very different value-added to the organization.

I know you must be thinking I'm going too deep into the sales examples, but it was intentional. I wanted to take down the case where equating performance with a result is the easiest and most tempting. If we move to other functions within a company, as we'll do, it's even trickier.

But bear in mind: Even though I'd say 99% of top sales organizations mix the concepts of performance and quota attainment so much that there seems to be no distinction between the two, I think it's ok to do so in the case of sales: weighing costs and benefits, you'll be fine. I just wanted to point out how severe the limitations of doing so are, especially given that doing so for most other functions shows even more severe limitations.

As an aside, let's think of what performance means in the reality of a developer, and most importantly, how limited the most obvious abstraction would be in measuring her performance.

Based on our definition, performance would mean the total value of the developer's behaviors. For example, shipping code that then becomes a feature that customers use could be an example of a performance sample, and the value of the use increment, the actual performance of the developer. Further, shipping that feature using clean, understandable, reusable code is an even better example since the developer prevents problems that will happen in the future (and subtract from future cash flows).

But would you feel comfortable equating the number of lines of code merged by the developer as value-added to the organization? How about if you threw in an additional code health? Or the number of comments said code gets from code reviewers within the team? Even then, I think you wouldn't want to do that. It's just too crude an abstraction. (Btw, it's not so distant an example. I've seen people argue for it. No kidding.)

A better proxy would be how much usage actually increased. Let's say the feature allegedly drove an uptick of 10% in MAUs/DAUs. Good? How much of the uptick was due to the developer, as opposed to the designer? Or the PM? How much have the holidays/weather/launch of the new Macbook Pro influenced the uptick?

Now that we've defined performance and discussed some of the risks in equating performance and the results - allegedly - produced by said performance, let's look at something that might be more interesitng: How can we enhance performance? What does it take for someone to perform? Or even better asked, what makes, et ceteris paribus, one person perform better and another perform worse?

That's the realm of performance determinants.

The determinants of performance

According to this guy Campbell (1990) [1c], there are three determinants of performance: declarative knowledge, procedural knowledge, and motivation.

In a nutshell, these mean, respectively, knowing what to do, knowing how to do it, and wanting to do it (the difference between what and how is sometimes blurry. I suggest just absorbing the geist).

Declarative knowledge, or knowing what to do, is a matter of education. Campbell defines it as “knowledge of facts, principles, and procedures— knowledge that might be measured by paper-and-pencil tests, for example”.

According to Wikipedia (which has some great easy-to-grasp examples):

In epistemology, descriptive knowledge (also known as propositional knowledge, knowing-that, declarative knowledge,[1][2] or constative knowledge)[3][4] is knowledge that can be expressed in a declarative sentence or an indicative proposition.[5] "Knowing-that" can be contrasted with "knowing-how" (also known as "procedural knowledge"), which is knowing how to perform some task, including knowing how to perform it skillfully.[1] It can also be contrasted with "knowing of" (better known as "knowledge by acquaintance"), which is non-propositional knowledge of something which is constituted by familiarity with it or direct awareness of it. By definition, descriptive knowledge is knowledge of particular facts, as potentially expressed by our theories, concepts, principles, schemas, and ideas.[6] The descriptive knowledge that a person possesses constitute her understanding of the world and the way that it works.[6]

I, for example, know what a DCF is, know how an income statement, cash flow statement, and balance sheet work, and so on and so forth). Erik-the-salesperson hopefully knows the product's features, the main objections he can face, and even where he can find the pricing calculator in the intranet.

But it’s also knowing what I'm expected to do: I need to perform a valuation analysis of Acme Inc., by Friday, with the goal of helping my boss make the case of whether our firm should invest or not in Acme Inc.). Erik must tackle objections as they arise, but never preempt it; he must always use the pricing calculator available on the intranet; his quota for the quarter is 100 dollars, 80 in product A and 20 in product B, all of it in multi-year contracts, etc.

Procedural knowledge, or knowing how to do it, is knowing how to apply declarative knowledge. Campbell, again, defines it as “facility in actually doing what should be done; it is the combination of knowing what to do and actually being able to do it. It includes skills such as cognitive skill, psychomotor skill, physical skill, self-management skill, and interpersonal skill and might be measured by simulations and job sample tests.” Having procedural knowledge means one is able to apply the declarative knowledge one has in actual work.

Again, Wikipedia does a great job of describing procedural knowledge:

Procedural knowledge (also known as knowing-how, and sometimes referred to as practical knowledge, imperative knowledge, or performative knowledge)[1] is the knowledge exercised in the performance of some task. Unlike descriptive knowledge (also known as "declarative knowledge" or "propositional knowledge" or "knowing-that"), which involves knowledge of specific facts or propositions (e.g. "I know that snow is white"), procedural knowledge involves one's ability to do something (e.g. "I know how to change a flat tire"). A person doesn't need to be able to verbally articulate their procedural knowledge in order for it to count as knowledge, since procedural knowledge requires only knowing how to correctly perform an action or exercise a skill.

In my example, it means I know how to actually go about doing the DCF. Probably because I've done it before, I've faced some challenges and overcame them. I can open a spreadsheet, spread the financials and project them, calculate some metrics, etc. It means Erik knows how to answer the objections without losing the proper tone and voice, etc.

Motivation, to end our triad, means to want to do something. Campbell further breaks down motivation into three components. The first is the choice to do something. I either want to do it or not. It's binary. The second is the level of effort I want to expend in doing it (if the choice was “yes”, do I wan't to go all in?) The final one is the duration of the effort I want to expend: for how long I'm willing to sustain the level of effort chosen.

Another way to put it is direction, amplitude, and duration (Campbell 1993).

In my example, do I actually want to do the DCF my boss asked me to? That's a biggie. Do I admire her? Did she ask nicely? Can I foresee the impact of doing the DCF correctly and in time with advancing my career? Or getting a bonus? Or fulfilling my self-image of a top-level analyst? (This opens up a whole other avenue of inquiry on the determinants of motivation, an avenue where I won't go for now.) Further, how much am I willing to give this DCF? Have I decided to double-check everything to make sure there are no typos? If I face challenges, am I willing to stay late? To moonlight at work?

Companies and managers both have a huge impact on the determinants of performance. Knowledge (procedural and declarative) can be recruited for (previous experience is frequently used as an indicator of skill) or trained, via onboarding, via in person or remote/async training (using software such as Learning.Rocks), or via on-the-job training performed by supervisors (a lot more here). Motivation can be promoted by making sure people know how they can advance within companies, by having an inspiring mission, by having inspiring leaders employees wish to emulate, etc.

Now we get to an important consequence of our present discussion of what performance actually is: measuring performance. Measuring performance is really important because it allows us to a) improve performance and b) reward performance.

Measuring performance

Measuring performance is the purview of a human resources discipline called performance management. Performance management programs are composed of set of processes based on the premise that helping individuals perform better will help the organization perform better (and based on our definition of performance, the assumption seems reasonable).

The cornerstone of a performance management program is the performance review. The performance review is a process (have you heard of Qulture.Rocks?) whereby employees' performances are measured alongside performance criteria. See, when a company measures performance, it doesn't actually measure the value of the sum of someone's behavior in a given time period, as our definition suggested. It measures some proxy of performance.

Most mature companies measure performance alongside two proxies, or let's call them “axes”: behaviors and results.

On the behaviors axis, they measure the presence of behaviors (according to people that work with the person whose performance is being measured) that are thought to be correlated with positive value-added for the company. It's a correlation, which is why it's a proxy. Anyway, most companies start to measure performance exclusively alongside the behaviors axis, because it requires less sophistication/ is easier to set up, with the frequent exception of salespeople, who can be measured with the help of some very easily collectable metrics such as, duhh, sales closed.

On the results axis, they measure the achievement of results against some expected level, where Achievement = Result / Target. (Each target is oftentimes articulated as a goal).

We'll get into more detail on the two axes.

Behaviors

Behaviors are usually evaluated using a set of behavior descriptions. These behavior descriptions might be general (everybody is evaluated with the same behaviors) or specific to different areas within a company (sales gets sales-relevant behaviors and finance gets finance-relevant behaviors).

Behaviors might be grouped according to the skills they require (e.g., “objection handling”,“written communication”, “financial modelling”) or to values the company (and its culture) requires everybody to show ("e.g., “don't be evil”, “do things that don't scale”, “move fast and break things”).

Results

Results are proxies for the value said behaviors generated to the organization.

In some functions, this is easyish to set up. For example, sales, as we've seen extensively. In some functions, this is very hard. For example, in-house counsel, or accounts receivable (AR). When it's hard, it's usually for two reasons.

First because for some functions it's hard to find KPIs that are relevant performance proxies (is the number of lines of code a good proxy for a programmer?); and even if such KPIs are found, they might not be easily attributable to an individual (how much of the product's engagement uptick was due to the programmer and how much was due to the designer?). Second, because they might be super hard to measure (how practical is it to even measure product engagement? Do we have somebody that knows SQL and has access to the DB to query it?).

As we've discussed extensively, results can be pretty flawed proxies. It's tempting because results can seem much easier to analyze; they can seem much more objective. It's much less time-consuming to not have to shadow a salesperson all year long and observe her behaviors, opting instead to just gauge how much she actually sold in dollar terms (or how much she sold against her quota) and triangulate some measure of performance. It's much less daunting to shoot numbers back when someone questions why a colleague's performance was measured as better.

As an aside, HR professionals usually think these are actually two different aspects or dimensions of performance. Results are “what” someone does. Behaviors are “how” someone does the “what.” But understanding what performance really is shows how wrong this line of thought is, and allows for a much better perspective on this. I'll explain: behaviors and results are not “what” and “how,” or different aspects/ dimensions of performance, but actually different ways to measure of the same thing.

Pairing both doesn't mean looking at the “what” and the “how.” It means looking at performance with two different tools that aim to measure the same thing. In the sales example, that even makes sense: since both tools are flawed, how about pairing them and getting the average?

Anyway, despite the confusion, it might be net good that HR teams use both “axes” in their performance management programs.

Improving performance: working “harder” or “smarter”

How does performance improve? Two ways: effort or development.

This is really interesting.

A salesperson (let's call her A, for consistency's sake) can improve her production by working 30 minutes more every day to do an additional call, resulting, given a constant conversion rate of calls to deals done, in an additional amount of dollars or logos sold.

Salesperson A has another way to improve her performance: tweak her sales pitch in order to improve her conversion rates and, given a constant commitment of hours, sell more.

We could call these two alternatives "working harder" and "working smarter," to be aggressively simplistic.

You can argue that improvements by "working smarter" are more durable because it's harder for them to recede. Once you do your better pitch, why would you revert to the old one? "Working harder", on the other hand, is more volatile. If you're not feeling well, lost a bit of motivation, etc., you can just work less and then produce less.

Takeaways

This was supposed to be an essay, in the sense of an unstructured exploration of a topic without a clear thesis or conclusion. But if you could take one thing with you, I'd urge you to take this: performance is the value that people's behaviors add to an organization. Results and goal attainment are not performance, but proxies for performance. In order to really assess performance, you'd have to shadow people around on all work-related situations and then add the value-added up.

[1a] The book can be found here.

[1b] Most of what I discuss in this article can be traced back to Campbell (1993). It's a great piece of work (even tough I don't have the file) [4].

[1c] Campbell, J. P. (1990). Modeling the performance prediction problem in industrial and organizational psychology. In M. D. Dunnette & L. M. Hough (Eds.), Handbook of industrial and organizational psychology (2nd ed., Vol. 1, pp. 687–732). Palo Alto, CA: Consulting Psychologists Press.

[2] This example still isn't as crisp as I wanted it to be. It doesn't feel 100% right, especially the “duration of effort” part.

[3] I usually hate these “what” and “how” analogies. In the realm of OKRs, John Doerr explains that objectives are “what” we want to achieve, and key results are “how” we will achieve these objectives. So much trouble has been caused by this terrible explanation.

[4] Campbell, J. P., McCloy, R. A., Oppler, S. H., & Sager, C. E. (1993). A theory of performance. In N. Schmitt & W. C. Borman (Eds.), Personnel selection in organizations (pp. 35–70). San Francisco, CA: Jossey-Bass.

Founding and running a startup [1] is kind of a game of its own when compared to founding and running other types of companies (e.g., consulting businesses [2], restaurants, travel agencies, industries, etc.), just like running 100m sprints at the Olympics is a discipline of its own and very different from the Marathon, and even more different from doing half-Ironmans as a hobby or jogging on weekends to lose weight.

The game of startups [3]

Why is the game of startups so specific? Or, what's really the game of startups?

The game of startups is not just the game of starting new companies. Startups are special kinds of new companies designed to grow really fast. The game of startups is the game of building the biggest companies in the shortest time possible [4].

Startups and venture capital

Venture capital is tightly associated with the game of startups but is not a requirement per se. One can totally play the startup game without it. It's just that in order to win, chances are that one will move faster if one uses the help of venture capital. It helps to think about Olympic swimming: fancy wetsuits are not a requirement per se of doing the freestyle 50m, but just about every successful swimmer uses them to some extent, just like just about every successful startup uses venture capital to some extent.

The game changes as the company grows

One specificity about the greater game of building and running companies is that the player playing the game usually changes modalities as the journey progresses.

From founding to pre-IPO, for example, the game may be indeed to build as big a business as possible quickly; from pre-IPO to IPO and beyond, the focus of the game changes slightly to building the most valuable business, even if that means sacrificing growth a bit; as time progresses, a founder may switch gears and play the game of trying to make the company endure the test of time. We could even argue that the best founders are the ones that correctly time these changes and make them successfully (e.g., founders of Airbnb, Facebook, Stripe, etc.)

Seeking the best advice

Seeking (and giving) advice that's good requires one to understand what game or discipline they are seeking advice for, and as important, what disciplines the people you ask is qualified to give advice on.

Notes:

[1] For a great defintion of what a startup is, read startup = growth, by PG. TL;DR, a startup is a company that's designed to become very big very fast.

[2] Whatever that means.

[3] Let's just recursively define "startups'' as the types of companies one tries to build when one plays the sport of startups.

[4] By the way, the sport of startups is not really to build the biggest companies in the shortest possible time, but to build the most valuable companies in the shortest possible time. It's just that in the early stages of a company, size [5] is probably the best proxy for value.

[5] We could also discuss what "size" means. Usually, it's revenue, but sometimes some other thing will be measured, such as the number of users a service such as a social network has, if the people playing the game think that users are a reliable indication of future revenue and then value.

This Marc Andreessen interview was one of the most insight-packed pieces of content I've read in a while. Read it here. Some of my highlights (formatted by myself) below.

I.

The second part of my answer has to do with what happens at a time like right now, when there’s a market downdraft. The minute tech stocks get hit, a lot of big companies basically say, “Oh, thank God, we don’t have to take this stuff as seriously.” This happened in a huge way after 2000. One of the reasons why Amazon took off is because all of the traditional retailers, after 2000, said, “Oh, thank God, we don’t have to worry about this e-commerce thing anymore.” And they just left the field. Borders famously outsourced their online business to Amazon, which, in retrospect, was maybe not the best idea.

This is already happening in this stock down-draft. So, Netflix stock is down 70 percent, 75 percent, 80 percent, whatever. And whereas before, you had all these stories talking about how Netflix was this permanent new, dominant Hollywood force, even a possible monopoly, now you get all these stories saying, “The Netflix model is broken, it’ll never recover,” with big, classic media companies saying, “Oh, thank God, this streaming thing is not going to be ‘the thing’ after all.”

Big companies tend to come in and out of tech this way, and it disadvantages them over time. They still have such a sense of palpable relief when they think that they don’t have to do this stuff anymore. Which goes to show that no matter what they say, they still aren’t technology companies first and foremost.

II.

The core thing that we [at A16Z] do is track talent flows [in order to predict big market/technology/etc inflection points]. And the thing that we know for sure is that the smartest people in the world, the smartest kids graduating from college, and the smartest industry professionals are flooding into those three sectors [AI, biotech and Crypto/Web3]. There’s an incredible wave of talent in the form of top-end engineers, scientists, executives, and founders flooding into those three sectors. In our world, that’s not completely predictive, but that’s as predictive as you can get.

…

The other set of people we focus on are the really good entrepreneurs. And yes, entrepreneurs respond to fads like anybody else. But when really good entrepreneurs pair with really good engineers, they start companies, they build products, and they tend to make what they’re working on a lot better. So they too are predictive.

IV.

There’s a finite number of super-smart engineers who know what to build. These people go to the companies that take them the most seriously. They go to companies where they think leadership really understands what they do and understands how to build a first-class technology development culture. They go the places where they think they’ll be appropriately rewarded but also where they’ll be taken seriously, listened to, and respected. And they want to be in a place where people like themselves form a critical mass.

The problem that big, classic Fortune 500 companies have is the same problem they had 20 years ago. I thought the problem would shrink over time, but I’m not sure it has. That problem is that the true technologists inside so many big companies are not the primary people at the company. They’re not treated as first-class citizens.

Just look at the org chart. For so long, companies put their technology people in the IT department. The IT department was so famously segregated and isolated that there are entire TV shows, like the great British comedy The IT Crowd, built around the idea of the nerds in the back room. Then, about 20 years ago, big companies got the message that maybe all their technologists should not be in the IT department. So they created what’s typically known as the digital division, typically led by a vice president of digital. The good news is that the programmers run the digital division and are taken seriously there. But it’s still a division. It’s still a unit. That’s a problem.

I’ll give you an example: at Tesla, the engineers working on self-driving cars are the most important people at Tesla. Elon talks about them all the time, he talks to them all the time, and they’re basically the leaders in the company. The people working on that stuff at traditional auto OEMs are not. Maybe they should be, but they’re not. They’re still in this kind of “back room” thing. The people who have led the business for 40 years are the same kind of people who are now in charge.

V.

Q: What do you say to these big companies, Marc? If you were to tell them how to digitally transform, what would you tell them?

Marc Andreessen: Find the smartest technologist in the company and make them CEO.

VI:

… the best technology doesn’t always win. There are all these case studies in business school: Betamax versus VHS, the Qwerty keyboard versus the Dvorak keyboard, Microsoft DOS versus Apple Macintosh, and so on. Example after example of, the best technology doesn’t always win. And sometimes that is true. But a lot of companies and a lot of MBAs take that to mean, “We don’t actually have to be good at technology. We can still win. We can market our way through.”

So, maybe the best product doesn’t always win. But really good tech companies build really good stuff. To compete with them you have to be in the game. You need the world’s best engineers. You have to have a world-leading technology culture to attract them. You need people who really know what they’re talking about to make really good decisions around this stuff. So, at some point, yes, you need to put the technologist in charge. But these companies will not do it. They are no closer to doing so than they were 20 years ago. We’re in some cycle of madness where they keep doing the same thing and expecting different results.

Let me know what you think of this format.

Cheers,

Kiko

This is crazy. It's been more than 6 months since I last posted. Sorry about that - at least I hope you'd like me to write more :) -. Things have been really hectic. We've crossed the 100 Q.Player line at Qulture.Rocks, have been growing like crazy, launched our new brand (a tiny bit more below), and personally, I had a new baby just a few weeks ago, Eduardo! (This is me giving him my first bath last week.)

Anyway, a recurring theme of this blog(?) is me writing an article about a term or a subject that people keep mindlessly repeating out there without, I reckon, really knowing what it actually means. I've previously discussed “disruptive innovation”, “flywheels”, “moats”, and today's subject is one that's very relevant to what we do at Qulture.Rocks: "mission” and, practically, “mission statement”.

(I've previously written about the Qulture.Rocks mission (to unlock the potential of people and organizations) in the context of my falling in love - again - with our mission, but decided to take a step back and discuss what a mission really is.)

Part of the trigger for writing about “mission” was the launch of our new rebrand. The major motivator for our rebrand was us feeling that our brand wasn't representing our mission as it should. We're all about unlocking people's potential so that organizations can unlock theirs, and ultimately achieve their missions, but our brand looked like the brand of a silly little happiness-as-a-service company, not one with such a noble and important raison d’etre.

It looked like this:

And now it looks like this:

By the way, you must check out our manifesto, which, I'm conflicted to say, is amazing!

So on to the article.

The relationship between Mission and OKRs: an important intro

“If people do not internalize the organization’s mission and vision, they will not use them to make day-to-day decisions, and if they do not use them in their daily lives, all the effort will have been in vain.”

–**Pete Babich** [1]

We’re big fans of OKRs (when done right) and believe that ultimately OKRs should come from an organization’s mission (that’s why, after all, we’ve named our OKRs book “OKRs: From Mission to Metrics” [2], which you can read for free on Amazon – and of which this article is an excerpt -). Therefore, we thought it would be important to spend some time properly defining the concept of mission (and a mission statement) and clarifying a very common mistake which is to confuse mission and vision, which is another important part of our OKR framework but that serves a different purpose.

OKRs are a strategy execution tool, that falls in a greater discipline in management that encompasses strategy formulation and strategy execution. Ideally, one shouldn’t exist without the other.

Strategy formulation always starts with the company’s mission statement. It should explain why the company exists, which is itself an interesting limiting factor that helps an organization figure out the boundaries of its footprint in the world. As we’ll see below, mission and vision are two sides of the same coin: whereas the mission describes why the organization exists (in a nutshell, a problem the organization wants to solve,) the vision describes a future world where the mission has been accomplished.

After the mission is considered, the next step is actually formulating the strategy of the company. Strategy is a very tricky term to define – we haven’t found a definitive answer we’re happy with – but a great way to think about it is as a series of decisions and tradeoffs that will maximize the chances of the company fulfilling its mission and achieving its vision. As we formulate the strategy, we can articulate a series of visions – future states of success – that are to be achieved along the journey. You may have heard, during the last decade, of companies’ 2020 Visions. These look different from a mission statement – it’s not so qualitative and inspirational, but more detailed and even quantitative where possible. An organization can have a 10-year vision, but also a 5-year vision, a 1-year vision, and even a vision for every quarter (we’ll talk about short cycles in a bit.) The important thing to consider about vision statements is that they articulate end states.

The visions and strategy go hand in hand: think about the strategy as the *why* behind the visions. For example, a Mexican company may have a 5-year vision of dominating its national market, a 10-year vision of dominating theNorth American market. In this context, strategy is an explanation of why the company chose to focus on the continent, and not expand to other emerging markets. Finally, there are OKRs, which help an organization articulate the gaps between its current state (let’s say, today) and its vision(s). OKRs are all about the critical few gaps that the company must close to get there, articulated as groupings of Objectives and their Key Results.
So to answer the question posed at the title of this chapter, an organization’s OKRs come from its mission, its strategy, and its visions/milestones.

Anyway, this is not an article about strategy formulation, nor about the broader discipline of strategy execution. It’s about the mission, and as you’ve probably noted at the start of this chapter, we think a mission has to do with why the organization or company exists. In the next lines, you’ll understand how we got there.

Towards a definition of organizational mission

At Qulture.Rocks we read many Silicon Valley pundits talking about mission statements and are amazed by how little clarity and how much confusion there is around what the term really means, why it exists, and how companies can leverage it the most.

Pundits (a general term for whoever thinks they know enough to try to define these terms) rarely agree on what mission is or what great specimens look like. As an example of how confusing it is, let’s look at what Jim Collins, the revered business author, says about the subject.

One of his most referred-to articles on the subject is “Building Your Company’s Vision,” which was published in the Harvard Business Review in 1996 and went on to become the core idea behind *Good to Great*, Collins’ very famous best-seller. In the first line of the article Collins and his co-author Jerry Porras write that “Companies that enjoy enduring success have a core purpose and core values that remain fixed while their strategies and practices endlessly adapt to a changing world.” I hope you can see why I’m citing this as an example of why it’s so hard to understand what mission and vision actually mean: in an article that’s about vision, Collins has a first-line talking about **core purpose**, which is something very close to the concept of a mission, and **core values**. Of course, they then try to clarify it a bit, and go on to say “[vision] has two principal parts: core ideology and envisioned future.” But hey, can we be a bit clearer?

To look a bit further, we Googled “how to build company vision.” On the first results page, we stumble upon an article on the Openview Venture Partners – a respected venture capital firm – website, written by portfolio CEO Firas Raouf. In the article, Raouf defines mission as “what a company is striving to be in the long term” and vision as “how it can get there,” asking “what things need to be executed to accomplish the mission?” This second excerpt is way worse, of course, and confuses more than only mission and vision, but also strategy and planning, but serves as a great example to further our point. Let’s move from pundits to the websites of large tech companies: Google and Amazon.

Google’s mission statement, for example, is “… to organize the world’s information and make it universally accessible and useful.” If we unpack it, there’s a first part that talks about how Google makes an impact on the world (organizing the world’s information and making it easier to retrieve), and a second part that talks about how Google wants the world to look (a world where information is organized and universally accessible and useful). By the way, the website (and the first results page for Google search for “Google vision”) has no mention of there being a vision. So it seems that what they call “mission” is actually a mix between a mission and a vision.

In talking about its mission, Amazon says “we aim to be Earth’s most customer-centric company. Our mission is to continually raise the bar of the customer experience by using the internet and technology to help consumers find, discover, and buy anything, and empower businesses and content creators to maximize their success.” Amazon doesn’t clearly define what is its actual mission statement, but when talking about it mixes what the company wants to become (“Earth’s most customer-centric company”) and why it exists (“to continually raise the bar of the customer experience…”).

It seems like these two giants don’t speak – at least to the outside – about their visions. They only talk about their missions. Now, if we read their missions, it’s still very hard to abstract a consistent pattern on what a mission should look like.

Now we go from the tech giants to the “old economy” for reference, and we get even more confused. Koch Industries, for example, defines its vision as: “Koch Industries is a trading, investment, and operating company that aggressively identifies and acquires companies in which it can leverage our strengths to generate superior earnings or market value.” Their mission is “Koch Industries seeks to maximize the present value of future profits. Doing so provides security and opportunity for stockholders and productive employees, while also benefiting customers and society….”

Now, unpacking Koch’s mission and vision is a trip. Their mission basically describes what business the company is in (buying out other companies for cheap, but companies where they have an edge). Their vision, on the other hand, describes what benefit the company wants to bring to its stakeholders (security and opportunity for stockholders and employees, and less explicit benefits to customers and society).
As you have probably inferred, after reading about how pundits, big tech, and the “old economy” talk about mission and vision, it’s impossible to deduce what these terms actually mean, and how a company should use them.

A startup-focused definition of mission

If you’re feeling more confused than when you started reading this article, that’s exactly how we felt when we tried to understand what an organization’s mission (and vision) statement looks like.

After this long journey, which included countless other references, we’ve settled upon the following definition:

The *mission* is the company’s purpose. It’s why it exists.

It helps to think, as Jim Collins suggests, “how would the world be negatively impacted if our company ceased to exist?” At Qulture.Rocks, for example, our mission is to “help people and organizations unlock their potential.” Google’s mission statement is almost there.

The *vision*, on the other hand, is how the world will look like if the company fulfills its purpose. Mission and vision statements are two sides of the same coin, and we think that this is a core reason why definitions are so confusing. Anyway, broad vision statements are pretty useless, so we recommend you stick with the vision for the purposes of defining what the organization is and what its boundaries are. More specific and time-bound visions, on the other hand, are very useful as a way to articulate what future end-states look like for the organization, and we’ll get into them in a bit.

Company x customer x shareholder-centric missions

Using broad strokes, the company’s mission can be company-centric or customer-centric. Google’s mission (or the part of it that looks like a mission by our definition) is customer-centric. If a customer reads it, she immediately relates the company’s impact on her own life. Koch’s mission, on the other hand, is company-centric. It basically talks about how the company makes money in very practical terms.

We believe missions should be as customer-centric as possible, and more abstract in nature. At Qulture.Rocks, for example, we’re not talking about anything other than the impact we were created to have on the world.

Another good guideline is to not mention your line of business in your mission statement. Paraphrasing Simon Sinek, it should be more about the *why*, and less about the *how* of your company’s impact. The how is more about strategy and tactics: how the company chooses, today, to bring that impact to the world. However, that can change if it ceases to be the most efficient way to do it.

For example, we don’t cite software or technology in our mission statement. That’s because we’re about cultures that rock, and not about software. Software is how we choose to pursue our mission, but it can change to another thing (content, consulting, wearables) if we think that would cause more impact.

Back to Google and Amazon

Now that we’ve – hopefully – agreed on the definition of a mission and vision statement, let’s go back to Amazon’s and Google’s mission statements and see how they pass our test.

Again, Google’s mission statement is “to organize the world’s information and make it universally accessible and useful.” It’s kind of a blend between a mission and a vision. The first part looks like a mission: Google exists to organize the world’s information and make it easy to access. However, it gradually blends into a vision, because it gives us a vivid description of what the world will look like if they’re successful: a world where all information is easy to access and useful. We’d give it a 7.

Amazon, on the other hand, has a worse mission statement. As we saw earlier, it reads “to be Earth’s most customer-centric company, where customers can find and discover anything they might want to buy online, and endeavors to offer its customers the lowest possible prices.” That looks much more like a vision statement to us, and quite a company-centric one, for that matter. It describes a future, but in terms of what the company will look like in the future, and not how the world will look.

Notes:

[1] I recently learned that this is called an epigraph.

[2] Drawing heavy inspiration from Ali Rowghani from Y Combinator in his “The Second Job of a Startup CEO“ [3]

[3] OKRs: From Mission to Metrics can be found in its latest version here and for free. Bear in mind that it's a work in constant progress. The good thing is that if you download it via Lean Pub, you'll be notified whenever I release a new iteration.

“Disruptive” innovation

The term "disruptive innovation" is widely used and often misunderstood. People typically associate the term with the creation of something new, like a radically new product, such as the Oculus Rift. Others think of it as whenever an incumbent firm gets its lunch eaten by a new competitor, particularly if said competitor is seen as a "tech company." However, neither definition is correct.

The term - and the model that gives substance to the term - was created by Clayton Christensen in his The Innovator’s Dilemma to mean the process in which a new entrant enters a market (oftentimes a weaker company, with a smaller war chest, less technological prowess, and so on) by offering a simpler, cheaper version of some incumbent(s)’ product or service, and ends up, after some time, totally displacing the incumbent's comfortable initial market position.

There's a lot to unpack here.

Incumbents and sustaining innovation

The first important ingredient for disruptive innovation to flourish is incumbents trying to outdo each other with incremental sustaining innovations. Sustaining innovations are those that “improve” a product along an established angle. According to Christensen himself,

Sustaining innovations are what move companies along established improvement trajectories. They are improvements to existing products on dimensions historically valued by customers. Airplanes that fly farther, computers that process faster, cellular phone batteries that last longer, and televisions with incrementally or dramatically clearer images are all sustaining innovations.

Faster or longer-range planes, faster computers, smaller phones, software that's more secure, faster, with more functionality, are all, et ceteris paribus, incremental innovations.

In most markets, competitors will try to outdo each other in terms of sustaining innovations in their never-ending quest for differentiation and pricing power which, when successful, lead to higher prices and better gross margins.

(It's important to note that some markets have even worse dynamics where products get better (along said historically valued dimensions) every year but prices go down every year. Think about flat-screen TVs, for example, that get thinner, smarter, more bright, higher resolution, but also much cheaper, or at least not more expensive. We'll ignore these markets for the purposes of disruptive innovation.)

Attackers and disruptive innovation

So on the one hand we have incumbents fighting each other by trying to improve their products around well-known product attributes. Products get ever more sophisticated and expensive.

What inevitably then happens is that some portion of the market's customers become overshot (a critical concept of the theory). For this portion of the market, products end up becoming too complex and too expensive. They don't really need all that functionality.

That's the opening a disruptive innovation needs to do its thing.

Disruptive innovators then enter this market by offering a cheaper, simpler product or service that is good enough (another key concept of the theory) for the overshot portion of the market.

What makes disruptive innovations so wicked is what comes next.

You might think “well, all pretty simple: why doesn't the incumbent build an offering to compete on the overshot segment of the market?” And the pernicious answer is “because it's not the ‘rational’ thing to do.” The ‘right’ thing for an incumbent to do given such a competitive setup is to give up this overshot segment because said customers don't value the stuff that's most dear to the incumbent: the innovative features it has worked so hard to introduce. Not only that but offering something simpler and cheaper will often cause a deterioration in margins and cannibalization of the incumbent's brand.

Therefore, most incumbents that are faced with disruptive innovation in their markets will usually ‘rationally’ retreat from the more overserved low-end of the market and double down on the less overserved high-end of the market. And that's their death kiss. The disruptive attacker enters the market by catering to formerly overshot customers but will start introducing sustaining innovations that will gradually eat its way up the market until the incumbent is totally displaced or relegated to a niche.

The fact that ‘rationality’ will almost always drive incumbents to their demise is what makes disruptive innovation theory so interesting. Again, Christensen in Seeing What's Next:

The theory holds that existing companies have a high probability of beating entrant attackers when the contest is about sustaining innovations. But established companies almost always lose to attackers armed with disruptive innovations.

Enterprise software: an incredibly fertile ground for disruption

Enterprise software is an incredibly fertile ground for disruptive innovation. Let’s take the example of Salesforce to illustrate that assertion.

Salesforce’s CRM offering goes much further than what a “simple” CRM is expected to do: aside from storing prospective customer information and allowing for a pipeline of deals to be managed, it has all sorts of bells and whistles, such as industry-specific reports, complex workflow automation, AI, an app store, and LOTS of flexibility that cater to all different sophisticated customer needs; it costs a lot and takes long to be implemented. And Salesforce keeps adding incremental sustaining innovations to the product every year and causing a bigger and bigger portion of the market to become overserved.

If we analyze Salesforce as the incumbent to be disrupted, one of the most obvious candidates for the role of the disruptive attacker would be Pipedrive: a simpler, cheaper version of Salesforce’s CRM offering that gets “the job” done just fine for a great number of customers, but for much less of a - time and money - hassle.

(Here, the concept of a “job” is important: a job, or job-to-be-done, is the actual progress a customer can make given a set of circumstances. Sales managers probably need more process, organization, and data so, therefore, they buy a software product and some services to do just that. And that product is usually a CRM system.)

If we take into account that Salesforce’s CRM is a multi-billion dollar business, there are opportunities for new entrants to build simpler, cheaper solutions that get the basic job done for half of the price, and half the time-to-value, and still, build huge companies in the process.

Watching a keynote by Stripe founder Patrick Collison the other day, I stumbled on the following passage:

This is really key a very important part of the stripe strategy is we're building for businesses of every size we're building for the highest potential graduates of accelerators around the world and we're building for some of the biggest companies in the world as you all know the sort of standards pattern in enterprise software for some innovative product to come along you get kind of pulled up market it gets a bit kind of longer the tooth a bit stagnant a bit a bit calcified and then in turn it's replaced by some you know nimbler and more agile upstart this is the software cycle of life this of course really bad for the customers who get stuck using the old version and so this is why we obsess over startups startup to some of those demanding customers in the world they can't tolerate complexity and they simply won't put up with outdated technology and so our strategy is very deliberately to serve both ends of the continuum and every point in between this ensures we can provide the most powerful functionality to the youngest companies in the world and that we can provide the most forward-thinking technology to the largest and the most established.

What Patrick is saying, essentially, is that by continuously focusing at least part of their efforts on the startup market, they will decrease their chances of being disrupted by a new entrant to their market.

Smart guy, isn't he?

Disruptive innovation in repeat mode

The most interesting thought experiment I've come to grapple with lately is that there seems to be no proper equilibrium in some more disruption-prone markets, such as enterprise software.

Even if no new technology paradigm comes into existence in a given application area, players will always innovate up-market until a portion of it is overshot and an up-and-coming disruptor arises. And said disruptor will go on to become the market “leader," only to predictably create a new wave of overshot customers and a wedge for a new disruptor. On and on and on and on. Am I crazy?

Of course, some factors may absorb this propensity. Most of what I can come up with seems to have to do with moats (a subject I explore extensively in this article). Salesforce, for example, doesn't seem to be undergoing a violent disruption, and I would guess a big reason for that is its brand, the (sort of) network effects that arise from its vast third-party developer ecosystem and its implementation partner network, and also the fact that most of the market's growth comes from the conversion of non-consumers into consumers, which gives players like Pipedrive (and Hubspot and RD Station and Base, only to name a few) a long runway to go until they start bumping into Salesforce.

Theodore Roosevelt in The Man in the Arena

I have been fascinated by the Slack x Microsoft Team s competitive battle for some time. I felt like I was witnessing one of the great competition stories of the enterprise software market firsthand. This week's acquisition of Slack by Salesforce jerked me back into thinking about it, and I decided to write a follow-up note in part to brag about past predictions and in part to build on them.

My previous take on Slack's prospects as a long-term investment

I'm not sure if you'll remember, but in June 2019, I wrote on this very same blog that I wasn't really bullish on Slack stock as a long-term investment. I'm not one to brag, but I feel that this has been one of my good calls (there are plenty of bad ones I shall not discuss). Back then, I concluded my piece by saying:

"The Microsoft threat is so big that I wouldn't be willing to hold Slack stock on the long run. I wouldn't buy it if I were Warren Buffet anyway."

Looking back, I think my thesis could have been 10x more well developed. Anyway, my reasons for that conclusion were mainly based on the strength of Slack's main competitor: Microsoft and its Teams offering. The supporting arguments were pretty straightforward.

The first was Microsoft's unbeatable partner ecosystem that would help it gain scale at a much faster pace, both in large enterprises as in middle-market enterprises, which were critical segments for Slack's growth. I wrote:

"Selling to the enterprise is a pretty specific muscle, that's hard to train. Microsoft has nailed it by cultivating a direct sales force and a huge channel partner ecosystem that sells and implements their products worldwide. Slack, which until only months ago boasted not having any salespeople, is still a toddler on that front. Slack is still a company hoping to be a long term success with a killer product; Microsoft has already gotten the fact that it is the distribution channel that matters in the end."

The second was that Teams was natively integrated with Microsoft Office, a staple everywhere outside of tech. I wrote:

"The point is: Microsoft and its Teams product is an incredibly strong contender on this fight. It has a huge share of wallet on basically all relevant enterprises globally. So a very relevant question would be how much better is Slack vs. Teams so that a CIO (and this is a CIO type of purchase) will buy in instead of just plugging Teams in its stack? Alas, Teams has much better Office integrations, which are still a huge part of “where work happens” [1] on the enterprise."

Both of these points would be made a bit over a year later by Ben Thompson in his Stratechery article called "The Slack Social Network." I am reasonably proud of that, him being one of the most well-read and well-articulated tech and strategy writers out there, even though I'm not one to brag, as I said.

Ben Thompson called Microsoft's partner ecosystem "a gargantuan moat":

Microsoft’s partner network is a truly gargantuan moat. When it comes to enterprise, it is easy to focus on the biggest companies, where Microsoft will engage directly, and challengers like Slack can build up sales forces to compete. Underneath those companies, though, are tens of thousands of smaller businesses that, even if they have IT directors of their own, rely on outside vendors to build up their technical infrastructure.

And mentioned the Microsoft suite of products and its native integration with Teams as a huge reason for Microsoft's win over Slack (to be very honest, Thompson was basically saying that Teams had already beaten Slack, and not that it was likely to beat Slack.):

This is what Slack — and Silicon Valley, generally — failed to understand about Microsoft’s competitive advantage: the company doesn’t win just because it bundles, or because it has a superior ground game. By virtue of doing everything, even if mediocrely, the company is providing a whole that is greater than the sum of its parts, particularly for the non-tech workers that are in fact most of the market. Slack may have infused its chat client with love, but chatting is a means to an end, and Microsoft often seems like the only enterprise company that understands that.

Having Ben Thompson repeat my arguments a year later already felt good. But Slack's stock price also pointed my way, even though Mr. Market is clearly not a very trustworthy arbiter of truth. After its direct listing, at around $38 bucks, it was trading at $25(ish) a year later (and days before rumors about a Salesforce deal started to emerge).

Curiously, Slack's stock didn't benefit from the tailwinds of COVID-19 and the remote work push we all got. If we compare its performance against Zoom's, we get a really bleak image, or shall I say graph?

My take is that Mr. Market was putting more weight on Microsoft's competitive pressure than on the supposed tailwinds that a more remote world would bring to Slack's back. And I think it was right in doing so. Before the listing, and as I pointed in my earlier article, a lot of Slack's growth had come from enormous enterprises. But lately, as noted by Nathan Baschez, Slack's growth among huge companies (I am guessing (year-on-year (run-rate ARR) growth) from large customers) was falling a lot.

That means Slack was having trouble selling into larger companies, probably because, as I predicted, CIOs preferred the cheaper-for-MSFT-customers, better-integrated Teams (over Slack).

There is one final argument supporting, at least in part, my argument for Teams' success: the sharp growth in its user base, especially relative to Slack's. Check out this chart from Statista:

Twelve million DAUs was the last stat published by Slack before the pandemic. Teams had just released it had already blown past Slack and most recently published an even more whopping number of more than 100m DAUs, whereas Slack remained silent ever since.

Now, I'm totally aware of how bad this argument is in a vacuum. First, because Microsoft gives Teams away for free for a bunch of Office subscribers. Second, because it is rumored to install Teams even though the customer may not have consciously clicked "install," and even worse, it is rumored to be a *very* difficult app to remove from your computer. Third, because Teams has all the voice and video capabilities developed by the MSFT acquisition of Skype, it is just fundamentally hard to compare with Slack's chat proposition.

I'm also aware that even though it looks terrible, Slack's decision not to publish DAUs anymore makes sense. DAUs x DAUs were probably comparing apples to oranges (Slack's engagement was probably much higher), and it was a metric on which Microsoft would look better no matter what. So why pour more fuel on the fire pit?

On the other hand, as flawed as these data are, they do reinforce my main point: Microsoft's distribution can make Slack's life really tough. First, because everybody uses Office, so being natively integrated with that is a plus. Second, Microsoft's huge partner network allows worldwide penetration to the enterprise ex-Fortune 1000 (where a lot of Slack's growth came from, but as we saw most recently started to slow). And I think solid growth on the (global) enterprise market is what would justify Slack's pretty rich valuation at the time of the listing (not to mention a reasonably healthy performance going forward,) and its absence, apparently perceived by Mr. Market, is what hammered its price performance.

Revising my thesis

As I said, I didn't really like my previous piece on Slack's prospects. It could have been 10x better, even though it had some good arguments in it and a conclusion I still kind of believe in. The part that was reasonably well developed was the reason I believed Microsoft Teams would beat Slack. The part that was very poorly developed was at what game such a beating would happen. In other words, what was Slack pursuing, and why that pursuit would be unfruitful because of Teams.

Slack's vision

I'll never know for sure what was Slack's vision for its success, but I'd guess at some point, Stewart, his team, and his investors believed Slack could become something like an enterprise super app. By enterprise super app, I mean a front-end to everything that goes on inside an enterprise, such as business apps (CRM, HR stuff, marketing automation, ERP, et al.).

The reason for my belief is many-fold: whoever is going to become the enterprise super app has to be a high engagement, company-wide app to begin with. And communications, such as email and now chat for a growing number of companies, are probably the most horizontal, highest engagement apps, at least in the western world.

Also, because Slack was showing behavior that pointed in that direction. It was trying to be a hub of integrations, such that Google Apps, Office, even Qulture.Rocks, had widgets that allowed said other apps' workflows and even files to be manipulated and/or consulted from inside Slack.

Enterprise super app being a hugely aggressive ambition, Slack's drive to transform its app into a platform, on top of which developers could build proper apps (and a marketplace that went with it) made sense.

Finally, Slack was expanding a lot away from chat. It built voice messaging. It built Slack calls. I think at some point, it even built video calls, just like the ones we do on Zoom or Google Meets.

The most important point, though, is that to really become the global enterprise super app, Slack had to become the global enterprise communications standard. It couldn't be a niche player. It had to become a $100 billion-plus company. Not a decacorn. A fucking crazy story, like what-the-fuck-is-going-on-here story. And because the company's listing was done at $38 (i.e., that would be my entry point), that's a very probable type of outcome that an investor in Slack for the long run (something I wasn't willing to be) could have been chasing.

Enter Microsoft Teams. Why I think Teams would beat Slack was something I've already elaborated upon. Distribution and Office.

The bullish case: The social network argument

Even though Ben Thomson was bearish on Slack's competition against Teams overall, he was bullish in two aspects. He thought Slack had a chance to win by a) focusing specifically on chat and b) creating a cross-company network of Slack instances (something that Slack had introduced a bit earlier called Slack Connect. With Slack Connect, two companies can essentially create joint channels where people from both companies can talk. Even more, individuals from both companies can DM each other as well. I wasn't very convinced).

Thompson's argument wasn't that Slack could win head-on, but that Slack could win by focusing on chat and pursuing the multi-company angle, while Teams could keep focusing on being "the front-end of all your apps in the cloud." He said:

Slack Connect is about more than chat: not only can you have multiple companies in one channel, you can also manage the flow of data between different organizations; to put it another way, while Microsoft is busy building an operating system in the cloud, Slack has decided to build the enterprise social network.

Stewart Butterfield, CEO of Slack, echoed this strategic angle, in an interview he would later concede to the very same Stratechery. First, he exalted the importance of chat:

There are two things I think that are worth talking about that I’d push back on. One is just the word “chat” because it indicates that this stuff isn’t important. If you’re a manager, executive, leader, more or less, a hundred percent of your job is communication. If you average it out across the entire company, something like 50% of people’s time will go into really basic acts of communication and coordination, so I don’t mean like interesting strategic creative conversations. I mean daily standups and status reports and basically every bit of effort that goes into the creation of a presentation that’s going to be shown at a meeting because the point of the meeting is just to tell everyone the same information so everyone has access to the same information and it’s calibrated. So if that’s what people are spending their time on, it is important. I can’t remember the exact line, but something like, “if your job is to chat all day, then Slack would be a good choice.”

Second, I don't remember where it was, but I remember reading some quote about Stewart downplaying Teams as a competitor. Yes. The guy who took the whole page ad on the New York Times to "welcome" his new frightening competitor started to change the narrative and say Slack was just another thing. It was very focused on chat, whereas Teams was another thing.

I could agree with Thompson that a narrower chat focus could be a good thing for Slack. Qulture.Rocks uses Slack, and I don't see us moving anywhere else. BUT, I don't think our enterprise customers will use Slack for the time being.

Most importantly, I don't think the Slack Connect thing can reverse the trend of losing to Microsoft Teams in the global enterprise arena.

I still think that IF this enterprise super app becomes a network effects play, Microsoft's big distribution advantage (which is its biggest advantage, to begin with) will give it an even greater chance of building said network. A bit after the post, I commented on the Stratechery forum:

IMHO, if Slack succeeds in making the shared channel a thing in the category (which will make Microsoft further copy it with all its horses and men), it will probably (and inadvertently) make its own competitive position even weaker in the future, since as you said Microsoft will just become much bigger (free, the partner network, and so on), and thus a much better “social network.” If there’s one company that can win in a network effects play in the enterprise it’s Microsoft with its Office penetration. WDYT? What am I missing?

I'm not saying Slack Connect isn't an amazing experience. It is. I experienced it firsthand a little while ago when one of the funds that invest in Qulture.Rocks set up a shared channel between themselves, myself, and all other founders of all other portfolio companies. Pretty seamless shit. I can essentially hop in that channel without changing instances. I can even DM somebody that's on that channel as if this person worked at Qulture.Rocks. It's really, really good.

My point is that it doesn't really help Slack become a global standard. It probably generates network effects within specific sectors like the tech + VC intersection, I won't argue. But how much of a return on a $25 billion valuation can you get by being the chat app of choice of the tech + VC ecosystem? Not sure. At gunpoint, I'd bet not much.

Slack and Salesforce

Anyway, this week Salesforce announced it was acquiring Slack for $27 billion (or was it Slack announcing it was being acquired by Salesforce? Anyway...). You might argue that a $27 billion is not too shabby, but let's note that it was clearly not the outcome the company was looking towards as it listed its shares for more than that a year ago. It was clearly trying to become something HUGE (we've already touched on that).

But I guess a year later, Stewart Butterfield was tired.

There are many reasons why he could be tired.

For one, he could have become tired of the fight against Microsoft (not judging - I would probably do the same if it were me) for large enterprise company-wide deals (they talk about the wins but don't really talk about the losses).

He could also have been very discouraged by the stock price gap between Slack and Zoom.

He could have been bearish about how Slack's stock might perform in the short to mid-term and the potential effects of such price action on company morale and customer buying decisions.

The point is, he decided to sell. And to sell for less than what the company was worth when it was listed in the first place. That could mean many things, but being bullish on his company's prospects could not have been one of them.

And, he sold to Salesforce. At that size of a deal, you don't have many potential suitors. He could have sold to Oracle, Google, SAP, Zoom, Amazon, Salesforce, and maybe a handful of other companies. Not Microsoft, because of antitrust concerns. But my main point is that none of these companies is really a dream home for your baby. Benioff attends calls from Hawaii on top of an elliptical (as relayed to me not only by some bulge bracket publication I can't remember but also by an enterprise software CEO with many years at Success Factors.) Oracle would frighten the shit out of any Slack employee - too frat of a culture. SAP kind of the same, but less colorful. Zoom is weird. Amazon is in Seattle. Google is evil (but might have been the best fit - was it frightened by its own antitrust worries?). You get the picture.

I'm not really going to go into the discussion of what were Salesforce's motives were, but I'm trying to reinforce the point that selling Slack to such companies was probably not a dream come true to Stewart. To reinforce the point that his stance for Slack was probably not that bullish. To reinforce the point that I was probably right in the first place, almost a year and a half ago.

Cheers,

Kiko

Notes and bibliography

https://unlockingpotential.kikomello.com/p/would-you-be-a-slack-shareholder

https://sec.report/Document/0001628280-19-004786/

https://www.theinformation.com/articles/slack-ceo-downplays-microsoft-as-factor-in-salesforce-deal

https://www.theinformation.com/articles/slack-ceo-downplays-microsoft-as-factor-in-salesforce-deal?utm_content=article-5093&utm_campaign=article_email&utm_source=sg&utm_medium=email

https://stratechery.com/?s=slack

https://www.nytimes.com/2020/12/02/technology/marc-benioff-salesforce-slack-microsoft.html