Founder-Market Fit Is Evidence a Stranger Can Check

Founder-market fit is a claim about your judgment, and a claim about judgment becomes evidence the moment a stranger can check it without you in the room. Three things make it checkable: dated calls that could have been wrong, access other people cannot buy, and contact with the market recent enough to still be true.

Every founder asserts insight. The assertion is free to make, so it carries almost no information, and the reader on the other side of the table discounts it to the pool average.

What moves you off that average is the format your insight arrives in. This piece builds that format, and gives the investor the questions that test it.

Why these three models

This runs the Wire Model: score the features of the decision, route to a small ensemble, then force the ensemble to produce dated actions. The scores that mattered:

• Verifiable versus unverifiable disclosure (0.9). Part of what you say can be confirmed by a third party today. The rest cannot. The two halves behave differently in the listener’s head.
• Small-sample inference (0.85). You are asking someone to infer a repeatable ability from a handful of observations, most of them unresolved.
• Knowledge decay (0.6). Market knowledge is a stock built by contact. Stop the contact and the stock falls.
• Cognitive distortion (0.7). Hindsight tidies a messy career into a clean line of sight. Both sides of the table do it.
• Regime-break risk (0.5). A fluent origin account now costs nothing to generate, which changes what the telling proves.

That routes to three models: the luck-skill continuum (the sample), verifiable disclosure and unraveling (the format), and Markov knowledge-stock decay (the clock). They span three outcome types, random, equilibrium and cycle-regime, so their errors point in different directions.

The behavioral layer is folded into the first card and the blind spot rather than shipped as a fourth. Hindsight bias changes how much weight you put on the same evidence and produces no lever of its own, so a separate card would have cost a hundred words and returned nothing. The governance member stays: the luck-skill model is it.

The framework: three tests a claim has to pass

1. The sample: one right call is one observation

Start with what the reader is actually weighing. In a survey of 885 institutional venture capitalists at 681 firms, the management team was mentioned as an important factor by 95% of firms and as the most important factor by 47%. Business-related factors, the model, the product, the market and the industry, were ranked most important by only 37% combined.¹

Now the uncomfortable part. A randomized field experiment sent nearly 17,000 emails to 4,500 active early-stage investors through AngelList, varying which facts each investor saw. Investors responded strongly to information about the founding team and did not respond to information about traction or about existing lead investors. The effect was driven by the most experienced and successful investors, measured by prior deal count, a success metric and network centrality. The least experienced investors responded to every category.²

Read those two together. The input carrying the most weight is the input with the least verification attached, and the sharpest readers are the ones leaning on it hardest. Your founder-market fit claim is doing more work than any number in your deck.

So how much does past evidence actually move the odds? Across several thousand venture-backed companies, entrepreneurs who had previously taken a company public had a 30% chance of succeeding in the next venture. First-time founders had 18%. Founders who had previously failed had 20%.³

Twelve points. That is what a completed, public track record buys. Skill here is real and modest, which sets the honest ceiling on any evidence sheet.

There is a fast test for whether skill exists in an activity: ask whether you can lose on purpose.⁴ Could you deliberately choose the wrong collection day for an agent float top-up, the wrong price break for a wholesaler, the wrong week to push credit at a distributor. If you can name the wrong answers, you hold real skill, and skill that can be named can be scored.

One correct call is one observation. A track record is a set of them, written down before the answer arrived.

The scoring method already exists. In a two-year geopolitical forecasting tournament, 85 questions resolved in year one and 114 in year two, open an average of 102 days. A 45-minute probability training module, team collaboration, and moving top performers into elite teams all improved calibration and resolution.⁵ The mechanism was mundane: questions with resolution dates, scored after the fact.

Founders can run the same machinery on a market. “Agent float runs dry around the 28th in this corridor” is a claim with a date and a check. “I understand this market deeply” cannot be scored.

Luck-skill continuum, the sample lens

Assumes: outcomes mix skill and luck in a stable ratio, and skill shows up only across repeated observations.

Fits because: small-sample inference scored 0.85, and prior success moves success odds by about twelve points.

Breaks when: the market is genuinely new, so there is no reference class and no honest base rate to regress toward.

Counteracts: reading one correct prediction as proof of a repeatable edge.

May reinforce: paralysis, and dismissing real early insight because the sample is small by construction.

2. The format: what a stranger can check, and what your silence says

The equilibrium result is clean. When a party’s claims are verifiable and the listener applies sophisticated skepticism, withholding becomes informative in itself. The listener assumes the worst about anything left out, so the informed party discloses, and disclosure unravels toward the full picture.⁶

Real listeners fall short of that, and the gap is measured. In a laboratory disclosure game with 324 subjects, roughly 95% of senders holding the best draws disclosed them. The majority holding a middling draw withheld it. Withholding actually maximized the sender’s expected return, because receivers overestimated what was being hidden. Receivers were insufficiently skeptical about undisclosed information, and repeated feedback narrowed the gap.⁷

Two actions follow, one for each side.

If you are the founder: the investor who has read four hundred decks is the receiver with feedback, and the discount they apply to your silence is larger than you think. State your weakest number yourself, in the first paragraph, with the reason. The disclosure costs you a sentence. Being caught costs you the meeting.

If you are the investor: your reading of an omission is almost certainly too generous. Write down what a deck does not contain before you write down what it does, and ask for the missing row by name.

Now the format itself. A founder-market fit claim becomes evidence when it points at an object with a date, a name and a third party attached. Working examples, all of which a stranger can confirm in under ten minutes:

• A WhatsApp order thread with customer names, dates and quantities, running back further than your company does.
• Ninety days of mobile-money settlement pulled from the operator rather than typed into your own spreadsheet.
• An agent roster with agent IDs and last-active dates.
• A signed LPO from a named distributor, plus their payment terms letter.
• A regulator’s correspondence, sandbox letter or licence file number.
• Two customers who have agreed in writing to take a verification call.

Each of these carries the same underlying property: it costs a founder with real market contact almost nothing to produce, and costs a founder without it either a great deal or an outright lie.

Verifiable disclosure and unraveling, the format lens

Assumes: claims split cleanly into checkable and uncheckable, and the listener is at least somewhat skeptical about what is missing.

Fits because: verifiable versus unverifiable disclosure scored 0.9, the highest feature in the set.

Breaks when: the reader has no feedback loop, which is the measured case. Then withholding pays and disclosure is unrewarded.

Counteracts: the habit of building a narrative around the numbers you like.

May reinforce: document theatre, where the founder optimizes the artifact set and stops talking to customers.

3. The clock: market knowledge depreciates

Treat market knowledge as a stock, accumulated by contact and drained by absence. That structure has been measured elsewhere. In commercial aircraft production, the estimated monthly depreciation parameter on a firm’s stock of production experience was 0.96, which implies that 61% of the experience held at the start of a year survives to the end of it.⁸

Match that on structure rather than surface: a stock built through repeated contact, depreciating through turnover and idleness, in a setting where nobody notices the loss until output falls.

The inheritance evidence points the same way. In the US laser industry from 1961 through 1994, 69 of the 465 entrants whose backgrounds could be traced were spinoffs from incumbent firms, and those spinoffs survived considerably longer than other startups, comparable to the longest-lived diversifying entrants.⁹ Knowledge carried out of an incumbent is a real asset, and it was fresh on the day it left.

Markets rewrite themselves faster than founders update their claims. Kenya capped commercial lending rates in September 2016. Banks rationed micro, small and medium enterprises out of the credit market, and the Central Bank of Kenya estimated that rationing lowered 2017 growth by 0.4 percentage points, with the MSME contribution to real GDP growth down 1.42 points.¹⁰ The cap was repealed in 2019. A founder whose lending-market knowledge was earned in 2015 held a stale asset by 2017 and a differently stale one by 2020.

Your founder-market fit carries a date, and the date is the last time the market surprised you.

One question tests it. What did you learn this month that changed a plan. A founder in live contact answers in about four seconds. A founder trading on a five-year-old career reaches for biography.

Markov knowledge-stock decay, the clock lens

Assumes: knowledge is a stock with a roughly stable depreciation rate, replenished by contact at a rate you control.

Fits because: knowledge decay scored 0.6, and the depreciation of experience stocks is measured, not assumed.

Breaks when: the market is structurally frozen. In a market whose rules have not moved in fifteen years, old knowledge stays accurate.

Counteracts: treating years of experience as a permanent asset on your balance sheet.

May reinforce: churn, where the founder mistakes constant meetings for learning and logs contact rather than surprise.

GEER: the levers, cheapest first

Four channels carry the exposure: verifiability (can a stranger confirm it), sample (how many dated calls exist), access (what you can obtain that others cannot), recency (contact rate this month). Pull the cheap, reversible levers first.

1. Rewrite each fit sentence as an object. Every claim gets a name, a date and an artifact, or it gets deleted. Hits verifiability. Costs an hour.
2. Open a call log. Three dated, falsifiable claims about your market, each with a resolution date inside 90 days. Hits sample. Costs twenty minutes.
3. Disclose your weakest number first. Hits verifiability, and it moves you out of the pool the experienced reader is discounting. Costs one sentence.
4. Document one call you got wrong and the decision it changed. Hits sample, and it is the single hardest artifact to fabricate.
5. Produce one access artifact. Operator data pull, distributor terms letter, agent roster, regulator file number. Hits access. Costs days.
6. Set a contact quota. Eight customer or channel conversations a month, logged with date and name. Hits recency. Costs a standing calendar block.

No-lever flag: if no claim you hold produces a checkable artifact and nothing you can reach is unavailable to a stranger, then the gap is in the fit itself and formatting will not close it. Write that down and re-plan the quarter around getting into the market.

RADAR: the portfolio, dated

DO NOW, by T+3 days. Reversible and dominant across every scenario.

1. Build the one-page evidence sheet: three claims, three artifacts, three names a stranger can call.
2. Log three dated calls with resolution dates inside 90 days. Send them to one person outside the company so they are timestamped elsewhere.
3. Put your weakest number at the top of the sheet with the reason it is weak.
4. Book eight customer or channel conversations for the next 30 days.

HEDGE, by T+14. Cheap insurance against the claim never being believed.

1. Obtain one artifact you can get only because of who you are and where you sit.
2. Get two customers to agree, in writing, to take a verification call.
3. Write the wrong-call note: what you believed, what happened, what you changed.

DEFER AND TRIGGER. Irreversible, so wait, and pre-commit the trigger now.

1. Defer: repositioning the company around a new market thesis, hiring a market-facing executive to own the insight, spending your strongest reference on a fund.
2. Trigger to spend: two of your three logged calls resolve correct by T+90 and at least one investor asks for a named artifact. Then put the thesis in the deck and spend the reference that week.
3. Counter-trigger: by T+28, no investor conversation has produced a question about how you know what you know. The claim is not landing as a claim. Rewrite it as a number with a date and a name attached.

If you are the one reading. DO NOW: ask every founder for one call they got wrong and what it changed, and score the answer for specificity. HEDGE: split each deck’s fit claims into checkable and uncheckable, log the ratio, and ask for one missing row by name. DEFER: changing your screen until you have twenty deals scored both ways and can compare the checkable-ratio against outcomes at 24 months.

CHAIN: what usually happens next

Match the reference class on structure. Forecasting tournaments, verifiable disclosure games and venture performance persistence all run the same machinery: a claim about future states, made by an interested party, scored later against what happened.^{5, 7, 3}

The base rate is the discipline. A complete, public track record moves success odds from 18% to 30%.³ A well-built evidence sheet should move an investor’s prior by something in that neighborhood, not to certainty. Price it that way when you plan your raise.

Present-state modifier: the most experienced investors react almost exclusively to team information.² That raises the payoff to a checkable team claim, and it raises the payoff to faking one by exactly the same amount.

Second-order consequence: as founders adopt the evidence format, the format becomes the new pool average, and the separating question moves from “do you have artifacts” to “who else confirms them”. Third-order: investors build verification into the process itself, and unverifiable fit claims stop being neutral and start being scored as negatives.

Subtract the counterfactual before over-crediting the founder. Some of what reads as founder-market fit is market timing. The laser spinoffs inherited an industry that was growing under them.⁹

Matrix-break flag. A fluent origin account now costs close to nothing to produce, which drains the information content of the telling. The same technology makes verification cheap: an operator data pull, a records lookup, a call to a named customer. The equilibrium moves toward verifiable disclosure faster than it otherwise would, and it moves in one direction only. Build for the verified side.

What this ensemble cannot see

Four things, and they are large.

Tacit knowledge with no artifact. Nine years trading in a market with no CRM, no receipts and no data trail is real expertise that this framework prices at zero. That is a failure of the framework, not of the founder.

Whether the market survives. Every model here assumes the thing you know still exists to be known. A licensing regime, a currency control or a platform policy can void the whole stock in a quarter.

Selection toward the legible. The evidence format rewards founders in markets that keep records, which in practice means formal, digitised, urban markets. It systematically underweights the operators with the deepest knowledge of informal trade. Investors reading only the sheet will keep funding the legible over the correct.

Whether your reader has feedback. The measured result is that receivers under-punish silence.⁷ An investor on their fifth deal reads the same evidence sheet completely differently from one on their five hundredth, and you cannot see which one is across the table.

So build the sheet anyway, because it is the only part you control. One page by T+14: three claims, three artifacts, three callable names, one wrong call, and the weak number stated first. If you cannot fill a single row today, make one dated, falsifiable call about your market in public this week and resolve it inside 90 days. That row is your first piece of evidence.

Sources and notes

1. Gompers, P., Gornall, W., Kaplan, S. N., and Strebulaev, I. A. “How Do Venture Capitalists Make Decisions?” Journal of Financial Economics 135(1), 2020, 169-190. Survey of 885 venture capitalists at 681 firms. Management team mentioned as important by 95% of firms and most important by 47%; business model 83%, product 74%, market 68%, industry 31%, with business-related factors most important for 37%. NBER working paper 22587, full text.
2. Bernstein, S., Korteweg, A., and Laws, K. “Attracting Early-Stage Investors: Evidence from a Randomized Field Experiment.” Journal of Finance 72(2), 2017, 509-538. Nearly 17,000 emails to 4,500 active early-stage investors on AngelList. Investors respond to founding-team information and not to traction or existing lead investors; the result is driven by the most experienced and successful investors, while the least experienced respond to all categories. Working paper full text. Publication record.
3. Gompers, P., Kovner, A., Lerner, J., and Scharfstein, D. “Performance Persistence in Entrepreneurship.” Journal of Financial Economics 96(1), 2010, 18-32. Working paper version: “Skill vs. Luck in Entrepreneurship and Venture Capital: Evidence from Serial Entrepreneurs,” NBER Working Paper 12592, 2006, which states the 30%, 18% and 20% success probabilities directly. Full text.
4. Mauboussin, M. J. “Untangling Skill and Luck: How to Think About Outcomes, Past, Present, and Future.” Legg Mason Capital Management, 15 July 2010. The lose-on-purpose test appears on page 2. Full text.
5. Mellers, B., Ungar, L., Baron, J., Ramos, J., Gurcay, B., Fincher, K., Scott, S. E., Moore, D., Atanasov, P., Swift, S. A., Murray, T., Stone, E., and Tetlock, P. E. “Psychological Strategies for Winning a Geopolitical Forecasting Tournament.” Psychological Science 25(5), 2014, 1106-1115. Eighty-five questions resolved in year one and 114 in year two, open an average of 102 days; probability training ran about 45 minutes. Author copy, full text.
6. Milgrom, P., and Roberts, J. “Relying on the Information of Interested Parties.” RAND Journal of Economics 17(1), 1986, 18-32. The unraveling and sophisticated-skepticism results. Author copy, full text.
7. Jin, G. Z., Luca, M., and Martin, D. “Is No News (Perceived As) Bad News? An Experimental Investigation of Information Disclosure.” American Economic Journal: Microeconomics 13(2), 2021, 141-173. Main sessions cover 324 subjects; roughly 95% of senders holding a draw of 4 or 5 disclose it, a majority holding a draw of 2 withhold it, and withholding raises expected returns because receivers are insufficiently skeptical about undisclosed information. Publisher bot-blocks direct access, so the identical working-paper text is linked. Harvard Business School Working Paper 15-078, full text.
8. Benkard, C. L. “Learning and Forgetting: The Dynamics of Aircraft Production.” American Economic Review 90(4), 2000, 1034-1054. The estimated monthly depreciation parameter of 0.96 implies that 61% of the firm’s stock of experience at the start of a year survives to the end of it. The paper builds on the organizational-forgetting findings of Argote, Beckman and Epple (1990). NBER Working Paper 7127, full text.
9. Klepper, S. “Spinoff Entry in High-tech Industries: Motives and Consequences.” Working paper, 2006, reporting the laser-industry results of Klepper, S., and Sleeper, S., “Entry by Spinoffs,” Management Science 51(8), 2005, 1291-1306. Sixty-nine of 465 traceable laser entrants from 1961 through 1994 were spinoffs, and they survived much longer than other startups. Full text. Journal record.
10. Central Bank of Kenya, “The Impact of Interest Rate Capping on the Kenyan Economy,” March 2018. Rationing of micro, small and medium enterprises out of the credit market is estimated to have lowered 2017 growth by 0.4 percentage points, with the MSME contribution to real GDP growth down 1.42 percentage points. The cap was introduced by section 33B of the Banking Act in September 2016 and repealed in 2019. Full text.