Claude pre-mortem on a planned project or decision
intermediateClaude OpusPersonal ProductivityDecision Makingframeworkmethodologydecision-makingriskpre-mortem
Use case
Use this prompt before launching a project, signing a vendor, or making a hire — any decision where the cost of failure is high enough that you want to imagine the post-mortem before living it. The XML scaffolding forces Claude to actually generate distinct failure modes rather than vague risk bullets.
The prompt
You are a seasoned operator running a pre-mortem on a plan that has already been approved. The goal is not to kill the plan — it is to surface the most likely ways it fails so the team can mitigate them in advance. <context> Plan or decision:{{plan}}Owner:{{owner}}Time horizon:{{time_horizon}}Definition of success:{{success_criteria}}What could go wrong (the user's current intuition):{{known_risks}}</context> <task> Imagine it is{{time_horizon}}from today and the plan has clearly failed. Work backwards. 1. Generate 8 to 12 distinct failure modes. Span at least four categories: execution, market, team, and external. Do not duplicate by rephrasing. 2. For each failure mode, score: - Likelihood (1 to 5) - Impact if it happens (1 to 5) - Detectability before it bites (1 to 5, where 5 = will be obvious early) 3. Rank the failure modes by Likelihood times Impact divided by Detectability. The highest scores are your priority risks. 4. For the top 5 risks, propose one specific, testable mitigation each — something the owner can put on a calendar or a checklist this week. </task> <output_format> Return a markdown table for steps 1 to 3, then a numbered list for the top 5 mitigations. </output_format> <constraints> - If a risk is generic ("team gets distracted"), make it concrete ("Q3 hiring backfill pulls the lead engineer off this project for 4+ weeks"). - If you do not have enough information about the plan to score a risk credibly, say so and ask one clarifying question rather than inventing. - Do not include risks below 6 on the combined score. Tight is better than long. </constraints>
Variables
{{{{plan}}}}Replace with your {{plan}}{{{{owner}}}}Replace with your {{owner}}{{{{time_horizon}}}}Replace with your {{time horizon}}{{{{success_criteria}}}}Replace with your {{success criteria}}{{{{known_risks}}}}Replace with your {{known risks}}
Example
Input
plan: Launch a self-serve tier of our B2B product in 90 days, targeting solo operators and 2-5 person teams. owner: Head of Product time_horizon: 6 months success_criteria: 500 self-serve signups, 8% convert to paid, CAC payback under 4 months known_risks: pricing might be wrong; engineering capacity is tight; sales team may resist
Output
| # | Failure mode | Cat. | L | I | D | Score | |---|---|---|---|---|---|---| | 1 | Self-serve tier cannibalizes mid-market sales motion before PLG flywheel takes hold | Market | 4 | 5 | 2 | 10.0 | | 2 | Engineering capacity gets pulled to enterprise reliability fires, slipping launch by 6+ weeks | Execution | 4 | 4 | 4 | 4.0 | | 3 | Solo-operator ICP churns at 15%+ monthly because product was built for teams | Market | 3 | 5 | 3 | 5.0 | | 4 | Pricing is set without willingness-to-pay research, lands either too low (margin) or too high (conversion) | Execution | 4 | 4 | 3 | 5.3 | | 5 | Support volume from low-ACV self-serve users overwhelms the team, degrading enterprise CSAT | Execution | 4 | 3 | 2 | 6.0 | | 6 | Sales reps actively steer prospects away from self-serve to protect commission | Team | 3 | 4 | 4 | 3.0 | ... Top 5 mitigations: 1. Cannibalization: Add a hard ICP gate (company size, use case) on the signup flow for the first 60 days. Review weekly with sales. 2. Pricing: Run 30 willingness-to-pay interviews in weeks 1-3, anchor pricing on outcomes not seats. 3. Support load: Stand up a self-serve-only support queue with deflection-first triage before launch day. 4. Cannibalization (sales comp): Pre-decide a comp adjustment so reps are neutral on which path a deal takes. 5. Capacity: Freeze enterprise feature scope for the launch quarter; publish the freeze to sales in writing.
Tips for best results
- 1The detectability score is what makes this version sharper than the typical 'list of risks' pre-mortem — risks you cannot detect early are the truly dangerous ones, and the formula surfaces them.
- 2Force category coverage in the prompt. Without it, Claude will generate eight versions of 'execution risk' and miss market and team failure modes entirely.
- 3Run this with the team after Claude generates the draft. Use Claude's list as a starting point, then have humans add risks they would never put in writing.
- 4If Claude scores everything 4 or 5, push back: ask it to re-score using the actual base rate of similar projects failing this way.
- 5Save the output as a decision artifact. Six months later, compare actual failures against the predicted list — it is the fastest way to calibrate your risk intuition.
Related prompts
Inverse thinking — how would this fail?
intermediateInvert the goal. Instead of asking how to succeed, ask how to guarantee failure — then avoid those paths. Forces Claude to map the failure surface, not just the success path.
Personal Productivityframeworkmethodologyinversion
Second- and third-order consequences of a decision
advancedMap the chain of consequences from a proposed action — not just what happens, but what happens because of what happens. Surface incentive shifts, downstream behaviors, and second-order risks.
Personal Productivityframeworkmethodologysystems-thinking
Disciplined 5 Whys with explicit causal chain
intermediateRun a 5 Whys analysis that follows a real causal chain and refuses to stop at 'human error,' surfacing the systemic cause and the testable countermeasure.
Operationsframeworkmethodologyroot-cause
Need help implementing this prompt in your workflow?
Book a call