Individual hypotheses — specific, testable ways to attack a bottleneck
A play is a single hypothesis: a specific way to attack a bottleneck. Plays sit at the canopy of the tree because that is where the most experimentation, variation and learning happens.
Many plays may be tried; few will graduate. That is the point.
Each Play has a clear structure: an Intervention (what we will do), an Expected Outcome (the falsifiable prediction), Test Boundaries (scope, cohort, duration), and an Abort Condition (when to stop immediately). Plays run in parallel within a Playlist, and their results are the raw evidence that feeds the Portfolio Decision Forum.
The discipline at this level is designing Plays that produce clear signal quickly — not building complete solutions. A Play should be small enough to learn from in days, not weeks.
6
Card fields
Play name · Linked Intent · Intervention · Expected outcome · Test boundaries · Abort condition. No Hypothesis field — Bottleneck + Intent Signal already encode the causal claim.
Falsifiable
Expected outcome
If we ship [X], then [signal] moves from [baseline] to [target] within [timeframe]. If not, we stop / pivot.
Days
Test horizon
A Play is sized to produce evidence in days, not weeks. Multiple Plays run in parallel inside one Cell.
Key principles
✓ A Play is a single, falsifiable hypothesis.
✓ Every Play has an Abort condition.
✓ Continuous Stop in the Cell on Abort trip.
✓ Forum decides Scale · Deepen · Pivot · Stop.