What success criteria would you agree on before starting an analysis?
Success criteria are the definition of done for an analysis. Without them, work expands indefinitely, stakeholders expect different outputs, and nobody can say whether the analysis succeeded — which is why agreeing them takes minutes and saves days.
The question, written down — one sentence naming the metric, population, time window and baseline, precise enough that two analysts would interpret it identically.
The decision it informs — who will act on the result and what their options are, since this determines how much precision is genuinely required.
The deliverable and audience — a one-page summary, a refreshable dashboard or an ad-hoc figure, aimed at an executive or an operational team; these demand very different work.
The decision threshold — what magnitude of result would change the action, which separates a finding worth escalating from routine variance.
Deadline and what is out of scope — the date the answer stops being useful, plus an explicit list of questions this analysis will not answer, which is the main defence against scope creep.
Key terms: definition of done, scope, deliverable, decision threshold, out of scope, scope creep, stakeholder sign-off, acceptance criteria
Emphasise that success criteria are written and confirmed, not discussed verbally — a short restated brief is the artefact that prevents the "this is not what I asked for" conversation. The out-of-scope list is the point most candidates omit and the one experienced interviewers react to, because it shows you have been burned by scope creep.
Tie precision to the decision rather than treating maximum accuracy as always correct; saying that a directional answer in two days can beat a perfect one in two weeks demonstrates business judgement. Expect the follow-up "what if the stakeholder will not commit to criteria?" — propose a small time-boxed first pass and use its output to force the conversation.