Engineering, production, governance, cost, communication · insight-1301
Write a time box and kill criteria into an exploratory modelling brief
- Evidence type
- practitioner heuristic
- Difficulty
- easy
- Roles
- Data scientist, Applied AI, ML engineer
- Industries
- cross-industry
Summary: situation, action, reason, boundary
Situation
An exploratory effort on call transcripts enters its fourth month with no go or no-go date (hypothetical).
Action
At kickoff write the time box, the baseline to beat, the minimum lift that would justify a build, the data and label assumptions that must hold, and the person who decides to stop. At the time box, review against only those criteria.
Reason
Stop rules fixed before results exist turn a vague hope into a testable bet and avoid deciding under sunk-cost pressure.
Boundary
Exploratory and proof-of-concept work with uncertain feasibility.
Symptom or risk
Sunk cost keeps it alive, new data sources and models keep being added, and stakeholders hear 'almost there' for months.
Exceptions
Research mandates that are explicitly open-ended; give them a review date instead.
Trade-offs
A short box can stop a project that needed one more month; allow one extension with a stated reason.
Pitfalls
Criteria written so loosely that any outcome counts as success.