A founder in your category gets exposed for inflating the numbers. The company dies. You had nothing to do with it. Then your own raise slows down, your data room grows longer, and a term you would have won last year is quietly off the table. Call it what it is: a transfer. The liar spent something that was not theirs to spend, and part of the bill arrived at your desk.
This is the sharpest corner of the idea behind everything we publish on capital. African venture does not really price risk. It prices the cost of checking. When one founder is caught lying, the market does not only punish that founder. It raises the price of checking everyone who resembles them, because it can no longer tell, cheaply, which of you is honest. The market cannot separate you from the liar without paying to look, so it prices you both at the blended average until someone pays to look.
The companion piece to this one stated the rule on the capital side: do not ask which capital is cheapest, ask who can check you cheapest, with the records you already hold. This is that same rule seen from the contagion side. A peer’s fraud raises your checking cost whether or not you ever met that peer. So the honest founder’s job is to make the checking cheap before the scandal breaks, not to protest innocence after it.
The evidence that this is real is now on the record. When the Ghanaian fintech Dash collapsed in 2023 after an internal audit found its founder had inflated user numbers, the reaction was not confined to Dash.2 Investors across the continent started talking, on the record, about tightening diligence on everyone.3 When the SEC charged Tingo’s founder with fabricating financials, alleging a subsidiary reported a bank balance of 461.7 million dollars against real accounts holding less than 50 dollars, the story became a headline about African startups, not about one man.1 Local investors said plainly that repeated reports of misconduct would worsen how outside capital sees the whole ecosystem.4, 9
Why these three lenses, and not one
One model would let me tell a single clean story about blame. Three models with different failure modes stop that. I am routing this to a set that spans three ways an outcome forms, because “one fraud re-prices the cohort” is really three separate machines running at once, and a lens tuned to one is blind to the other two.
The first is an equilibrium lens: statistical discrimination, the pricing of an individual off the estimated fraud rate of a group they visibly belong to. It explains why an honest founder is discounted at all. The second is a cycle-regime lens: SIR contagion, the epidemic shape of a reputational shock. It explains the timing, why the chill spreads fast, peaks, and then fades. The third is a complex lens: Granovetter thresholds. It explains why the diligence bar ratchets up and does not come back down, because each scandal tips more investors past their tolerance into a permanent verify-everything posture.
Two models I am deliberately not shipping as their own cards. Costly signaling belongs in the levers, because it is the founder’s response, not the market’s mechanism, and I fold it into the moves below. Behavioral pattern-matching, the investor who over-reads one vivid fraud, I fold into the threshold card and the blind spot, because it changes how fast the tipping happens, not the shape of it. Neither earns a lens of its own.
The framework: three machines behind one discount
1. The pricing lens: you are quoted the average of you and the liar
An investor cannot observe your honesty. They can observe your category, your stage, your geography, your founder profile. So they price you off the estimated base rate of fraud inside that observable group. This is statistical discrimination, and it is what a rational buyer does when quality is hidden and checking is expensive. When a fraud is exposed, the estimated base rate for your group jumps, and the price applied to every member of it drops in the same motion.
The cruelty is in the word “resembles.” Financial misconduct is not scattered at random. It clusters among firms of comparable size and among leaders of similar age, which is to say it clusters exactly along the lines a diligence process uses to sort you.5 So the founder who most looks like the exposed fraud, same sector, same stage, same story, pays the steepest unearned discount. A peer’s fraud does not make you riskier. It makes you unreadable, and unreadable gets the same discount as risky.
Assumes the investor prices you off the fraud rate of a group you visibly belong to, because your individual honesty is hidden and costly to verify.
Fits because misconduct clusters by size, sector and founder profile, the same axes diligence uses to classify you.
Breaks when you carry a cheap, hard signal that lets the investor price you as an individual instead of a member of the class.
Counteracts the instinct to argue you are honest. Words are the one thing the liar also had. Only a costly signal separates.
May reinforce unfair pattern-matching against whole categories, sometimes on surface traits that carry no real information.
2. The contagion lens: the chill spreads, peaks, then burns out
A reputational shock moves through a cohort the way an infection moves through a population. There is a susceptible group, the founders who look like the exposed company. There is an infectious event, the scandal and its coverage. And there is transmission: limited-partner letters, investor group chats, a wave of press, a diligence checklist copied from one firm to the next. The chill rises fast, reaches a peak, and then decays as attention moves on and as clean records accumulate around it.
Two things follow, and both are actionable. First, the shock has a clock. The tightening that followed the 2023 scandals was sharpest in the months right after, when funding to African startups had already fallen to roughly 2.3 billion dollars from over 4 billion the year before.4 Second, transmission to you is not fixed. How hard the contagion hits your raise depends on how connected your cohort is and how much you resemble the source. You cannot stop the epidemic. You can lower your own exposure to it, and you can choose when to stand in the road.
Assumes reputational damage transmits through a connected cohort with an epidemic shape: fast rise, peak, decay.
Fits because the 2023 African fraud cases produced a visible, dated chill that spread across the category and then eased.
Breaks when the cohort is loosely connected or the scandal is contained, so transmission never reaches you.
Counteracts the urge to raise into the peak. The same round is cheaper a quarter later, once the wave has passed.
May reinforce panic timing, where every founder waits at once and a queue forms behind the decay.
3. The threshold lens: the bar ratchets up and stays up
Each investor holds a private tolerance for soft-information diligence, a point past which they stop trusting the story and start demanding proof. A fraud nudges a batch of them over that line. Once enough flip, the market’s default diligence regime tips, and here is the part that stings: it does not tip back. Founders reported that diligence requests in 2023 looked nothing like 2021, heavier on financial controls, governance and founder behaviour.3, 4 That is a threshold system with hysteresis. The new baseline of proof becomes the permanent baseline, because no single investor wants to be the one who loosened up right before the next blow-up.
Why rational people over-correct is the behavioral piece I folded in here. One vivid fraud reads as more common than it is, so investors set the new bar higher than the true fraud rate would justify. That over-correction is a cost to you, and it is also a fact you cannot argue away. You meet the bar the market actually sets, not the one it should.
Assumes investors each have a diligence threshold, and a scandal pushes enough of them across it to tip the market’s default.
Fits because the post-2023 bar rose across the ecosystem and stayed risen, not just for the failed firms.
Breaks when the shock is too small to move the median investor, so the regime holds.
Counteracts the hope that things “go back to normal.” They rarely do. Build for the higher bar as the new floor.
May reinforce a proof arms race, where the cost of raising climbs for honest and dishonest founders alike.
GEER: the moves, cheapest and least binding first
Read together, the three lenses point one way. Talk will not move the discount, because talk is exactly what the liar also had. What moves it is a signal a liar would not pay for. Start with the moves that cost little and lock in nothing.
- Hand over a record the investor can pull without trusting you. A bank feed, a mobile-money till history, a payment-processor export. These are hard signals a stranger verifies directly, so they price you as an individual and lift you out of the group average.
- Attack the specific fraud pattern in your category. If the exposed fraud was inflated user counts, lead with transaction-level proof of real usage, not a headline metric. Meet the exact fear the scandal planted.
- Publish a dated record over time. A monthly note, sent in bad months too, builds a history that predates the scandal and that no founder can manufacture after the fact. Its value is that it was already there.
- Bring a counterparty who has already paid to check you. A named, reachable reference or a local co-investor absorbs a verification cost the distant investor cannot pay, and their willingness to vouch is itself a costly signal.
- Buy the audit last, and only by the numbers. An audit is the most expensive signal here, and it works because it is costly: education-style separation, where the honest founder can afford the signal and the liar’s cost is higher.8 Buy it when the discount it removes is larger than its fee, not before.
RADAR: what to line up before you open the round
Sequence by reversibility. Do the cheap dominant things now. Buy the tail insurance only if it pays for itself. Pre-commit the timing move to a trigger you name in advance, so you do not open the round in a panic at the worst moment.
- Do now (T+3 to T+14). Assemble the pack of records that verify each other, name one reachable reference, and secure one third-party confirmation of your headline number. Reversible, cheap, and dominant across every scenario, because it lifts you out of the group price regardless of what the market is doing.
- Hedge (by T+14). Get a quote for an audit or an independent metric verification. Hold it as tail insurance. Trigger the spend only when your own estimate of the cohort discount exceeds the fee.
- Defer and trigger (T+28 and beyond). Do not open into the peak of a category scandal. Pre-commit the open to an observable trigger: the news cycle has turned, a comparable clean company has raised, and your verification pack is complete. When all three fire, open. Until then, hold and keep building the record.
CHAIN: what usually happens to the founder next to the fraud
Match the comparison group by structure, not by surface. The right reference class is not “startups that raised after bad press.” It is honest founders who shared an observable profile with an exposed fraud, split into those who could hand over checkable proof and those who could only offer their word. The second group eats the full cohort discount. The first group is often waved through, because they gave the investor a way to price them alone.
The base rate here is unkind and worth stating flatly. Fraud is common enough that treating any category as clean is itself a bet: in normal times only about a third of corporate frauds are ever detected, so the rest sit unpriced in the background.6 And the market punishes the revealed ones hard, with reputational losses that dwarf the legal ones: for every dollar a firm falsely inflates, it loses that dollar plus about 2.71 more in lost reputation once the truth lands.7 Investors know both numbers in their bones, which is why one exposure makes them re-price a whole group rather than shrug.
Now adjust for your present state. The more you look like the exposed fraud, same sector, same stage, same claim, the heavier your discount, and the more a hard signal is worth. Then subtract the counterfactual. Some of the tighter terms you are seeing are the macro downturn, not the scandal, and you would have faced them anyway. Net that out and do not spend on removing a discount the fraud did not cause.
Matrix-break flag. If a shared verification rail becomes standard in your market, a common KYC utility, a verified-metrics feed investors trust by default, then the group signal stops being noisy. When every honest founder is cheaply checkable, statistical discrimination has nothing to bite on, and the cohort discount collapses. Watch for that rail. Building on it early is the cheapest exit from this whole problem.
What these three lenses cannot price
The ensemble assumes the fraud is real and exposed. It is blind to the false accusation, and to the founder who looks nothing like the liar yet still gets caught by lazy pattern-matching on a trait that carries no information. It cannot tell you the exact half-life of a given scandal, only that one exists. And it treats the investor as coldly rational, when some of the re-pricing is fear that will overshoot and misfire.
So here is the decision that survives the ignorance. You cannot time the next scandal and you cannot argue your way out of the one already here. You can control one thing: whether a stranger can check you without trusting you. Buy the signal a liar would not, before anyone asks, or you will be priced as though you already lied. This quarter, build one hard, checkable proof and hold your raise until the pack is complete. That is the move that pays whether the cohort is calm or on fire.
Sources and notes
- U.S. Securities and Exchange Commission, “SEC Charges Tingo Mobile Founder, Three Companies with Massive Fraud and Obtains Emergency Relief,” press release 2023-254, 18 December 2023. Verified: body states the FY2022 Form 10-K reported a cash balance of $461.7 million in Tingo Mobile’s Nigerian accounts while those accounts allegedly held a combined balance of less than $50, and that Mmobuosi fabricated financial statements. sec.gov
- BitKE, “Ghanaian Fintech, Dash, Shuts Down Operations After Raising Over $86 Million Following Reports of Inflated User Numbers,” October 2023. Verified: body states an internal audit found the CEO had inflated user numbers, the company had raised about $86 million including a round led by Insight Partners, and it shut down. bitcoinke.io
- TechCrunch, “In the wake of Dash’s closure due to fraud, 5 investors talk due diligence in Africa,” 24 October 2023. Verified: body frames Dash’s closure as due to fraud and gathers investors discussing how they are tightening due diligence across the market. techcrunch.com
- Techpoint Africa, “African investors say reports of founder misconduct could hurt future fundraising efforts,” 2023. Verified: body states only about $2.3 billion was raised by African startups in 2023 against over $4 billion in 2022, that misconduct reports could worsen outside perception, and that much of the continent’s capital comes from foreign VC firms. techpoint.africa
- Christopher A. Parsons, Johan Sulaeman, Sheridan Titman, “The Geography of Financial Misconduct,” NBER Working Paper 20347 (2014), later Journal of Finance 73(5). Verified: body states misconduct rises with the misconduct of neighbouring firms through peer effects, and that the effect is stronger among firms of comparable size and CEOs of similar age. nber.org (PDF)
- Alexander Dyck, Adair Morse, Luigi Zingales, “How Pervasive Is Corporate Fraud?” Verified: body states that in normal times only about a third of corporate frauds are caught, leaving roughly two-thirds undetected. haas.berkeley.edu (PDF)
- Jonathan M. Karpoff, D. Scott Lee, Gerald S. Martin, “The Cost to Firms of Cooking the Books,” Journal of Financial and Quantitative Analysis 43(3), 2008. Verified: for each dollar a firm inflates its value it loses that dollar plus about $2.71 in lost reputation, and reputational penalties far exceed legal ones. Primary PDF and a plain-language mirror. cambridge.org (PDF), washington.edu
- Michael Spence, “Job Market Signaling,” Quarterly Journal of Economics 87(3), 1973. Verified: body develops signaling costs and shows how a costly, observable signal separates types when productive capability is hidden at hiring. sfu.ca (PDF)
- Businessday NG, “Investor confidence takes hit as promising African startups fail.” Verified: body reports investor confidence taking a hit and stricter due diligence following a run of African startup failures. businessday.ng