You raised on the strength of a reputation you spent three years building. Eighteen months later the same investors are asking questions you thought you had answered permanently.
Nothing went wrong. Reputation is not a balance. It is a stock with an outflow. The outflow was running the whole time, and you stopped topping it up the day the round closed.
Why these three models
The decision is what cadence of proof to maintain when nothing is currently on fire. The features that fire are a counterparty forming beliefs over time, a quantity that accumulates and drains, and a disclosure choice with a verifiability condition attached.
Three lenses, three error structures. Reputation decay produces a cycle answer about why the effect expires. Stock and flow produces a complex answer about the rate and about what you can actually control. Unravelling produces an equilibrium answer about what refilling requires and when it is possible at all. The first two agree on the diagnosis and disagree usefully on the remedy: one says replenish uncertainty, the other says you cannot set a stock, only its flows.
1. Why the effect expires
The formal account of reputation is less flattering than the popular one, and more useful.
A reputation works because the other side assigns non-zero probability to your being a type that cannot behave badly. Not a type that chooses well: a type with no other option. That probability mass is what makes your promises informative, and it is the whole mechanism.
The consequence follows immediately. Every period they observe you, they update. Observation converges belief. Absent something that keeps the uncertainty alive, the probability attached to the committed type falls until reputation effects stop operating altogether, and the literature is explicit that the finding that reputations are temporary indicates a model of long-run reputation needs some mechanism by which uncertainty about types is continually replenished.1
Read that carefully, because it inverts the usual advice. Your reputation does not decay because you did something wrong. It decays because you did nothing, and because being observed behaving normally is exactly what resolves the uncertainty your reputation was made of.
Reputation decay, the expiry lens
- Assumes: reputation rests on the counterparty’s uncertainty about your type, and observation reduces uncertainty.
- Fits because: the relationship is ongoing and they have been watching for a while.
- Breaks when: your type is genuinely verifiable, in which case there is no uncertainty to replenish and this is a certification problem instead.
- Evidence: grade B plus. Formally established in the repeated-games literature; the rate of decay in any real relationship is not measurable.
- Counteracts: treating credibility as a permanent asset once earned.
- May reinforce: performative activity, since the model rewards being seen and does not distinguish signal from noise.
2. The leak, and what you can actually touch
The second lens converts the diagnosis into something you can operate.
Credibility behaves as a stock. It accumulates through inflows and drains through outflows, and it is conserved in the sense that it does not jump. A stock cannot be set directly. You can only move what fills it and what empties it, which is why “we need to rebuild trust with the board” is a sentiment rather than an action.
Name the flows and the plan writes itself. The inflow is verified evidence arriving at a counterparty: a delivered commitment they can check, a number they can confirm, a forecast that came true. The outflow is time passing without any of that, plus every event that raises their uncertainty without a corresponding update from you.
This is also the reasoning people are measurably worst at. Highly educated adults routinely fail to infer the behaviour of simple stock and flow systems, and the failure is not attributable to an inability to interpret graphs, to contextual knowledge, to motivation, or to cognitive capacity.2 So expect to underestimate this leak specifically, not as a personal failing but as the default.
The operational consequence is a cadence rather than a campaign. A quarterly update is an inflow of four units a year against an outflow that runs continuously. The monthly investor update is not politeness. It is the replenishment mechanism.
Stock and flow, the rate lens
- Assumes: credibility accumulates and drains through identifiable flows and cannot be set directly.
- Fits because: the problem presents as a level that fell without an incident.
- Breaks when: a single event resets the stock discontinuously, which fraud and litigation do. Then this is the wrong model entirely.
- Evidence: grade A. Structural, and the difficulty people have reasoning about it is well replicated.
- Counteracts: planning a credibility push instead of a credibility rate.
- May reinforce: mistaking volume of communication for inflow, when only verifiable content counts.
3. What actually refills it
The third lens is the constraint on the second, and it is the one that decides whether any of this is available to you.
Not all disclosure counts as inflow. Where information is verifiable, silence becomes informative: the strongest type discloses to separate from the next, which forces the next to disclose, and the chain runs down until even the weakest is disclosing. Where information is not verifiable, none of that happens, everyone pools, and your update is read as narrative rather than evidence.
The empirical version is unusually clean. When Los Angeles County required restaurant hygiene grade cards in windows, the cards caused inspection scores to increase, consumer demand to become sensitive to hygiene quality, and foodborne illness hospitalisations to fall.3 A verifiable disclosure changed what firms did, not merely what customers believed.
There is a corollary worth stating because it is counterintuitive. The chain only runs where the receiver can check cheaply. In a market with no verification infrastructure, disclosing more does not separate you from a weaker operator, because they can disclose the same things and nobody can tell the difference. That is not a reason to stay quiet. It is a reason to understand that the binding constraint is the verification layer rather than your willingness to be open, and those have completely different remedies.
So the test for anything you are about to send is simple and unforgiving. Can they check it without asking you? A delivered milestone with a customer reference is inflow. A revenue figure your accountant will confirm is inflow. A paragraph about momentum is not, and it consumes the attention that inflow would have occupied.
Verifiable disclosure, the inflow-quality lens
- Assumes: disclosure separates types only when the receiver can verify it cheaply.
- Fits because: you are choosing what to put in an update, not whether to send one.
- Breaks when: verification is expensive, in which case types pool and the good operator is priced as the average.
- Evidence: grade A. Robust theory with a clean field demonstration.
- Counteracts: the belief that more communication is more credibility.
- May reinforce: over-disclosing checkable trivia because it is easy to verify.
The levers, cheapest first
- Set a cadence, not a campaign. Monthly beats quarterly not because monthly is virtuous but because the outflow is continuous and four inflows a year will not match it.
- Audit your last update for checkable content. Mark each claim as verifiable without contacting you, or not. If most are not, the update was activity rather than inflow.
- Report the miss before they find it. A missed commitment disclosed by you is inflow. The same miss discovered by them is outflow, and the difference is entirely in who spoke first.
- Attach one third-party confirmation per cycle. A customer who will take a call, an accountant’s figure, a signed order. One is enough to change the category of the whole document.
- Keep a small amount of uncertainty alive deliberately. A commitment that is genuinely uncertain and then met is worth more than a certainty restated, because it is the only kind of evidence that moves a belief.
What to do before the next update
Do now, sized at one hour, effect visible next cycle. Take your most recent investor or board update and mark every claim as checkable or not. Reversible, free, and dominant across every scenario about why the questions came back.
Hedge, where the premium is the whole loss, live before the next month closes. Add one third-party-confirmable item to the next update. If your credibility was never the issue you have spent an afternoon collecting a reference, and that is the entire downside.
Defer and trigger, size fixed now. Do not rebuild your reporting stack this quarter. Pre-commit the trigger instead: the first time a counterparty asks a question you believe you already answered, the cadence moves up one step and stays there. Decide now what one step means, because a cadence chosen in response to a difficult conversation is chosen under pressure.
Watch the arrivals as well as the dates. The audit lands this week. The cadence change is felt after two or three cycles, because a stock responds to a changed flow with a delay, and the temptation to abandon the new cadence will peak exactly before it starts working.
What usually happens next
Run the break test first. Has a rule changed, has an actor entered or left, has a measurement become a target? A new partner joining the investor’s side resets the uncertainty to its starting value for that person, and your accumulated stock does not transfer to them. That is a genuine structural break and it is the most common one here.
If nothing broke, the shape is a slow decline with no incident. That trajectory is the signature of an outflow with no matching inflow, and it is why people reach for an explanation involving something that must have gone wrong. Usually nothing did.
There is a second regularity. The cadence tends to collapse in exactly the months when the news is worst, which is when the inflow is most valuable and hardest to produce. That is not weakness of character. A month with bad numbers offers fewer checkable positives, so the update takes longer to write and delivers less, and the rational short-run move is to skip it. The stock does not care about the reason.
Subtract the counterfactual before crediting a recovery. Credibility that returned in the same quarter you shipped a large customer probably tracked the customer, not the communication. The test is whether the questions stop returning in a quarter when nothing large ships.
What this ensemble cannot see
All three lenses treat credibility as one quantity held by one counterparty. It is neither.
It is held separately by every person on the other side, each with their own observation history, and their stocks do not move together. A board that appears to have lost confidence is often one member who did, and the aggregate behaviour looks like a single declining level when it is a distribution changing shape. Nothing here tells you which, and the remedies differ completely.
There is also a limit on the central model. Reputation decay is formally established and its rate is unmeasurable in any real relationship. This framework can tell you the direction and that it is continuous. Anyone offering you a decay rate is selling something.
And one property none of these models contains: raising your disclosure cadence changes what your counterparty expects as a baseline. The new rate becomes the reference point, and returning to the old one later will read as a deterioration even if it is where you started.
The one action that survives the ignorance: before the next update goes out, name the single person on the other side whose confidence you would least like to lose, and write one checkable item aimed specifically at what they last queried. If you cannot name that person, that is the finding, and it matters more than the cadence.
Who has to move
This only changes anything if whoever writes the update reads it, and in most companies that is the founder writing at the end of a hard month, when the checkable items are the ones hardest to face. The cheapest first test is the one-hour audit of the last update. If most claims turn out to be unverifiable, you have learned that your communication has been activity rather than inflow, and you learned it before the next round rather than during it.
Sources and notes
- George J. Mailath and Larry Samuelson, Repeated Games and Reputations: Long-Run Relationships, Oxford University Press, 2006. The adverse-selection approach to reputation and the role of commitment types is section 15.1 and 15.2. Temporary reputations, including the result that the posterior attached to the commitment type eventually falls below the level at which reputation effects operate, and the authors’ own conclusion that a model of long-run reputations should incorporate a mechanism by which uncertainty about types is continually replenished, are section 15.5. Reputation with replacements as one such mechanism is chapter 18.
- Matthew A. Cronin, Cleotilde Gonzalez and John D. Sterman, Why don’t well-educated adults understand accumulation? A challenge to researchers, educators, and citizens, Organizational Behavior and Human Decision Processes 108(1), 2009, pages 116 to 130. Author copy: https://www.mit.edu/~jsterman/CroninGonzalezSterman061210.pdf. The abstract states that highly educated people are often unable to infer the behaviour of simple stock-flow systems, and that persistent poor performance is not attributable to an inability to interpret graphs, contextual knowledge, motivation, or cognitive capacity.
- Ginger Zhe Jin and Phillip Leslie, The Effect of Information on Product Quality: Evidence from Restaurant Hygiene Grade Cards, Quarterly Journal of Economics 118(2), 2003. Open-access copy: https://drum.lib.umd.edu/bitstreams/1fce72fc-166b-49ba-8167-fc69cd0b13a6/download. The abstract states that the grade cards cause restaurant health inspection scores to increase, consumer demand to become sensitive to changes in restaurants’ hygiene quality, and the number of foodborne illness hospitalisations to decrease.
A note on the uncomfortable implication. If reputation rests on the counterparty’s uncertainty about your type, then perfectly predictable good behaviour eventually stops carrying information, which is a strange thing for a framework to recommend against. The resolution is not to manufacture doubt. It is to keep making commitments that were genuinely uncertain when you made them, which is the only way a met commitment stays informative.
Joshua Agonya Pi’Rwot, Founder.