Why low-competition ecosystems discourage innovation

Episode 1 — Fear is not a character trait. It is a price.

I come from Congo-Brazzaville. A country where many people have ideas, but many hesitate to start a business for fear of having them stolen.

I write this sentence without bitterness and without intention to denounce anyone. I write it because it seems to me, on reflection, poorly understood — including by those who utter it. It is heard as a complaint. I propose to hear it as a datum.

Widespread hesitation is not a lack of courage. It is a behaviour. And behaviours repeated across an entire population are rarely read as character accidents. They are read as responses to a structure. When thousands of people make the same choice, it is more fruitful to look at what is in front of them than at what is inside them.

That is the bet of this series: to treat the fear of starting a business as an economic signal, not as a cultural trait.

The cultural explanation explains too much, and therefore nothing

There is a ready-made answer, and it circulates widely: "among us, people don't trust each other." The problem with this explanation is not that it is false — mistrust exists, it is measurable, it has effects. The problem is that it explains too much.

A variable that accounts for everything predicts nothing. If cultural mistrust explained entrepreneurial torpor, we would have to explain why the same Congolese person, settled in Paris, Montreal or Johannesburg, registers a company, partners with strangers, signs confidentiality agreements and takes risks they would not have taken in Brazzaville. We would have to explain why Estonia, emerging from a regime where mistrust was a survival reflex, produced in twenty years a density of startups per inhabitant that few European countries reach. We would have to explain why northern Italy and southern Italy, culturally close, do not have the same business creation dynamics.

The individual moves. Their culture follows. Their behaviour, however, changes. So it is not primarily the culture that changed: it is the game.

I therefore prefer a narrower and more demanding hypothesis: the fear of having an idea stolen is a rational estimate of the expected return from disclosure. It is high where disclosure costs a lot and returns little. It is low where disclosure costs little and returns a lot. The rest — temperament, boldness, confidence — comes after, and often because the prices have moved.

The founding paradox: you cannot sell an idea without giving it away

Kenneth Arrow formulated it in 1962, in a text that remains the starting point of all innovation economics. Information has three properties that make it economically odd.

It is non-rival: if I give you my idea, I have not lost it, we both have it. It is cheap to reproduce: what took two years of trial and error can be copied in three months by someone who now knows it works. And above all, it suffers from a disclosure paradox: to convince someone to buy an idea, you must show it to them; once they have seen it, they no longer need to buy it.

This paradox is not Congolese. It is universal. Every founder, in San Francisco as in Pointe-Noire, must open their hand to raise funds, recruit, approach a distributor, respond to a tender. Innovation is an activity that forces you to reveal yourself.

The interesting question is therefore not: why are Congolese entrepreneurs afraid to disclose? They are afraid for the same reason as everyone. The question is: why does this same paradox produce, elsewhere, a functioning market, and here, paralysis?

What neutralises the paradox elsewhere

In dense ecosystems, three mechanisms defuse the disclosure paradox. None is magical; all are institutional.

The first is repeated reputation. A Sand Hill Road investment fund sees several hundred pitches a year and must continue to see them. If it copies a project that was presented to it, the information circulates within weeks in a narrow circle, and its deal flow dries up. Theft is punished not by law but by the repetition of the game. Where interactions are few and unobserved, this mechanism does not exist: you cannot lose a reputation in a market where you play only once.

The second is the existence of an alternative buyer. If only one company in the country has the means to monetise my idea, it sets the price — and the price it can set includes the option of paying nothing at all. If there are twelve, they compete to acquire it, and the value of my idea ceases to be fixed by my weakness and is instead fixed by their rivalry. That is the heart of this series, and I will return to it at length.

The third is the ability to run faster than the copier. A copied idea is not a lost idea if the inventor has the capital, the team and the channels to build a three-length lead while the imitator assembles theirs. The real protection of the innovator is almost never legal. It is kinetic.

These three mechanisms have one thing in common: they do not depend on the virtue of the actors. They depend on the structure of the market. They can be built.

The thesis I will defend over seven episodes

I state it here in one sentence, and I will spend the next six episodes supporting it, qualifying it and testing it against its best objections.

In a competitive ecosystem, the least costly path for a dominant company to obtain an innovation is to buy it. In a low-competition ecosystem, the least costly path is to wait, observe, then rebuild. It is not the same company behaving differently. It is the same rationality encountering a different arithmetic.

Put this way, the thesis displaces the problem. It ceases to be a morality trial and becomes a question of architecture. And an architecture, unlike a morality, can be modified by public policy, taxation, financing structure and public procurement.

I want to be clear on one point, because it conditions the reading of the six texts that follow: a company that rebuilds a product rather than acquiring it is not cheating. It is responding to prices. If the price of waiting is low, if the price of imitation is low, if the price of acquisition is high relative to what it would obtain by waiting, then rebuilding is the decision that any competent investment committee would recommend, anywhere in the world. Blaming a manager for following the incentives presented to them is misidentifying the target. The incentives have an author: they are institutions, and institutions are decided.

The cost nobody counts

There is one last, more disturbing reason to take this question seriously.

When a startup gets overtaken, you see something. There is a company, founders, a product, a truncated trajectory. It is visible, it is narrable, and it is what people talk about.

What you do not see is the thirty-year-old engineer who has had a project in a notebook for two years and will never bring it out. They do not go bankrupt. They do nothing. They appear in no statistics, no business register, no incubator report. Their renunciation produces no trace.

Yet that is where the bulk of the cost lies. An economy does not suffer primarily from projects that fail — failure is the normal mode of innovation, even in the most performing ecosystems, where the vast majority of funded startups do not return their capital. An economy suffers from projects that do not start, because those produce neither learning, nor trained talent, nor created supplier, nor spurred competitor, nor next idea made possible by the previous one.

It is an invisible cost, and that is precisely why it is never weighed.

The itinerary of the seven episodes

This series is a single line of reasoning, cut into seven parts.

1. Fear as signal — this text: framing the problem as a question of incentives, not of morality. 2. Wait or buy — the arithmetic of the decision, from the dominant company's perspective. 3. Why Silicon Valley buys — what venture capital, speed and network effects do to the price of waiting. Comparisons with South Korea, Estonia, China, Europe. 4. The cost of what is not born — the psychology of renunciation, the option of waiting, and the lost externality. 5. 'An idea is worth nothing, only execution counts' — the most serious objection, taken seriously. Where it is right, where it ceases to be. 6. Changing prices — local capital, public procurement, taxation, intellectual property, open innovation: what actually affects incentives, and in what order. 7. What I would do with a year and a budget — the synthesis, and who bears the burden of proof.

I am not claiming to describe an African fatalism. The mechanisms I will discuss produced the same effects in 1990s Europe, in 2000s Japan, and they are producing vigorous debate in the United States itself, where people wonder whether platform concentration has begun to discourage the entry it was funding yesterday. Congo-Brazzaville is not my subject: it is my observation point.

Episode 2 on Tuesday: why, on paper, waiting can be the most rational decision a large company can make.

Episode 2 — Wait or buy: the silent arithmetic of dominant companies

In the previous episode, I stated a thesis: when a large company sees an innovation, it chooses between buying it and rebuilding it, and this choice depends less on its morality than on its arithmetic. I now want to do that arithmetic. Seriously, and from its point of view.

Because that is the blind spot of most conversations on this subject. We reason from the founder's perspective, who has everything to lose, and never from the perspective of the executive committee facing a capital allocation decision. Yet it is that decision that structures the ecosystem.

The decision, as it actually arises

A startup launches a product. It works. Not enormously, but enough for an industry executive to notice and ask their teams for a memo.

That executive has four options, and only one is ever discussed publicly.

Ignore. The product does not threaten its core business, or the market is too small to justify a line on the P&L. Statistically, this is what happens most often, everywhere in the world. Most innovations die of indifference, not predation.

Acquire. Pay to immediately obtain the product, the team, the customers, the accumulated learning — and remove the asset from the market.

Rebuild. Fund an internal development of an equivalent product, relying on its own resources.

Wait, then rebuild. Let the startup spend its capital proving that a market exists, educating customers, discovering what works — then rebuild, at lower risk, once the uncertainty has been lifted by someone else.

The fourth option is the most interesting, because it is the most rational under certain conditions and the most destructive for the ecosystem in all of them.

What imitation actually costs

The immediate objection is that a large company does not have the skills to rebuild an innovative product, and that it will take too long. This is sometimes true. It is less true than people think.

Edwin Mansfield's work on imitation costs in American industry established a now-classic order of magnitude: imitating a product costs on average significantly less than inventing it, and takes significantly less time. The reason is simple and structural: the imitator does not have to pay for uncertainty.

The innovator pays for three things the imitator does not.

They pay for dead ends: features developed then abandoned, customer segments tested then discarded, pricing models that did not work. These expenses do not appear anywhere in the final product, but they often constitute the majority of its real cost.

They pay for market education: convincing the first customers that the problem is worth solving, and that it is solved this way. This is the most expensive line item and the most generously handed to the next entrant. The second entrant finds customers who already know what they are buying.

They pay for the risk of the market's existence itself. The most expensive question a startup answers is not 'can I build this?' but 'does anyone want to buy it?' Once they have answered, the answer becomes public and free.

Waiting, then, is not slowness. It is the purchase of an option: the right, without obligation, to enter once the uncertainty has been lifted. In finance, this option is very well understood. It is worth more the greater the uncertainty and the lower the cost of waiting. And waiting is cheap precisely when the company does not fear that a third party will seize the market while it observes.

Remember that last proposition. It contains, I believe, the whole problem.

Complementary assets, or why the inventor often loses

In 1986, David Teece published the article that, to my knowledge, best explains what we are describing. His question: why does the innovator so rarely capture the value of their innovation?

His answer rests on two variables.

The first is the appropriability regime: how protectable is the innovation, by patent, secrecy, technical complexity? A pharmaceutical active ingredient is highly appropriable. A merchant payment interface, a distribution model, a logistics organisation are almost not.

The second is the control of complementary assets: to turn an invention into a sold product, you need a sales force, a distribution network, after-sales service, a regulatory licence, a known brand, production capacity, relationships with key accounts. These assets are not the innovation. They are what makes the innovation worth anything.

Teece's theorem is blunt: when appropriability is low and complementary assets are held by an incumbent, it is not the innovator who captures the value. It is the asset holder. Not through predation, but through arithmetic — they own the bottleneck.

Let us apply this to Congo-Brazzaville, or any comparable economy.

Appropriability there is low: most local innovations concern services, interfaces, distribution models — objects that are hard to patent everywhere in the world. The country is a member of OAPI, the regional intellectual property organisation, and the OHADA legal framework exists and works; but a title has value only through the cost and duration of the litigation it enables, and a founder without cash flow does not engage a three-year proceeding against a better-resourced opponent. The law exists; the procedural power balance empties it of part of its effect.

Complementary assets, meanwhile, are extremely concentrated. In an economy where a few telecom operators, a few banks, a few distributors and a small number of groups hold the customer relationship, the physical network, the regulatory approval and the campaign capacity, a founder can practically bring nothing to market without passing through one of them.

Teece's conclusion then follows mechanically, without assuming the slightest bad intention: under these conditions, the value of the innovation goes to whoever holds the assets, regardless of who had the idea.

Why acquisition, in this context, is a bad calculation

It remains to understand why the executive does not prefer to buy. After all, acquiring is also buying time and a team.

Three reasons, all defensible in committee.

A startup's price is an option price, and the option has value only if another bidder exists. A startup valuation is almost never backed by its earnings. It is backed by what someone else would be willing to pay. In a market with a single credible buyer — a monopsony of acquisition — the reference value collapses, not because the company is worth less, but because the mechanism that revealed its value does not exist. The executive who is considering buying asks, quite legitimately: why pay today for what nobody else will contest tomorrow?

Acquisition brings costs that rebuilding avoids. Integrating a team, absorbing technical debt, inheriting poorly documented fiscal or social liabilities, fitting a culture of ten people into a company of two thousand. These costs are real and systematically underestimated, even in mature ecosystems. Faced with a technically reproducible product, a CTO will often say, without ulterior motive: "my teams can redo it in six months." They will frequently be right.

The incumbent company gains more by not moving fast. This is what Arrow called the replacement effect: the one who already dominates a market gains less from innovating than the entrant, because what innovation brings them partly cannibalises what they already earn. Conversely, Gilbert and Newbery showed that a threatened monopolist benefits from pre-empting — acquiring or locking up — precisely when the threat is credible. These two results do not contradict each other: they say that the incentive to act fast depends entirely on the credibility of the threat. Without a credible threat, waiting is free. With a credible threat, waiting becomes the main risk.

That is the whole problem of low-competition ecosystems: they do not produce a credible threat.

The same product, two arithmetics

Let us take an illustration — made-up figures, solely to make the structure visible.

A startup has built, in eighteen months, a product that cost the equivalent of €300,000 and that reaches €200,000 in annual revenue. A dominant player observes it.

Dense ecosystem. Three other groups are watching the same segment. A fund has already proposed a funding round that would let the startup triple its sales force in one year. If our executive waits twelve months, they will face a funded competitor, or pay a price multiplied by five, or watch a rival seize it and lock up distribution. Rebuilding would cost €200,000 and twelve months — but twelve months during which the target will have gained a head start they will have to buy back at a higher price. They buy now, at a price their CFO will find excessive and their board reasonable.

Thin ecosystem. No other credible buyer. No local fund capable of writing a cheque that changes the startup's scale. The founder cannot recruit twenty salespeople because they have no way to pay them before their sales come in. Waiting twelve months costs nothing: the target will be in the same place, with a better-validated market. Rebuilding will cost €200,000 to a company that already has the developers, the brand, the customers and the distribution network — so, in reality, a much lower marginal cost. The committee decides to wait, then rebuild. And it is right, from the perspective of its mandate.

It is the same product. It is the same reasoning. It produces opposite results.

The variable that tips the result is not virtue, nor culture, nor even market size. It is the existence of a competitor capable of making waiting costly.

What this implies

If this analysis holds, then one consequence follows, and it is counterintuitive for much of African innovation policy.

Training entrepreneurs will not be enough. Funding incubators will not be enough. Organising startup competitions will not be enough. As long as waiting remains free for incumbents, these programmes will produce better-prepared projects — that will encounter exactly the same arithmetic.

What must be changed is the price of waiting. And for that, we must understand what, elsewhere, has made it so high.

That is the subject of episode 3. I will examine why, in Silicon Valley, a company with a monopoly on a market prefers to pay a billion dollars for a team of thirteen people — and why this behaviour, often described as strategic generosity, is in reality an admission of fear.

Open question: in your industry, can you think of an example where a large company bought rather than rebuilt? And if so, do you know who the other potential buyer was?

Episode 3 — Why Silicon Valley buys: acquisition as admission of fear

In April 2012, a company that already dominated its market paid roughly a billion dollars for a thirteen-person, revenue-free photography application eighteen months old.

This fact is told in two ways.

The romantic version: Silicon Valley recognises talent and pays it generously. The cynical version: a giant eliminates a rival before it grows.

Both miss the point. This company employed thousands of engineers. It owned the audience, the infrastructure, the brand, the data, the ad platform. Technically, it could rebuild that application. It in fact did so several times afterwards, for other products, with mixed results.

So why pay?

Because in its ecosystem, waiting cost more than a billion dollars. The acquisition is not a gesture of generosity. It is the price of a risk one refuses to run. It is an admission.

This episode is about what makes waiting so costly in certain ecosystems — and about the fact that these mechanisms are constructed, therefore reproducible.

Mechanism 1 — Venture capital turns an idea into a credible threat

This is the central mechanism, and it is poorly understood. Venture capital is often presented as a funding window. That is not its main economic function.

Its main economic function is to make a small company dangerous before it is profitable.

A startup that raises the equivalent of ten million dollars can hire forty people, saturate a customer acquisition channel, sell at a loss for two years and take a market by speed. Overnight it becomes a competitor that can hurt. The executive opposite no longer reasons about what the startup is. They reason about what it can become in eighteen months if someone gives it the means.

It is this transformation that makes waiting costly. Without it, observation is comfortable: the target cannot accelerate, so it will still be there in a year, smaller than you.

Hence a reformulation that seems important to me, and that applies to all public innovation policy: venture capital does not only fund startups, it changes the behaviour of companies that are not funded. Its most powerful effect is a deterrent effect. An ecosystem without growth capital is not merely an ecosystem where startups lack money. It is an ecosystem where nobody fears them.

And you only buy what you fear.

Mechanism 2 — Network effects make falling behind irreversible

On an ordinary market, being second costs market share. On a network-effect market — messaging, marketplace, payments, social network, operating system — being second can cost the entire market.

Katz and Shapiro formalised this in the mid-1980s: when a product's value to each user grows with the number of users, the market tips. Past a threshold, the leader's lead becomes self-reinforcing, and the laggard cannot catch up with a better product — they would need to convince users to leave together, which no marketing campaign can do.

For the dominant player, this completely changes the nature of the wait. They are no longer playing for market share; they are playing for position. The calculation becomes asymmetric: a billion overpaid is a recoverable mistake, a market lost by tipping is not. Facing this asymmetry, a 300% acquisition premium remains a prudent decision.

This is worth noting for Africa: mobile money, last-mile logistics, marketplaces, digital identity, linguistic interoperability are all network-effect markets. The positions taken today will be very difficult to contest tomorrow. That is no reason to despair; it is a reason to consider that the timeline matters more than people say.

Mechanism 3 — Talent mobility makes it impossible to confiscate the secret

There is a California legal quirk that is rarely cited and explains a great deal: California does not enforce non-compete clauses between employer and employee. Ronald Gilson argued, in an article that became a classic, that this rule is one of the structural causes of Silicon Valley's advantage over Boston's Route 128, where Massachusetts law allowed them.

The consequences are considerable.

An engineer can leave their employer on Friday and found a competitor on Monday. Knowledge therefore circulates faster than organisations can contain it. A dominant player who decides to wait before rebuilding risks that three of its best engineers leave to build the product themselves, funded by a fund that understood before them.

In other words: in this system, you cannot confiscate an innovation, you can only buy the people who carry it. Hence the practice of acqui-hires, often mocked as an extravagance and in reality the expression of a power dynamic: talent is mobile, therefore it must be paid for.

Where careers are immobile, employers few, and an employee's departure to a competitor socially costly, this mechanism disappears. And with it, a good part of the founder's negotiating value.

South Korea: the country that disarmed its own giants

The obvious objection, when talking about Silicon Valley from Brazzaville, is one of scale. An internal market of 340 million solvent consumers produces mechanisms that a market of 6 million never will.

South Korea makes this objection insufficient, and it does so on the exact ground of this series.

Because Korea is not an open-ecosystem country. It is an economy where a few family conglomerates — electronics, automotive, chemicals, distribution, telecoms — hold the bulk of industrial value, distribution channels and customer relationships. That is to say: exactly the configuration described in episode 2, where waiting is free because no second buyer exists.

For a long time, the arithmetic there produced what it produces everywhere. An innovative supplier was not bought: it was absorbed as a subcontractor, or rebuilt in-house.

What changed amounts to one instrument. In the mid-2000s, the state endowed a vehicle that does not invest in companies, but in private venture capital funds. Public money enters as a minority subscriber alongside private investors, leaves investment decisions to managers, and withdraws as the market deepens.

The distinction seems technical. It is decisive. A public fund that invests directly funds companies and creates a queue. A fund of funds funds managers, and creates an industry. In the first case, the state becomes a window; in the second, it becomes the first client of a profession that did not exist.

The result affects the price of waiting. When several dozen independent funds have capital to deploy, a Korean startup ceases to have a single credible interlocutor. The conglomerate that was considering waiting must now factor in that a fund can finance the target until it becomes too expensive — or that a competitor buys it first.

And the second buyer, when it appeared, was not Korean. The largest acquisition of a Korean startup in the previous decade was made by a German group, for an amount far beyond what a domestic buyer would have offered. The credible threat was imported.

Estonia: when the alternative buyer was never local

Second case, more radical on scale: a country of just over a million inhabitants, with no internal market of any kind, no domestic global platform, and yet one of Europe's densest technology ecosystems relative to its population.

Estonia never had a world-scale national acquirer. It did not try to fabricate one. It built the opposite: the conditions for any foreign buyer to enter, value and exit without friction.

The country's first big success — a voice messaging service that went global — was acquired by a US player two years after takeoff, then resold to another. The next, a trade management software, was taken over by a US fund. None of these buyers was Estonian. None could have been.

What the country did, however, fits in a list more useful to Africa than any funding programme: a company legal form readable by a foreign investor, without offshore structuring; standardised fundraising documentation, therefore low in legal costs; fully digitalised company incorporation; an employee share participation scheme that actually works, making talent mobile; and frictionless convertibility, so that an investor knows how they will exit before entering.

The alternative buyer does not need to be local. They must be credible and reachable.

This shifts the problem to very concrete questions, far less glamorous than a national innovation plan. Does the legal structure of companies allow a foreign investor to enter and exit cleanly? Is capital repatriation practicable, and within what timeframe? Is the currency convertible? Is fundraising documentation standardised, or is each deal renegotiated from scratch with a lawyer?

These topics seem technical and boring. They determine the very existence of the second buyer.

China: imitation as a regime, but in an immense market

China is the honest counter-example, and I do not want to gloss over it.

For two decades, rapid imitation was a dominant and embraced strategy there. Many major Chinese platforms began by replicating Western models. If imitation mechanically destroyed innovation, China would not have become one of the two global technological poles.

What happened? Two things.

First, the copying happened in a market of over a billion consumers, where the imitator immediately found a scale that funded adaptation work, then original innovation. Copying was a starting point, not an endpoint, because the market's size made going further profitable.

Second — and this is the decisive point — imitation was followed by an internal competition of unusual ferocity. Ten players copied the same model and competed to exhaustion. Competitive discipline was restored not by law but by numbers. Two major groups then formed rival camps, investing massively in startups to prevent the other from doing so: another alternative buyer, this time produced by the rivalry of two blocs.

The Chinese lesson is therefore not 'imitation is harmless'. It is: imitation is tolerable where it is itself exposed to competition. An ecosystem where a single player can copy without being copied in return has nothing to do with an ecosystem where ten players copy each other.

Europe: the intermediate case, and what it reveals

Europe has excellent researchers, a solvent market, protective law, abundant seed capital. It produces, nevertheless, few global champions, and its best technology companies are frequently acquired by American buyers before reaching scale.

The commonly accepted diagnosis is not a deficit of ideas but a deficit of growth capital: seed money exists, late-stage money much less. Add market fragmentation — twenty-seven jurisdictions, as many tax regimes, about twenty business languages — and you get an ecosystem where scaling is structurally slower than elsewhere.

The European case therefore proves something important for our subject: talent is not enough, seed capital is not enough, legal protection is not enough. What is missing is the ability to convert an early lead into a dominant position fast enough for waiting to become dangerous for others.

If Europe, with its means, achieves this only imperfectly, one can measure the distance elsewhere. One can also see that the problem is not African.

Africa: the signals to read correctly

Three observations, without complacency.

First: when a credible foreign buyer appears, local valuations change immediately in nature. The acquisition of a Nigerian payments company by a US player in 2020, the entry of a global card operator into a Nigerian group's equity in 2019, did more for the credibility of exits in West Africa than ten years of speeches. These deals told founders that an exit existed, and investors that a price existed.

Second: Africa's most transformative innovations have often been carried by dominant players, not startups — Kenyan mobile money is the canonical example. This is a fact to face head-on rather than skirt around. It does not contradict my thesis: it narrows its scope. A dominant player is perfectly capable of innovating; it does so when incentivised, notably by competition or external risk-sharing. The question is never 'do large companies innovate?' but 'what pushes them to?'

Third: the presence of pan-African groups — telecoms, banks, distribution, logistics — present in fifteen or twenty countries creates the conditions for an acquisition market that did not exist fifteen years ago. Two competing operators in a single country are, literally, two possible buyers. It is thin. It is not nothing. It is exactly the material from which a price is built.

The one-line summary

In ecosystems that buy, it is not the morality that is better. It is that someone else was going to buy.

Everything else — venture capital, network effects, engineer mobility, foreign buyers, bloc rivalry — is just a way of producing that 'someone else'.

There remains a question I have set aside for two episodes, and which may be the most serious. We have talked about startups that exist. We have not talked about those that do not exist — and which, in my hypothesis, constitute the bulk of the loss. That is the subject of episode 4.

Open question: if you had to create a second credible buyer in your country, without waiting for an international fund to get interested, where would you start?

Episode 4 — Fear does not kill projects. It kills the very desire to start a business.

So far, I have reasoned from the company's perspective. Now we must look at the other side, and I want to do so without sentimentality, because that is where, I believe, the bulk of the economic cost lies — and it is never counted.

One sentence sums up what I want to establish: fear does not only kill projects, it kills the very desire to start a business. This is not a slogan. It is a proposition I will try to prove.

The fundamental asymmetry: two waits, one winner

Let us go back to episode 2. I showed there that waiting, for an incumbent, is the purchase of an option: the right to enter later, once the uncertainty has been lifted by someone else.

The founder has exactly the same option. They can choose not to launch, keep the idea, observe, tell themselves they will see later. And real options theory — Dixit and Pindyck gave the reference formulation — says something very clear: the more irreversible the investment and the more uncertain the environment, the more the option to wait is worth.

Except that the two waits are not of the same nature, and that is where everything is decided.

The incumbent's wait preserves the value of its option. Its assets — customers, distribution, brand, engineering capacity — do not depreciate because it observes. It waits from a position that remains good.

The founder's wait destroys the value of theirs. An idea is not a financial instrument, it is a perishable asset. The technological window closes, someone else gets there, the founder ages, takes out a mortgage, gets a stable job, takes on family responsibilities. Their option does not sleep: it melts.

The founder is therefore placed before a decision where waiting is costly and acting exposes them to dispossession. Both branches lose. In this configuration, renunciation is not cowardice: it is the solution to a badly posed problem.

Why this fear is not a cognitive bias

One might be tempted to file this under biases: loss aversion, overestimation of salient risks. Kahneman and Tversky showed that a loss weighs psychologically more than a gain of equal magnitude, and that striking events are overweighted.

I believe we must resist this reading, because it is too convenient. It places the fault in the founder's head, which dispenses us from examining their situation.

Yet, in an economy where seed capital comes from family, where there is no unemployment insurance for the self-employed, where insolvency proceedings do not truly free the debtor from their debts or their reputation, where entrepreneurial failure is a household event before it is a balance-sheet event — the founder's risk aversion is not a bias. It is a correct assessment.

The difference between one ecosystem and another is therefore not that people there are more courageous. It is that the cost of failure has been socialised. Where failure results in a line on a CV and a job six weeks later, taking a risk requires no particular virtue. Where it results in a ten-year family debt and a loss of status, prudence is intelligence.

This is a public policy point, not a psychology point: you increase risk-taking by lowering the cost of failure, not by organising seminars on boldness.

Reverse selection: who enters when entering is dangerous

Here is the most serious consequence, and the least often stated.

When the expected return from disclosure becomes low, the population of entrepreneurs does not merely shrink in number. It changes in composition.

Who keeps entering?

First, those who do not need to disclose: low-capital-intensity projects, self-financed, in small and uncontested markets. Excellent projects, often viable — but not the ones that produce technological breakthroughs.

Then those who are protected other ways: by a relationship, a position, a family, an access. Their advantage is not product quality; it is the assurance of not being dispossessed.

Finally, those who have not evaluated the risk.

What disappears from this population is precisely the profile every ecosystem seeks: the competent, clear-headed engineer, without special protection, capable of measuring what they risk — and who, having measured it, chooses salaried employment. Such an ecosystem does not select on talent. It selects on protection. And a selection mechanism that ignores talent produces exactly what we observe: many businesses, little innovation.

That is, in my view, one of the heaviest and most silent costs of the problem I am describing.

Secrecy as strategy, and the destruction of fertile ground

Second consequence, equally important.

A founder who fears copying adopts behaviours that are perfectly rational at their scale and collectively ruinous.

They do not talk about their project. They do not present it publicly. They refuse partnerships. They recruit only within their family circle. They publish nothing. They do not seek a mentor in their field, because the most competent mentor is also the most dangerous competitor.

Yet innovation is an agglomeration phenomenon. AnnaLee Saxenian documented this difference with remarkable precision by comparing two American regions of comparable technological level in the 1980s: Silicon Valley, open culture, mobile employees, information circulating in cafés and conferences; and Route 128 around Boston, vertically integrated companies, secretive, hierarchical. The first produced a cumulative dynamic, the second stagnated. It was not a difference of talent or funding. It was a difference of circulation.

Secrecy protects the project and sterilises the environment. It prevents encounters, cross-learning, the formation of specialised suppliers, the building of a shared vocabulary. You get a collection of isolated entrepreneurs, which is not the same thing as an ecosystem — and which will never produce one.

The loop that closes on investors

Third consequence, and the one that makes the problem so hard to undo.

An investor considering creating a fund in a country looks at deal flow: how many serious pitches would they see per year?

But the best projects stay hidden, precisely because their holders fear revealing themselves. The investor therefore sees only what agrees to be shown, which is an impoverished selection. They conclude, honestly, that there is nothing to justify an investment vehicle. They do not raise the fund.

And without a fund, founders remain without growth capital; without growth capital, they do not constitute a threat; without a threat, waiting remains free for incumbents; waiting remaining free, copying remains rational; copying remaining rational, fear remains justified; fear remaining justified, projects remain hidden.

The loop is complete, and every actor in it behaves impeccably. Nobody is wrong. Nobody cheats. The result is bad all the same.

Economists call this a low-activity equilibrium: a stable situation from which no actor has an interest in deviating alone, even though all would gain from exiting together. That is a capital point for the rest of this series, because it completely changes the nature of the diagnosis. A stagnant ecosystem is not necessarily 'behind' on a development path. It may be stuck on the wrong one of two possible equilibria. Falling behind is caught up with time. An equilibrium is left only with a coordinated impulse, strong enough to tip everyone's expectations at the same time.

This distinction governs everything I will propose in episodes 6 and 7.

What we see, what we do not see

Frédéric Bastiat wrote in 1850 a text whose title suffices to summarise the argument: That Which Is Seen, and That Which Is Not Seen. The bad economist, he said in substance, considers only the visible effect; the good one also considers the effects that do not happen.

Let us apply it.

What we see: a startup gets overtaken. There are names, a story, a felt injustice, possibly an article. One can debate it.

What we do not see: the projects that do not start. The software engineer who keeps their job. The pharmacist who does not automate the stock management they had envisioned. The veterinarian who abandons their idea for herd health monitoring. None of them goes bankrupt. None appears in a business register, an incubator report, a failure statistic. Their renunciation is perfectly invisible.

And that is where the mass of the cost lies. Not because all these projects would have succeeded — most would have failed, even in the world's best ecosystem, where the overwhelming majority of funded startups do not return their capital. But because even failures produce value for the rest of the economy, and projects that never launch produce none.

A startup that fails has trained engineers who will go work elsewhere. It has validated or invalidated a hypothesis for all those that follow. It has created a supplier, a usage, a vocabulary. It has pushed an incumbent to improve its offering. It has provided an investor with a data point they did not have.

A project that never launches does none of this. It costs nothing and produces nothing. It is the only type of failure from which one learns absolutely nothing.

Why the social cost exceeds the private cost

One final element completes the argument, and it is central in innovation economics.

Work on innovation returns — notably William Nordhaus's — converges on a finding: the innovator captures only a modest share of the total value created by their innovation. The rest goes to consumers, imitators, neighbouring sectors, subsequent innovations made possible.

This positive externality is the classic economic justification for public support for research and innovation: if the inventor captures only a share of the social gain, then the level of innovation chosen spontaneously by the market will be below the socially desirable level. Everywhere. Even in a perfect ecosystem.

Now add our mechanism. If, in addition to this universal under-capture, the inventor anticipates that a better-resourced player will take the rest, their expected share collapses further. Entry adjusts downward accordingly — and the gap between what society would gain and what the market produces spontaneously widens.

In other words: where appropriability is low, under-investment in innovation is not a marginal market failure. It is its normal result. Which means that waiting for 'the market to mature' is not a strategy. It is a description of immobility.

Where we stand

Four episodes, and a complete chain: fear is rational → because waiting is free for incumbents → because no second buyer exists → because there is no growth capital → because projects stay hidden → because fear is rational.

Before proposing anything, we must confront the most serious objection raised every time I present this reasoning, and which has the merit of being partly true: 'an idea is worth nothing, only execution counts'.

That is episode 5. I will give it its full force before saying precisely where it ceases to hold.

Open question: do you know someone who gave up on a project without ever launching it? And if you remember the reason they gave, was it really the true one?

Episode 5 — 'An idea is worth nothing, only execution counts': the most serious objection

Every time I present the reasoning of the four previous episodes, the same objection returns. It is stated with assurance, often by people who have built something, and it deserves better than a quick rebuttal.

'An idea is worth nothing. Only execution counts.'

I will start by giving it its full force. Then I will examine three other equally legitimate objections. And only then will I say precisely where this argument ceases to hold — for it does not collapse, it shifts.

The objection at its strongest

It rests on at least four solid facts.

Ideas are simultaneous. The history of science and technology is a history of multiple discoveries: calculus, the telephone, evolution, the integrated circuit — almost all major inventions emerged independently in several people, within months or years of each other. William Ogburn and Dorothy Thomas compiled an inventory as early as 1922. If an idea were rare, it would be precious. It is not: it is in the technological air of its time, available to anyone who looks in the right place.

The first entrant often loses. The dominant search engine was not the first. The dominant social network was not the first. The first touchscreen phone did not found the smartphone market. This is not anecdotal: in the majority of categories, the winner is a late entrant who executed better.

Experienced investors refuse confidentiality agreements. This is not arrogance: they see hundreds of pitches, often similar, and signing a non-disclosure with each would make their job legally impracticable. Their implicit position is coherent: what they buy is not the idea, it is the team and its ability to execute it.

Execution is intrinsically hard to copy. What gives a company its value — the quality of an organisation, the speed of iteration, the fine-grained knowledge of the customer, a thousand accumulated micro-decisions — belongs to what Michael Polanyi called tacit knowledge: a know-how that cannot be fully articulated, therefore not fully transferred. Copying a product is easy. Copying a team that learns fast is very difficult.

This objection is therefore right. I believe it is even mostly right. And yet it is incomplete.

Three other objections to take seriously

First objection: imitation is socially useful. Imitation is a diffusion mechanism, and diffusion is what turns an invention into growth. A patent is a temporary monopoly, therefore an accepted efficiency loss in exchange for an incentive. Serious economists — Boldrin and Levine among the most radical — argue that the net balance of intellectual monopolies is negative. Historically, today's rich countries massively imitated before protecting: nineteenth-century United States did not recognise foreign copyrights, post-war Japan and Korea built their industry on reverse engineering. Demanding very strong intellectual property protection in an economy that is not yet a technology producer can cost more than it brings. This objection is solid, and I will take it into account in episode 6.

Second objection: acquisitions are not always virtuous. I presented acquisition as the sign of a healthy market. That is one-sided. Two recent literatures show it. The one on killer acquisitions — Cunningham, Ederer and Ma documented it in pharmaceuticals — describes buyouts intended to bury a competing project rather than develop it. The one on the kill zone — Kamepalli, Rajan and Zingales — suggests that beyond a certain degree of concentration, the prospect of being acquired by the dominant player discourages entry and funding in the segments it occupies. In other words, an excess of acquisitions produces the same impoverishment as their absence. What I defend is therefore not 'more acquisitions', but something more precise, which I will formulate at the end of this text.

Third objection: the real problem is elsewhere. Electricity, logistics, education, market depth, macroeconomic stability, access to credit. A Congolese founder facing an intermittent power grid and a narrow internal market has more immediate difficulties than the structure of competition. That is true. I will simply answer that these constraints are real but they do not explain the variation: they do not say why, at equal constraints, one country sees technology companies emerge and its neighbour does not. The structure of incentives is one of the variables that explain this difference — not the only one, but one of the few on which one can act in two years rather than twenty.

Where the execution argument turns against itself

Let us now revisit the main objection and examine its logical structure. 'Only execution counts' is a statement that contains a hidden assumption, almost never made explicit.

It assumes that the ability to execute is distributed according to merit.

Yet executing is not a moral quality. It is a production function. It consumes specific inputs: capital, engineers, time, access to distribution, a customer relationship, a regulatory approval, the capacity to absorb losses during scale-up.

The question then becomes: how are these inputs allocated?

In a deep ecosystem, they are allocated by relatively contestable markets. An unknown founder can raise capital on the quality of their project. They can hire an engineer by paying them more, or by giving them equity. They can buy distribution. They can, in eighteen months, build an execution capacity they did not have. In this context, saying 'only execution counts' is a valid meritocratic statement, because the means of executing are accessible to those who can persuade.

In a thin ecosystem, these same inputs are not allocated by a market. They are held. Growth capital does not exist or goes through relationships. Experienced engineers are in the same three companies. Distribution belongs to a few operators. Approval depends on an administration whose response time is not the same for everyone. Here, execution capacity is not something you acquire: it is something you already have, or will not have.

And then the statement changes meaning without changing words. 'Only execution counts' ceases to be a statement about merit and becomes a tautology about position: the one who already held the means to win wins. It is no longer a rule of the game, it is a description of the starting ranking.

That is exactly where the argument turns. It is true where the factor market is deep. It becomes a circular argument where it is not. And the debate between those who invoke it and those who contest it is most often a dialogue between two people correctly describing two different ecosystems.

The threshold: when resource asymmetry becomes qualitative

We must be more precise still, because resource asymmetry exists everywhere. A startup is always smaller than the group opposite. It is not the asymmetry that causes problems, it is a threshold.

I would propose formulating it thus: asymmetry becomes disqualifying when it renders the founder's learning useless.

The objection's reasoning supposes that if the founder just executed better, the market would reward them. But in an ecosystem where the means of execution are held by incumbents and access depends on relationships the founder does not have, better execution does not change the outcome. The ceiling is not the founder's talent; it is the structure of who holds what.

The honest formulation of the objection would then be: 'only execution counts, provided you have access to the means of executing'. That is true everywhere. It just happens to be trivially true in deep ecosystems and devastatingly true in thin ones.

What to actually change

Four priorities, ranked by effect.

First: change the price of waiting. Make it costly for an incumbent to observe without acting. The instruments are known: encourage competing funds to enter the market, create public co-investment vehicles that reduce the startup's dependency on a single buyer, and build legal frameworks that allow foreign investors to enter and exit cleanly. The first effect is immediate: an incumbent who knows a fund can finance a startup to scale will acquire earlier, at a higher price, and the founder captures more of the value.

Second: change the cost of failure. Make it possible to fail without being destroyed. The instruments are known: modernise insolvency law so that failure does not carry a decade-long stigma, build unemployment insurance for the self-employed, and ensure that a failed founder can raise again. The first effect is behavioural: more people try, because trying no longer means risking everything.

Third: change the circulation of knowledge. Create the conditions for encounters, cross-learning, the formation of a shared vocabulary. The instruments are less glamorous than a national innovation plan but more effective: open-access industry data, public competitions with mandatory knowledge-sharing, and investment in the technical communities that already exist online but are underfunded. The first effect is informational: founders learn what others have tried, investors learn what works, and the ecosystem thickens.

Fourth: change the allocation of public procurement. Make it a tool for market creation, not just cost minimisation. The instruments are targeted: set-aside for local suppliers in technology contracts, pilot-project funding for startups that address public needs, and outcome-based contracts that allow unproven solutions to compete. The first effect is commercial: a startup with a public contract has a reference customer, a revenue stream, and a story that private investors can evaluate.

The answer is a price

The answer to all of this is not courage, not culture, not education. The answer is a price. The day waiting costs something to those who wait, ideas will come out of notebooks. Not because their authors will have become brave, but because they will have stopped being right to be afraid.

Open question, and last: among the four priorities described here, which one could your country launch before the end of the year — and what really prevents it, beyond the reasons usually given?