Before week one: name the decision
The sprint starts by writing down the decision it exists to serve, in one sentence, along with what would have to be true to go each way.
If you cannot write the sentence, the research will be unfalsifiable and you will finish with a document instead of an answer.
Week one — recruit and listen
Eight to twelve conversations with people who match the profile the decision depends on. Not customers who are easy to reach; customers whose answer changes the outcome.
The questions are behavioural throughout: last time, what happened, what did it cost, show me. No demos, no roadmap discussion, no pitch.
Week two — analyse the language, not the summaries
Transcripts get coded for behaviour and for vocabulary separately. Behaviour tells you whether the problem exists; vocabulary tells you how to sell the answer to it.
This is also where contradictions surface — the places where what someone said in minute four is undermined by what they described in minute twenty. Those contradictions are the most valuable material in the sprint.
Week three — decide, in writing
The output is a recommendation: build, pivot, or kill, with the evidence for it and the evidence against it laid out separately so anyone can audit the reasoning.
A recommendation with no counter-evidence section is a sales document. Every honest decision has a case against it.
What gets cut
- —Surveys. Cheap to run, slow to interpret, weak at the sample sizes available in three weeks.
- —Competitive teardowns. Interesting, rarely decision-changing at this stage.
- —Persona documents. A decision needs one profile, not a cast.
- —Anything that produces a deck rather than an answer.
Why a kill is a good outcome
A kill decision in week three returns you the eleven months you were about to spend, with the customer relationships intact and the reasoning documented for the next bet.
The teams who find that hardest are the ones who framed the sprint as validation of a plan rather than a test of it. Which is why the decision sentence gets written first.