research-question-framing

Extract the decision behind the topic by asking what the requester will do differently based on findings, then decompose the topic into question types (descriptive, comparative, causal, predictive, design) and pick one. Pin every noun — population, metric, baseline, timeframe — until two people would answer the same question.

Write a one-line brief with the pinned question, decision, timebox, and an explicit out-of-scope list where tangents get parked instead of adopted. Re-check the frame at the timebox midpoint and require re-signing for any legitimate re-framing.

A question is answerable when evidence exists or can be generated within the timebox, the question's weight fits the available time, and it is falsifiable — a 'no' answer must be genuinely acceptable to the requester. Questions failing these gates get renegotiated before work starts.

Do not use it to conduct the research itself — it only produces the question that methods like literature review, survey design, or experiment design then answer. It also requires access to the requester, since framing is a negotiation rather than solo wordsmithing.

Most failures are framing failures: no statable downstream decision, unpinned nouns like 'better' or 'at scale', untested load-bearing assumptions discovered in week three, and silent scope drift. A day-one assumptions register and signed brief prevent these.