ARTICLE · Research

Customer needs research: what to check before you invest

Boris Kaptelov · 02.08.2026 · 7 min

Only one decision justifies all of this work: fund the launch or not. The cost of getting it wrong is the budget and six months of the team's time sunk into a product no one will hire for their job.

Needs research offers no guarantees. It screens out the hypotheses whose weakness is visible before the money is committed, and puts the rest into testable form. That is all it has to do, and it is enough to make it pay for itself.

Why "let the market decide" costs more

The market is the most honest test and the most expensive one. It delivers its verdict six months later, exactly once, with no explanation: they didn't buy. Then the guessing begins. Was it the price? Did they not see the point? The channel? Or is the job the product was built for already handled on the client's side by some other means — and handled reasonably well?

Each of these versions calls for its own rework, and all of them are expensive. The company picks one at random, spends another quarter, and gets another one-word answer.

Skimping on research does not reduce the cost of the mistake; it shifts the mistake to a stage where fixing it costs more. There is no universal multiplier here, and quoting one would be dishonest: the effect depends on the stage, the industry, and how the findings are used. The mechanism, though, is simple. Before launch you change a hypothesis. After launch you change production, contracts, trained people, and promises already made to the market.

Four questions the research must answer

Good needs research differs from bad not by the number of interviews or the thickness of the report. It differs in that it closes four questions, not one.

What job the customer is doing. Not which product they like, but what task they are solving and under what circumstances. The product is one way to get that job done, and usually not the only one. Until you understand the job, you cannot judge whether your solution is even relevant.

What they compare it against, including doing nothing. The real alternative rarely matches whoever you consider your competitor. More often it is a familiar workaround, someone else's product from an adjacent category, or the decision to put it off. The most underrated alternative to your product is buying nothing and carrying on as before.

What blocks the purchase and who decides. Price, risk, switching effort. And, separately, how many people take part in the decision. The dividing line here does not run between corporate and consumer markets: a private individual choosing housing, medical treatment, or education also faces multiple participants, high uncertainty, and a long approval process. Wherever the decision is not made by a single person, the enthusiastic user usually controls neither the budget nor the priority. More on this in our article on why the organization is not a respondent.

What the customer is willing to pay for. Not how much — for what. The difference is fundamental, and it deserves a section of its own.

Why you cannot ask about price directly

The direct question "how much would you pay for this" produces an answer that is elegant and almost useless. People name a figure based on their idea of the category, their budget, and the wish to sound reasonable in conversation. A real purchase works differently: it means parting with money, giving up a competing expense, accepting risk, and doing all of that at a specific moment. None of this is present in the conversation.

The gap between intention and action has been measured. Webb and Sheeran's meta-analysis of experimental studies showed that a medium-to-large shift in intentions is accompanied by only a small-to-medium change in behavior. Intentions move noticeably more easily than actions do.

For purchase intentions, the link to sales depends on what exactly you ask. The review by Morwitz, Steckel, and Gupta found a stronger link between intentions and purchases for existing products than for new ones, for short horizons than for long ones, and for a specific model than for the category as a whole. In other words, intention works worst precisely where you are most likely to ask about it: on a new product that does not yet exist.

What an interview genuinely delivers on price is the structure of value. What the customer pays for gladly, what they consider a mandatory part of the offer and refuse to pay for separately, where they already have a familiar reference point. That structure determines how to package the offer and what to build the price on. The number itself is tested later and by other means: reactions to different price levels, conversion, churn, behavior in the deal.

The scope of the research is not set by a quota

There is no single correct number of interviews, and no formula will produce one unless you are measuring frequency.

There is one working rule: interviews are counted not per study but per analytical group within which you intend to draw a conclusion. A narrow question in a homogeneous group closes with about ten. A comparison of three segments will not close even with thirty, because each segment ends up with ten.

Where the "twelve interviews" norm came from, what the empirical tests of saturation showed, and how to read all this in a vendor's estimate is a separate breakdown in our article on how to read a research quote.

How to ask

One distinction matters more than any technique: ask about an episode, not an opinion.

An opinion gets assembled right there in the conversation, out of whatever seems reasonable, and it will sound confident. An episode is recalled, and what is recalled can be checked against dates and documents. "Tell me how you last solved this" yields verifiable material. "Do you think you would need such a feature" yields nothing but politeness.

Seven questioning techniques, and the type of data each one produces, are covered separately. Only one consequence matters here: a founder in love with their solution imperceptibly stops probing the need and starts selling. The conversation, meanwhile, goes beautifully.

What this research proves and what it does not

We put this on record before the work begins, because it determines what decision the results can support.

Interviews show which jobs and barriers exist, how they are structured, what words the customer uses to describe them, and how one case differs from another. A recurring motif points to a stable pattern within the group studied.

It says nothing about market share. The phrase "six out of ten said this matters" describes those ten people. Prevalence is measured by a survey with a defensible sampling procedure, and a survey requires the categories to be defined in advance — which means the qualitative stage must already be complete.

Nor do interviews establish the size of an effect. A customer's explanation shows the version of the cause available to them, not the cause itself. That is a separate topic, and we take it up in the next article.

This sequence is cheaper than the reverse. A quantitative study will diligently count exactly what you put into it, so first you need to understand what to count.

What you get at the end

Not a report but a decision with its rationale: go or no-go, for which segment, around which job to build the offer, in which price zone, what to test next and by what method.

If you have no customer data at all, this is the first step, not the last. If the decision is reversible and cheap, it is smarter to test it in the field. Research pays off where the stake is large and rolling back is expensive.

On timing, the benchmark is this. One narrow question in one segment is a five-day diagnostic: we frame the decision, run about ten interviews, cross-check against your data, and deliver our position on Friday. Comparing segments, mapping the price contour, and handling a corporate purchase with several people involved in the decision is a two-month project. The difference is not in rigor but in the number of analytical groups within which a conclusion has to be reached.