Research / 22 July 2026 / 6 min

Past behavior questions: three rules of an honest interview

Why "what did you do last time" works better than "would you buy": three interview rules from the CustDev Agent method.

Short answer: Three rules of an honest interview from the CustDev Agent method: one question at a time, relying on the facts of past behavior instead of the hypothetical "would you buy", and recording direct quotes with an assessment of the strength of evidence. Past behavior is more predictable than future promises.

Three points of the method

The passport of the CustDev Agent fixes three methodological principles: "One question at a time", "Facts of past behavior", "Quotes and the strength of evidence". This is not decoration of the card — the whole coding and synthesis rest on these three points. Let us take each one apart.

Rule 1: one question at a time

A double question ("How do you choose a service and what irritates you about it?") gives one answer to one of the halves — usually the one that is more convenient for the person to answer. The rule "one question at a time" protects the purity of the data: every fragment of the interview can later be referenced as the answer to a specific question. During the coding, such fragments are broken down into observations, situations, jobs and obstacles without guessing what the person was answering.

Rule 2: facts of past behavior

The question "Would you buy a solution that…?" measures the politeness of the person, not the market. The question "How did you solve this task last time: what did you do, how much did you spend, how did it end?" measures behavior. Past behavior has already happened — it can be checked with details: dates, amounts, names of tools. The hypothetical future cannot be checked with details.

The practical consequence: if the respondent answers about the future, bring them back to the last real case. If there was no real case — that is a valuable observation in itself: the problem may not be worth solving.

Rule 3: quotes and the strength of evidence

Every conclusion in the report rests on a direct quote, separated from the researcher's interpretation. But quotes are not equal to each other: "I changed the tool three times in a year" is stronger than "well probably, it would be convenient for me". The assessment of the strength of evidence is explicitly built into the artifact — the report with quotes, contradictions and hypotheses.

How these rules work together

One question at a time gives clean fragments; questions about past behavior give fragments that can be checked with details; quotes with the strength assessment give the ground for every conclusion. Together they produce an opportunity map where the hypotheses are separated from the facts — and the team understands which decisions can be made right now and which require the next series of interviews. More about the structure of the final report — in the article "Opportunity map".

Read also: Five interviews - data or illusion? The limits of the sample in CustDev · Opportunity map: how interviews become verifiable hypotheses

Questions and answers

Why must we not ask "would you buy?"?

Hypothetical questions measure politeness, not behavior: people readily promise to buy what they never buy. Past behavior is checked with details, the future is not.

What to do if the person answers with generalizations?

Bring them back to the concrete last case: when it was, what exactly they did, how it ended. A generalization without a case is a signal of low strength of evidence.

Who assesses the strength of the evidence?

The agent separates the quotes and the interpretations and shows the strength of the ground of every conclusion; the decision whether the evidence is sufficient for a step is made by the researcher and the team.

Why one question at a time?

This way the answer stays concrete and verifiable: a chain of questions blurs the facts.

What are the facts of past behavior?

The real actions of the person in the past instead of their forecasts about future intentions.

Why are intentions a weak signal?

Intentions often diverge from actions; past behavior is observable and verifiable.

How does the agent distinguish a quote from an interpretation?

Direct quotes are separated out; interpretations are marked and not mixed with the facts.

What does the strength of evidence show?

How much the conclusion rests on direct quotes, single signals or assumptions.

Can template questions be used?

They can, as a base, but the wording is adapted to the segment and the goal of the research.

Who benefits from this approach?

Researchers and product teams that need verifiable hypotheses instead of opinions.

First step

Choose a ready-made solution or describe your process.

Pick one of the twelve agents and subagents, or fill in the questionnaire for custom development.

Search agentseffect.com

Quick links