They ran six weeks of diligence on you: customer calls, cohort data, a background check, two reference calls you did not arrange. You looked at their portfolio page and spoke to a founder they suggested.
The asymmetry feels normal because it is normal. You are about to give someone information rights and a board seat for a decade on less checking than you would do on a supplier.
Why these three models
The decision is which investor to take money from when you have a choice, and what to find out before you do. The features that fire are a judgement made by category because individual assessment is expensive, a set of references whose value depends entirely on how they were collected, and a reputation that describes a period rather than an institution.
Three lenses. Statistical discrimination produces an equilibrium answer about what you are actually doing when you rate a fund by its name. Aggregation produces a random answer about how to make founder references informative. Reputation decay produces a cycle answer about why the fund’s reputation is evidence about a past you are not buying. Together they turn a vague instinct that you should check into a specific set of calls.
1. You are pricing them by their category
Start with what you are currently doing, because it is the same thing they do to you and you have strong opinions about it when you are on the receiving end.
Evaluating an individual investor properly is expensive: it needs several conversations with people who have no reason to talk to you. So you substitute the average of an observable group. A recognised fund name, a partner who has been on a panel, a portfolio containing a company you admire. That is statistical discrimination, and it is a reasonable response to a screening cost rather than laziness.1
It is also exactly what you complain about when a fund prices you off your category rather than your company, and it has the same failure mode. The average of a group tells you very little about a specific member, and the variance inside a fund is often larger than the variance between funds. The partner matters more than the firm. The partner’s behaviour in a difficult month matters more than their behaviour in a good one, and the brand tells you nothing about either.
The move that breaks this is the same one you would want a fund to make about you: gather individual evidence that the category cannot supply. Which means the specific question is not is this a good fund. It is what does this partner do when a portfolio company misses, and that is answerable.
There is a version of this that fails even when you do it well. In much of East Africa the person you are diligencing is not the person who decides. A fund with a Nairobi or Kampala office often has a local partner who sources and champions, and an investment committee in another country that votes. You can collect four excellent accounts of that partner’s conduct and still be wrong about the institution, because the behaviour you were told about belongs to someone who cannot approve your next round alone. Ask how many of this partner’s recommendations the committee rejected in the last year, and what happened to those companies afterwards. A partner who has never had a deal turned down is either unusually good or not the one bringing deals.
If you cannot run the full exercise, run the cheapest fragment of it. Ask the partner to name the three companies they most recently led an investment into, then find out for yourself what happened to each. That is one question in a meeting and an hour of your own work, and it is harder to dress than a portfolio page, which shows the survivors.
2. How to make references actually mean something
The second lens is about method, and the method most founders use guarantees an uninformative answer.
Aggregated judgements beat individual ones under specific conditions: the judgements have to be roughly independent, their errors have to point in different directions, and there need to be enough of them.2 Almost every reference process founders run fails all three.
A reference the fund selects is not independent, because the selection is the whole point. Two founders from the same portfolio who have spoken to each other are one observation. And a call in which you ask whether they would recommend the fund produces a socially obligatory answer from someone who still needs that fund’s support.
The repairs are mechanical. Choose the founders yourself, from the portfolio page, including at least one company that is no longer operating, because the informative conversation is with someone who has nothing left to protect. Call them separately and do not tell them who else you are speaking to. Ask about events rather than opinions: what happened when they missed a quarter, how long the partner took to answer an urgent message, what they did in the down round, whether they ever went around the founder to a customer or an employee.
Ask what happened, not what they think, because a story has details and an opinion is a courtesy.
The method has its own failure mode, and it is the mirror of the one it fixes. Weight the sample entirely toward companies that died and you collect grievance, which is as predictable as courtesy and no more informative. A founder whose company failed has a story in which somebody is responsible, and the fund is a candidate. Include one, not three. The comparison that carries information is between a founder who had a hard year and stayed and a founder who had a hard year and did not, because the difference between those two accounts is where the partner’s conduct actually shows.
Some founders will not talk to you at all. A scheduled call with a written agenda raises the cost of candour in a small ecosystem, so lower it: a voice note asking one specific question gets answered when a calendar invitation does not. If the fund objects to you contacting founders it did not nominate, do not argue the principle. Note it, and treat it as you would a customer reference a supplier refused to let you check.
Three or four independent conversations conducted this way produce something usable. One suggested call produces nothing, and its function in the process is to let both sides feel that diligence occurred.
3. Their reputation describes a period you are not buying
The third lens is the one that catches the most experienced founders, because it operates on exactly the funds with the strongest names.
A reputation is not a permanent record. It is a flow, replenished by recent conduct and decaying without it, and under imperfect observation it is temporary rather than durable.3 A fund’s reputation reflects the partners who built it, the period they built it in, and the market conditions of that period.
Several things break the link between that reputation and what you will receive. The partner who created it has moved to a different fund or stopped leading deals. The fund is deploying a much larger vehicle, which changes what outcome it needs from you. It is late in a fund’s life, which changes its patience. Or the market that made its behaviour generous has turned, and behaviour that was never tested under pressure is about to be.
Which makes the horizon of your questions the important variable. Ask about the last two years specifically. Ask which fund number you are in and where in its life. Ask how many companies this partner currently sits on the board of, because attention is finite and a partner with fourteen boards is not the partner in the story you were told.
Fund life is the variable founders skip most often, and it has a mechanical effect on how you will be treated. A ten-year vehicle in its second year can wait, follow on, and let a thesis develop. The same vehicle in its seventh year is looking at a distribution schedule, and the patience it once had has become expensive. Ask the fund number, the vintage year, and whether the next fund has closed. A partner raising a new vehicle needs marks and exits from your cohort, and that need arrives as pressure to raise at a higher price before the numbers support one.
Then ask what happens to the board seat if the partner leaves, and get the answer into the documents rather than into the meeting. The reputation you diligenced attaches to a person. The seat attaches to the fund. When those two separate you inherit a director you never met, chosen by people you never spoke to, carrying none of the history that made the relationship work. A right to be consulted on the replacement costs nothing to ask for during the term-sheet period and cannot be obtained afterwards.
Guard against the opposite error while you are at it. One vivid story from one founder, good or bad, will feel more decisive than three measured accounts, and it should not.4 A single catastrophic experience may describe a genuine pattern or a bad match between two specific people, and the only way to tell is whether it recurs in conversations that were not connected to each other. One dramatic reference is a story, and three quiet consistent ones are a finding.
None of those questions are hostile and all of them are answerable in one meeting. A partner who is irritated by them has told you something useful in the cheapest possible way.
What the three say together
- Diligence the partner, not the firm. The variance inside a fund is usually larger than the variance between funds.
- Pick the references yourself, including one company that failed, and call them separately without naming the others.
- Ask what happened, not what they think. Events have details; opinions are courtesies.
- Bound the questions to the last two years, and ask which fund number and how many boards.
Where they disagree
The aggregation argument and the decay argument disagree about who to call.
Aggregation says get more independent observations, which means talking to as many portfolio founders as you can reach, including from older vintages, because sample size is what makes the exercise work at all. Decay says most of those observations describe a fund that no longer exists in the relevant sense, so a large sample drawn across ten years is a confident answer about the wrong institution.
The resolution is to weight by recency rather than to choose between them. Prioritise founders who have dealt with this specific partner in the last two years, and treat older conversations as evidence about the firm’s stated values rather than about the behaviour you will get. Where those two disagree, and they sometimes do, the recent behaviour is what you are buying and the older reputation is what you will be told about.
What none of them contain
None of the three prices the cost of asking. In a small ecosystem, calling four founders about a fund is visible, and a partner who hears about it may read it as ordinary professionalism or as a lack of trust depending on who they are. That risk is real and it is also a test: a fund that penalises careful diligence has answered your question.
None of them handles the case where you have no choice. Much of this assumes competing offers. Where one term sheet exists, the exercise is still worth running, but its purpose changes from selection to preparation: you are learning what to expect and what to get in writing rather than deciding whether to sign.
None of them handles the investor who is excellent and wrong for you. A partner can be honest, responsive and genuinely helpful, and still push a company toward an outcome that suits their fund’s shape rather than yours. Every check in this article measures conduct, and conduct is not the same as alignment. The alignment question is answered by fund size and vintage rather than by references, and it deserves its own conversation.
And one property the ensemble will not produce: you are also being chosen. An investor who has seen a great deal more of these relationships than you have is forming a view of how you will behave under pressure, and the way you run this process is part of what they observe. Doing it carefully and courteously is not only diligence, it is the first evidence they have about how you handle a difficult conversation.
The one action that survives the ignorance: pick three portfolio founders yourself, at least one from a company that no longer exists, and call them separately this week. Ask what happened when things went badly. Thirty minutes each, and it is the only part of this process nobody has arranged for you.
Who has to move
The founder, during the term-sheet period rather than after signing, because that is the only window in which the questions are cheap to ask and the answers can still change what you do. The instinct is to avoid anything that might slow a live offer, which is exactly the pressure that makes the asymmetry persist. The cheapest first test is one call to a founder whose company failed, chosen by you rather than suggested. That single conversation usually contains more information than the rest of the process combined, and nobody has an interest in arranging it for you.
Sources and notes
- Jeffrey Carpenter and Andrea Robbett, Game Theory and Behavior, MIT Press. Statistical discrimination, in which a decision-maker substitutes the average of an observable group when individual assessment is costly, is developed alongside signalling and screening. Used in section 1. The symmetry noted there, that founders do to investors exactly what they object to investors doing to them, follows from the mechanism being about screening cost rather than about preference.
- Jeffrey Carpenter and Andrea Robbett, Game Theory and Behavior, MIT Press. The conditions under which aggregated judgements outperform individual ones, and the failure of aggregation when the individual estimates are correlated rather than independent, are treated alongside information cascades and social learning. Used in section 2, including the specific consequence that a reference list supplied by the subject of the reference violates the independence condition by construction.
- Martin W. Cripps, George J. Mailath and Larry Samuelson, Imperfect Monitoring and Impermanent Reputations, Econometrica 72(2), 2004, pages 407 to 432, developed at book length in George J. Mailath and Larry Samuelson, Repeated Games and Reputations: Long-Run Relationships, Oxford University Press, 2006. Under imperfect monitoring a reputation is temporary rather than permanent and must be replenished by recent conduct. Used in section 3 for the claim that a fund’s reputation is evidence about a period and a set of people rather than about an institution.
- Sanjit S. Dhami, The Foundations of Behavioral Economic Analysis, Oxford University Press. The tendency to draw confident conclusions from a small number of vivid cases, and to prefer a compelling individual story to a frequency, supports the recommendation in section 2 to ask about events rather than impressions and to collect several independent accounts. Direction only; no magnitude claimed.
A note on a number this article does not give. There is no number of reference calls that constitutes sufficient diligence. Three independent and recent conversations beat ten arranged ones, and the quality is entirely in who chose them and what was asked. The method transfers, the count does not.
Joshua Agonya Pi’Rwot, Founder.