Ask the Right Person the Right Question

Hand the same questionnaire to a frontline analyst, a department head, and a board member and you have made a quiet error that shapes everything downstream. You have assumed all three can speak to the same things. They cannot, and pretending otherwise costs you twice.

The analyst gets asked about board level strategy they only see in fragments, so they guess, and you record the guess as data. The board member gets asked about the texture of a daily process they have not touched in years, so they answer from memory or impression, and you record that too. Everyone fills in the boxes, the response rate looks healthy, and you end up with a dataset that is partly real and partly people being polite about questions that were never theirs to answer.

A better interview starts from a more honest premise. Different people see different parts of the organization, and the job is to ask each person about the part they can actually see.

What someone can credibly speak to

Think about a single issue, say how forecasting works. A dozen different people touch it, and each one holds a different slice of the truth.

The analyst who builds the forecast knows where the numbers come from, which inputs arrive late, and which assumptions everyone quietly fudges. The manager who reviews it knows how much it gets reworked and how often it is wrong. The business partner who presents it knows whether anyone downstream trusts it. The executive who relies on it knows whether it helps them make decisions. Ask all four "is forecasting working," and you flatten four genuinely different vantage points into one mushy average that belongs to no one.

The richer move is to ask each of them about the part they live in. You get the mechanics from the people who run the mechanics, the credibility from the people who consume the output, and the strategic fit from the people who bet on it. Assembled together, those slices form a far more complete picture of forecasting than any single question could, precisely because nobody was asked to speak past what they know.

Adaptive, without losing comparability

There is an obvious objection. If everyone answers different questions, how do you compare anything across the organization at all. You would have hundreds of bespoke interviews and no way to add them up.

So the interview is not fully bespoke. It is built in two layers.

There is a shared core that everyone answers, the same handful of dimensions and a few standardized questions asked of every single respondent regardless of role. That core is what makes the population comparable. It is how you can say that one part of the organization rates something far lower than another, because everyone answered the same anchor questions on the same terms.

On top of that core sits the adaptive layer. Once the interview knows who it is talking to, it extends into the areas that person can credibly speak to and skips the ones they cannot. The analyst goes deep on process mechanics. The executive goes deep on strategic fit. Both answered the shared core, so both still sit on the same comparable spine, but each spent their remaining time on the questions where their answer carries real weight.

You keep the thing structured surveys are genuinely good at, comparison across everyone, and you add the thing they cannot do, depth that is matched to the person.

Adapting to the role, not leading the witness

Adapting the questions to a person is not the same as steering them toward an answer, and the difference matters enormously.

A leading interview decides what it wants to hear and nudges people toward it. "How frustrating is the new approval process" has already told you the answer it expects. That is how you manufacture a finding rather than discover one. A role-adaptive interview does something different and narrower. It changes which topics it raises based on what the person can speak to, then asks about those topics as neutrally as possible. It chooses the subject, never the verdict.

So the analyst and the executive might be taken into different territory, but within that territory the questions stay open and unloaded. The interview follows what each person actually says, probes when an answer is vague, and lets them describe their own experience in their own terms. Adapting the route is fair. Tilting the answer is not, and a diagnostic worth trusting holds that line hard.

The quiet benefit: it costs less to run

There is a practical reason this matters beyond accuracy, and it is the reason deep interviews at scale used to be impossible.

If you ask every person every question, you pay for a long interview multiplied by everyone on the roster, and most of that length is people laboring through questions they cannot really answer. Scope each interview to what the person can speak to and the conversations get shorter and sharper at the same time. You are not buying less insight. You are not spending on the dead weight of irrelevant questions. That is a large part of how a real conversational interview becomes affordable across hundreds or thousands of people rather than a privilege reserved for the dozen names a consultant has time to sit down with.

Better data and lower cost usually pull against each other. Here they happen to point the same way, because the same discipline that stops you collecting hollow answers also stops you paying to collect them.

The shape of a good interview

So the right interview is not one fixed list sent to everyone, and it is not a free for all with no common ground. It is a shared core that holds the whole population together, plus an adaptive extension that meets each person where their knowledge actually is, asked in language that never tips the scales.

Ask the right person the right question, in their own area, in neutral terms, and let the shared core do the comparing. What you get back is not a stack of answers about everything from people who knew a little about all of it. It is a layered picture, assembled from people who each spoke with authority about the part they live in.