FounderWiseDecisions, not feeds
← All articles FounderWise · Long-form

They are using it wrong, and that is the finding

A third of your accounts are doing something the product was not designed for. The instinct is to correct them. The information is in why nobody had to be told.

31 Oct 2026 13 min read By Joshua Pi’Rwot
Share X LinkedIn

Your reporting tool has a notes field. A third of your accounts are pasting entire client updates into it and exporting them, which is not what it is for and works badly.

Support has written a guide explaining the correct workflow. Several hundred people independently reached the same wrong answer, and the guide is an attempt to argue with a measurement.

Why these three models

The decision is what to do about customers using your product in a way you did not intend. The features that fire are an aggregate pattern that no one designed, a cause you cannot determine by reasoning about it, and a use that locates you somewhere other than where you positioned yourself.

Three lenses. Self-organising behaviour produces a complex answer about where the pattern came from. The black-box probe produces a random answer about how to learn from it without a theory. Position produces an equilibrium answer about what it tells you. The first two are diagnostic and the third is the one that changes the roadmap.

1. Nobody designed this, which is what makes it evidence

Start with what the pattern is. Individual users, each solving their own problem with their own local rules and almost none of them talking to each other, arrived at the same behaviour. That aggregate was not planned by anyone, including them.1

That property is what makes it worth more than any survey you could run. Nobody was prompted, nobody was selecting an option you offered, and nobody was being polite to an interviewer. A behaviour that emerges independently across many users is a strong statement about a need, precisely because no one designed the question.

Which reframes what the workaround is. It is not misuse. It is your customers doing product development on your behalf, at their own cost, and reporting the result in a form you can observe for free.

The instinct to correct it is worth examining because it is so consistent. It comes from the reasonable belief that you know what the product is for, and the belief is true and beside the point. What the product is for is a statement about your intentions. What it is being used for is a statement about the market, and only one of those is a fact about the world.

The first move is therefore to stop writing the guide and start counting. How many accounts, what proportion of usage, whether those accounts retain better or worse than the rest, and whether they arrived through a particular channel. That is an afternoon of queries and it converts an irritation into a measurement.

Then check that the convergence is actually independent, because the whole argument rests on it. A hundred accounts doing the same odd thing is strong evidence when they found it separately and weak evidence when one person taught them. They may share an industry association, a consultant, or a WhatsApp group where somebody posted the trick with a screenshot. The test is arrival timing read against the account list. Independent discovery appears steadily across cohorts and channels. A taught workaround appears in clusters: nothing for a year, then eleven accounts in a fortnight, all from one source. The first is a market signal. The second is one person’s opinion counted a hundred times, and building on it is building for that person.

Frequency does the other half of the separating. Of three hundred and forty accounts, if a hundred and twelve have touched the workaround and only nineteen use it more than once a week, you have a curiosity rather than a workflow, and the nineteen are the population to study. The accounts who tried it once tell you the idea occurred to them. The accounts who depend on it are the only ones who can tell you what the need is.

2. You cannot reason your way to why, so probe

Having counted, the temptation is to explain, and explanation is where this usually stalls. Everyone in the room has a theory, the theories are plausible, and none of them is checkable.

The alternative is to treat the users as a system you cannot see inside, and to learn by changing one thing and reading what comes back rather than by modelling the mechanism.2 Ship a small, cheap version of the thing they are improvising toward, to some of them, and watch. Do not announce it, do not survey, do not explain it. Change one input and read the output.

Do not ask them why they do it. Build the smallest version of what they are doing and see who moves.

The probe has to be designed so that it can come out either way, which is the part that gets skipped. If the only outcome you would accept is confirmation, you have built a demonstration rather than a test. Decide before you ship what result would tell you to stop, and write it down.

Choose the probe group carefully, because the obvious choice ruins the reading. Shipping the small version to your happiest and most engaged accounts guarantees a positive result, since those accounts adopt things partly because you built them. Include accounts you have never spoken to, and a few you would not mind losing. If the demand is real, strangers will find the built version without being told, and that is the outcome worth having.

There is also a probe available when you cannot ship anything, which is the usual position for a small team. Do it by hand. Have one person perform the improvised task for ten accounts for two weeks, manually, and charge for it if the workflow is worth charging for. That tests the demand with no engineering, and it produces something the shipped version cannot: you learn what the work actually involves, including the parts customers never mention because they stopped noticing them.

Interviews are still worth doing and they answer a different question. Asking a user why they do something produces a rationalisation constructed on the spot, which is unreliable about cause and quite reliable about what they were trying to achieve. Use interviews for the goal and probes for the demand.

There is a reason to be firm about that split rather than treating it as a nicety. People do not have reliable access to the causes of their own behaviour, so a why question returns a plausible story invented during the call, and the story is then repeated in your product meetings as a customer insight.4 The reason they give you is generated by the question, and it will outlive the evidence that contradicts it.

And keep the probe small. The failure mode here is a company that discovers an unintended use in March and has rebuilt around it by June, on a signal that was never tested against anything.

3. Where the misuse puts you

The third lens is the one that turns a product observation into a strategic one.

Think of the space of things a buyer might need, with your product occupying a position in it and competitors occupying others. An unintended use is a report that your actual position is not where you placed yourself: you are nearer to some neighbouring need than your positioning claims, and customers navigating that space found you anyway.3

That has consequences beyond the feature. It tells you which competitor you are actually being compared against, which is frequently not the one in your deck. It tells you what to put on the website, because the words those users would search for are not the words you use. And it tells you where the adjacent market is, which is the same question as what to build next, arrived at from evidence rather than from a workshop.

The strategic decision that follows is genuinely three-way and should be made explicitly rather than by drift. Serve it, by building the small version properly and repositioning toward it. Bound it, by supporting it well enough to retain those accounts while keeping the product’s centre where it is. Or refuse it, deliberately and with a stated reason, which is a legitimate answer when the use pulls toward a market you do not want.

Repositioning carries a cost that lands outside the product, usually on sales, and it is rarely counted. The pipeline in front of you was assembled under the old positioning. Those buyers arrived for the thing you said you were, and new words on the website will not convert them. Expect a quarter of weaker numbers while the top of the funnel refills with people who came for the new claim. Where that gap would break your cash position, bound the use now and reposition when there is room to absorb the pause.

Drifting is the fourth option and the most common. It looks like tolerating the behaviour, half-supporting it, and letting the roadmap be shaped by whoever complains, which produces a product that serves two positions weakly.

And plan for the improvising accounts rejecting what you build for them, because it happens often enough to design against. Their workaround gave them control: they could change the format, correct it before it went out, and do something you never anticipated. A tidy built version removes all of that in exchange for a convenience they did not ask for. Ship the new path alongside the improvised one and leave the old route open until the usage moves across on its own. Closing the workaround on the day the replacement launches is how a good finding becomes a support crisis.

What the three say together

  • Count before you correct: how many accounts, what share of usage, how they retain, where they came from.
  • Probe rather than explain. Ship the smallest version to some of them, decide the stopping result in advance.
  • Use interviews for what they were trying to achieve, not for why. The reason is constructed on the spot.
  • Choose explicitly: serve it, bound it, or refuse it with a stated reason. Drifting is not one of the three.

Where they disagree

The emergence argument and the position argument disagree about how much weight to give this.

Emergence says treat it as strong evidence, because independent convergence on the same workaround is exactly the signal that surveys and interviews cannot produce, and it arrived without you spending anything.

The position argument says that a use which pulls you toward a different location in the market is not automatically good news, that the accounts doing it may be a small and unrepresentative segment, and that following it can move you into a space where a much larger competitor already sits.

The reconciliation is that emergence establishes that the need is real and says nothing about whether it is yours to serve. Both questions have to be answered and they are answered by different evidence: the count and the probe tell you the need exists, and the competitive map tells you whether moving toward it puts you somewhere you can win. A company that answers only the first ends up in an adjacent market it entered by accident.

What none of them contain

None of the three prices the support cost of the current situation. Every unintended use generates tickets from people doing something difficult in a tool that does not help them, and that cost is running whether or not you decide anything.

None of them handles the misuse that is a compliance or safety problem. All of this treats unintended use as information, and some of it is exposure: data being stored somewhere it should not be, or a workflow that breaks a term you have agreed with someone else. That case is not a product decision and the counting exercise is the wrong first step.

None of them prices the option you close by serving the use. Building for the improvising accounts commits engineering and support to a second workflow, which narrows what you can do next in ways that are invisible at the time and binding two years later. The three-way choice is presented here as a product decision and it is also a decision about which futures remain available.

And one property the ensemble will not produce: the users who improvised are more attached to your product than the ones who did not. They invested effort in making it work, and effort creates attachment. If you serve the use properly they become advocates, and if you remove the workaround without a replacement they take it personally, which is a stronger reaction than a feature change normally earns.

The one action that survives the ignorance: before support publishes another guide, count the accounts doing it and check whether they retain better or worse than the rest. If they retain better, you are about to correct your most engaged customers, and the guide is the wrong artefact.

Who has to move

The founder, because support will document the intended workflow and engineering will treat the misuse as a defect, and both responses are locally sensible and jointly close the finding. The instinct is to reduce the confusion, which is the correct instinct applied to a symptom. The cheapest first test is the retention comparison, which takes one query and reframes the entire discussion when the improvising accounts turn out to be your best ones.

Sources and notes

  1. Stefan Thurner, Rudolf Hanel and Peter Klimek, Introduction to the Theory of Complex Systems, Oxford University Press. Agent-based treatments, in which a population of individuals following local rules produces an aggregate pattern that no participant designed, are developed in the chapters on complex adaptive systems. Used in section 1 for the claim that independent convergence on a behaviour is informative precisely because nobody coordinated it. Used qualitatively; no simulation is claimed.
  2. Stafford Beer, Cybernetics and Management, English Universities Press, 1959. The black box, and the argument that a system whose internals cannot be inspected is still controllable and learnable by manipulating inputs and observing outputs, run throughout the volume. Used in section 2 for the recommendation to probe rather than to model the mechanism, and for the requirement that a probe be designed so that it can come out either way.
  3. Jeffrey Carpenter and Andrea Robbett, Game Theory and Behavior, MIT Press. Spatial models of choice, in which options occupy positions in a space and buyers select by proximity to their own requirements, are treated alongside location and voting models. Used in section 3 as a framing for reading an unintended use as a report on your actual position. The application is analogical rather than measured; no position is being estimated numerically.
  4. Sanjit S. Dhami, The Foundations of Behavioral Economic Analysis, Oxford University Press. The unreliability of introspective reports about the causes of one’s own behaviour, alongside the general treatment of stated versus revealed preference, supports the distinction drawn in section 2 between what interviews are good for and what they are not. Direction only; no magnitude is claimed.

A note on a number this article does not give. There is no share of accounts at which an unintended use becomes worth building for. It depends on how those accounts retain, what they pay and whether the adjacent position is defensible. What transfers is the order: count, probe, map, then choose one of three options rather than drifting into the fourth.

Joshua Agonya Pi’Rwot, Founder.

Lock in your calls.

You’ve marked 0 of 5. Now choose how often you want the signals.

Step 1 · Pick your cadence

The DispatchWeekly · your Monday 5 callsFreealways

Step 2 · Where to send it

Personalize your BriefThe Brief

Tune every edition to the markets and industries you actually act on.

🔒 Unlock personalization — The Brief, $19.99/mo →
Free Dispatch forever · upgrade anytime · we never share your details.
Know where you stand?
The Dispatch tells you what changed. Knowing what to do about it is a different question, and it is the one FounderWise answers. Start with the free Traction Audit: 12 questions, about 3 minutes, scored out of 100.
Find out where you stand →

For teams, syndicates & programs

Recommended
Team
$15/seat · mo
Daily Brief for the whole team (min 3 seats).
  • Everyone on the same signal
  • Admin + shared watch-list
  • One invoice · ~25% off solo
Get Team →
Cohort License
$1,500/cohort
One cohort of up to 30 founders, for accelerators, funds & programs.
  • MVP Studio and GTM Studio for every founder
  • Your logo, your cohort
  • A funder report on what each founder did
Talk to us →
Pass the Dispatch on
Know a founder making these calls blind? Send them this week’s five — free, every Monday.

Decisions, not feeds. · Curated by Joshua Pi’Rwot · FounderWise · Free Audit · Store · parent of Business Growth Accelerator

Call committed. We’ll hold you to it.
Know where you stand →