Run an inception when the team can't state the problem the same way twice
When: Ask three people on the team what problem they are solving. You get three different answers, and nobody seems bothered by that.
The action: Run a short inception to agree the problem, the people it affects, and what the team will not do. Keep it to days.
Forces: An inception takes the team away from delivery, and teams under pressure resent that. Skipping it means the disagreement surfaces later, in rework. A long inception slides into big design up front.
Signals: Team members describe the problem in matching terms within a week. Fewer backlog items get rejected for being out of scope.
When not to: The work continues something the team already understands. An inception there is ceremony. Also skip it if the problem is genuinely unknowable without building something. Build a probe instead.
Licensed under CC BY-SA 4.0.