Kill Criteria × Sunk Cost — Decision Hygiene Field Guide

Date: 2026-02-22 Category: systems / explore

Why this matters

Most teams don’t fail because they lacked effort. They fail because they kept investing after the original thesis was already broken.

Sunk cost bias turns past effort into fake evidence for future success.

This guide is about building a practical operating pattern: define kill criteria early, monitor reality honestly, and exit quickly when assumptions fail.

Core principle

Past cost is immutable. Future decisions should optimize future expected value only.

A useful question:

“If we were not already in this project, would we start it today with what we now know?”

If the answer is no, continuing usually means identity protection, not rational allocation.

Typical failure patterns

1) Identity lock-in

• “We are the team that does X.”
• Strategic review becomes emotional self-defense.

2) Escalation framing

• “We’ve spent too much to stop now.”
• Additional spend is justified by prior spend, not updated evidence.

3) Goalpost drift

• Original success criteria quietly change to avoid admitting failure.

4) Vanity progress

• Activity metrics (commits, meetings, prototypes) are mistaken for outcome progress.

Practical anti-bias protocol (KILL-6)

K1) Pre-commit kill criteria before major spend

Define at least 3 hard stop conditions:

• Market signal: no user pull / conversion floor not reached
• Technical signal: reliability or latency target repeatedly missed
• Economic signal: CAC/payback/maintenance burden outside threshold

Write them down before execution, not during stress.

K2) Separate owner and reviewer roles

• Owner presents progress.
• Reviewer (or red team) evaluates against pre-defined criteria.
• Reviewer is explicitly rewarded for truthful stop calls.

K3) Time-box with decision gates

Use explicit gates (e.g., 2 weeks / 6 weeks / 12 weeks):

• Continue
• Pivot
• Kill No implicit “keep going by default.”

K4) Use pre-mortem + stop-mortem

• Pre-mortem at kickoff: “Why might this fail?”
• Stop-mortem when killing: “What signal did we ignore longest?” Capture lessons for future selection quality.

K5) Track opportunity cost explicitly

For every ongoing initiative, list:

• What this blocks
• What we could do instead with same people/time

When opportunity cost is visible, bad projects die faster.

K6) Normalize graceful exits culturally

Treat clean shutdown as competence, not embarrassment.

• Celebrate “saved runway” and “freed focus.”
• Promote people who kill early for correct reasons.

A simple kill-scorecard template

Score each item weekly (0=bad, 1=uncertain, 2=good):

1. User pull evidence
2. Unit economics direction
3. Technical feasibility trend
4. Strategic fit
5. Opportunity cost burden

Interpretation:

• 8-10: continue
• 5-7: pivot with strict deadline
• 0-4: kill or freeze

This is not truth; it is a forcing function against narrative bias.

Signals you should kill now (not later)

• Repeatedly re-baselining deadlines without new constraints
• “One more sprint” language for >3 cycles
• Success depends on one unproven external assumption
• Team confidence rising while external evidence is flat/down
• No one can clearly state the falsification condition

Field checklist (10-minute review)

Before next sprint planning, ask:

1. What specific evidence improved this week?
2. Which pre-committed kill criteria are close to breach?
3. If forced to re-allocate 30% budget tomorrow, what gets cut first?
4. Are we protecting outcomes, or protecting ego/history?

If answers are defensive or vague, you already have your signal.

Closing thought

Disciplined stopping is not pessimism. It is portfolio intelligence.

Winning teams are not the ones that never start bad bets. They are the ones that detect quickly, exit cleanly, and redeploy relentlessly.