Research Brief Interrogation

What questions should I ask before starting a research project?

Only the ones where two different answers would produce two different designs. Typically: what decision does this feed and when is it made, what happens if nothing is done, who else has to agree with the answer, what is explicitly out of scope, and what has already been tried. Three to six questions get answered properly. Fifteen get delegated and answered badly.

The brief says "run a survey" but does not say why. What do I do?

Strike the method from the brief and re-read what is left. If a question survives, the method was a preference and can be tested on its merits. If nothing survives, no question was ever written and the brief is a request for an artefact. Then ask what question that method would genuinely answer well, and whether it is the question the decision needs.

How do I tell if a client has already decided and just wants evidence?

Ask what they would do if the study found the opposite. If the answer names a changed action, the question is live. If the decision stands regardless, or the finding simply would not be published, the brief is a ratification exercise. Other signals: the expected finding appears in the brief, the timing sits after the decision meeting, and the language is "demonstrate" and "prove" rather than "understand" and "establish".

What is the difference between what a brief says and what it assumes?

A brief states facts you can check against a source, preferences about how the work should be done, and assumptions about the world presented as settled. Most of a brief is assumptions, and they hide in noun phrases: "why customers are abandoning the portal" assumes abandonment; "the barriers to adoption" assumes barriers. Tag each line, then ask what would change if the assumption were wrong.

Should I write down what is out of scope?

Yes, and clients read it as competence rather than retreat. Write scope in three parts: explicitly in, explicitly out, and the third category that causes most disputes, the things silently assumed to be in. Markets nobody mentioned excluding, the segment everyone pictures when they say "customers", the languages fieldwork will not run in.

How many clarifying questions is too many?

Any number where the important ones stop getting answered. Split them: blocking questions, which stop design starting, and non-blocking ones, which travel with a stated default. Send the blocking list alone, keep it short, and say what each one unlocks.

Can AI interrogate a brief for me?

It can do the mechanical parts well: the line-by-line register split, the assumption consequences, the stakeholder slots, the question list. What it must not do is infer a decision, a decision owner or a deadline that nobody stated, or resolve an ambiguity to make the brief read coherently. Ambiguity is the finding. Whether to challenge a client about a ratification brief is a professional judgement with a relationship attached.

What should I send back to the client after reading the brief?

A one-page restatement in your own words: what we understand you to be deciding, what we understand you to be asking, what is in and out, what we are assuming, and what we still need to know. Write it so they can disagree with it specifically. Disagreement now is worth more than agreement, because the alternative is discovering it at the debrief.