01 — Pivot Culture
Pivot culture often buys time, not learning.
Founders are taught to pivot as courage. Sometimes it is. Often it is a socially acceptable way to avoid a kill decision while sunk cost whispers. Years of effort and capital make the conversation gut-wrenching—which is exactly why kill criteria cannot live only in the retrospective.
Incubation environments amplify the tension: multiple bets, thin teams, pressure to show momentum. A pivot deck can read as strategy when it is still avoidance. The design job at validation stage is to make stop/pivot legible from evidence, not from narrative.
Kill criteria should be written in the same document as the wedge hypothesis. If you cannot name what would make you stop, you are not running validation—you are running morale theater.
02 — Recommend Against
Recommending against is a product deliverable.
Recommending against a direction is not failure of the design function. It is the product function when evidence says the idea should not receive the next tranche of build.
Pivots that preserve the same unexamined user, the same regulatory fantasy, or the same data model are often pivots in naming only. The deliverable is a stop/pivot memo backed by evidence: what we tried, what broke, what we will not fund next.
Incubation portfolios multiply this risk because parallel bets compete for narrative airtime. The discipline is per-bet falsification, not portfolio-average storytelling.
03 — Kill in Sprint
Evidence that should end a wedge belongs in the sprint.
Kill criteria are not a post-mortem template. They are inputs to the next cycle: what signal would we need to see to continue, and what signal already means stop. Without that, iteration becomes motion—tweaks that feel productive while the core bet stays unexamined.
Pair this with honest MVPs: behavior that can fail in public, not slides that cannot. The team that can kill quickly preserves trust with partners and with itself. The team that pivots forever trains everyone to distrust the roadmap.
On agentic civic work, stage gates only matter when falsifiers are named up front—wrong wedge, wrong channel, wrong regulatory path. Run the loop to break the thesis, not to decorate it.
The test: did this sprint produce a stop signal, a revised kill criterion, or another decoration slide? If only the last, you bought morale theater, not validation.
