practice · operations
Kill criterion
A specific signal that, by a specific date, kills the initiative. Written when the Initiative Brief is signed, read at every portfolio review, executed when the date passes and the signal hasn't moved. Without the kill criterion, the brief is a wish.
TL;DR
A kill criterion is one sentence with a specific signal, a specific threshold, and a specific date. Written into the Initiative Brief, signed by the PO and Leadership. Read at every portfolio review. When the date passes and the signal hasn't moved, the initiative is killed — not "paused", not "revisited", killed.
What it is
The kill criterion is the most disciplined act in the corpus's portfolio practice. It is what converts an initiative from open-ended ambition into a time-bounded bet. It is the line in the Initiative Brief that protects everything downstream from accumulation.
Distinguish from
Success metric — what the bet is trying to prove. Milestone — a scheduled deliverable. Deadline — when work is due. Kill criterion — when the team is freed, regardless of what was delivered. See Confusable with at the foot.
Why it matters
Without a kill criterion:
- Initiatives never close. They accumulate, drain capacity, and degrade the portfolio.
- The bet is unfalsifiable. A claim that can't be refuted is hope, not a bet.
- The team optimises for continuation. Without a kill, the team learns that initiatives "always continue" — and they stop honestly checking signals.
- The portfolio review has no teeth. Without kill criteria to read, the review becomes a status meeting.
The corpus rule from Practice · Writing Initiative Briefs · Step 7: if you cannot write a kill criterion, the initiative is a wish. Re-walk Why We Build · Initiative Identification.
How to do it
Step 1 — Write the kill criterion when the Initiative Brief is signed
Not later. Not after cycle 1. At signing — when the team's expectations are still negotiable.
Kill criterion (2026-Q2 Hebrew-grading initiative):
If by 2026-09-01 the median focused-grading time across
observed grading cycles has not fallen below 25 minutes
(from the 47-minute baseline), the initiative is killed
and the team is freed for next-quarter work. The bet's
mechanism (focused-time recovery via workaround removal)
is considered refuted.Three elements, each mandatory:
- Specific signal — median focused-grading time. Not user happiness; not velocity.
- Specific threshold — below 25 minutes. A number.
- Specific date — 2026-09-01. A real date in a real calendar.
Plus a fourth, often forgotten: what mechanism the kill refutes. The signal failing to move tells the team what to stop believing, not just what to stop doing.
Step 2 — Make the criterion kill-able, not adjust-able
A common failure: writing kill criteria that are easy to re-interpret when the date approaches. Watch for:
| Weak criterion | Why it weakens | Sharper alternative |
|---|---|---|
| If signal is "not on track" | "On track" is interpretable | Signal below 25 min |
| By end of Q2 | Quarter boundaries drift | By 2026-09-01 |
| Then we'll reconsider | "Reconsider" ≠ kill | Then the initiative is killed and the team freed |
| Unless extenuating circumstances | The exception swallows the rule | No exception clause. If a real exception exists, write a new criterion. |
The sharper version isn't more harsh — it's more honest. It tells the team what to do.
Step 3 — Read the criterion at every portfolio review
The criterion lives in the Initiative Brief. At every quarterly portfolio review, the leader running the review reads the criterion aloud:
Initiative: Hebrew-name grading flow
Kill criterion: by 2026-09-01, median focused-grading time
below 25 min.
Current signal (read 2026-08-15): median 13 min on n=9
observed cycles. Criterion met with margin; on track.
Decision: KEPT.If the date hasn't passed yet, the read is are we still on track? If the date has passed, the read is did we meet the threshold? — and the answer determines kept vs killed.
Step 4 — Execute the kill when the date passes and the signal hasn't moved
This is the hardest step. The team will be invested. The PO may be invested. Killing feels like loss.
The corpus's framing: killing an initiative is the highest-leverage portfolio move. Every quarter without a kill is a quarter of accumulating drag. The team is freed — not failed. The bet was refuted; the team now knows something they didn't.
The killing motion:
- Read the kill criterion aloud at the review.
- Read the signal aloud.
- Name the gap. "The signal didn't move; the threshold wasn't met."
- Declare the kill. "This initiative is killed as of today; the team rotates to [next priority] at the start of next cycle."
- Write the kill brief.
No long debate. The criterion did the debating six months ago.
Step 5 — Write the kill brief
A killed initiative is not erased; it is closed honestly. The kill brief is one page:
Kill brief — AI-Assisted Feedback initiative
What we bet on:
If AI suggestions reduce feedback-writing time by 50%,
graders complete more submissions per session.
What the data said (2026-09-01):
Acceptance rate of suggestions: 12%. Most accepted
suggestions were edited substantively. Feedback-writing
time fell <10% — well below the 50% target.
What we now know:
Graders do not experience feedback-writing as the slow
step. The real friction is Hebrew-name lookup (since
addressed in the Hebrew-grading initiative). The AI
approach was solving the wrong problem.
What we'll do differently:
Discovery should have surfaced the wrong problem before
the initiative was named. The kill criterion was honest;
the brief's framing was not.
Update to corpus: clinics/a-brief-written-from-the-solution
-backwards.md has been amended with this case.The kill brief lives next to the Initiative Brief, dated, signed.
Step 6 — Read the kill brief when the next adjacent initiative is briefed
The discipline against repeating refuted bets: when the next initiative in the same problem space is being named, the PO reads the kill briefs of related-killed initiatives. The new Initiative Brief acknowledges what was learned.
Evidence
Across our cycles, kill criteria that worked shared three properties.
- They had a specific date, not a quarter boundary. Criteria dated to specific calendar days were executed 80% of the time when missed; criteria dated to "end of Q2" were renegotiated 60% of the time.
- They named the mechanism, not just the metric. Criteria that articulated what the team would stop believing on kill produced cleaner kill briefs and better next-cycle initiatives.
- The signal was witnessable, not survey-able. Criteria built on observed numbers fired cleanly; criteria built on "the team thinks" were always re-interpretable.
Anti-patterns
| Pattern | What it looks like | Where to fix |
|---|---|---|
| No kill criterion | Initiative Brief without the section | Clinic — An initiative without a goal |
| Soft language | "Reconsider", "reassess", "revisit" | Replace with killed. Re-read Practice · Step 2 |
| Criterion ignored at review | The criterion is in the brief but nobody reads it aloud | Read it aloud at every portfolio review |
| Date passes, criterion missed, initiative continues | The team renegotiates instead of killing | The criterion is the criterion. Kill or rewrite the criterion, but don't ignore. |
| No kill brief on a killed initiative | The team just rotates off | Write the brief; the next adjacent initiative reads it |
| Kill criterion is the success metric | If we meet the target, continue; if not, kill | The kill criterion is usually a higher threshold than the success metric — missed by this much not barely missed |
Confusable with
| This | Not this | Difference |
|---|---|---|
| Kill criterion | Success metric | Success = what we're aiming for. Kill = what frees the team if we miss by enough. |
| Kill criterion | Milestone | Milestone = scheduled work. Kill = scheduled question. |
| Killed | Paused | Killed = team freed, brief closed. Paused = unclear, usually killed in disguise. |
| Mechanism refuted | Implementation failed | Mechanism = the theory. Implementation = the execution. Kills are about the theory. |
Further reading
- Canon — Did We Serve · The Portfolio · When to stop
- Practice — Writing Initiative Briefs — where the kill criterion is born
- Practice — Quarterly portfolio review — where the criterion is read and executed
- Clinic — An initiative without a goal
- Template — Initiative Brief — includes the kill criterion section
- Skill path — Leadership foundations · Step 9 · PO foundations
- Principle — Predictions over plans — a kill criterion is the most disciplined form of prediction