What is a pre-mortem? CATCS 2030 planning turns uncertainty into decisions

Takeaway	Detail
A pre-mortem only changes decisions when every named failure gets a probability, a tripwire date, a named owner, and a pre-decided response.	Thesis rule: without all four elements, the exercise is theater, not analysis.
Run the pre-mortem in 90 minutes before committing any 2030 plan.	Reader rule: the 90-minute session must output a failure register, not discussion notes.
Three estimation methods — reference class forecasting, Monte Carlo, and pre-mortem adjustment — cut schedule overruns from 67% to 28%.	Adam Analytics figure: 67% overruns reduced to 28% by communicating uncertainty.
Make risks, assumptions, scenarios, and uncertainty explicit before action.	Premortem purpose per Systematyka: reduce surprises before commitments are made.

This guide defines the pre-mortem and shows how a 2030 planning pre-mortem converts uncertainty into decisions. It delivers the failure-register format: probability, tripwire date, named owner, and pre-decided response for every named failure.

How a Pre-Mortem Actually Works

The pre-mortem was developed by psychologist Gary Klein as a structured exercise in what he calls "prospective hindsight" — the practice of imagining that a project has already failed and then working backward to explain why (theuncertaintyproject.org). Rather than asking "what could go wrong?" in the abstract, the team assumes the 2030 plan has already collapsed and generates specific reasons for that collapse. This temporal shift — from future possibility to past fact — is the engine that makes the technique work.

The mechanism is asymmetric in a way that matters for planning. Imagining failure is cognitively easier and produces more specific output than imagining success, because failure has a concrete endpoint to reason backward from while success remains diffuse (theuncertaintyproject.org). A team asked to list reasons the 2030 plan failed will surface risks — supplier concentration, regulatory shifts, key-person dependencies — that a standard "what could go wrong?" brainstorm often misses because the question feels hypothetical rather than real.

Systematyka frames the pre-mortem as a decision technique whose core function is to make risks, assumptions, scenarios, and uncertainty explicit before commitments are made (systematyka.ireshev.com). The goal is not prediction but reduction of surprises: by surfacing what could go wrong while there is still time to act, the team converts vague anxiety into named, addressable conditions. This is why the pre-mortem belongs before a 2030 plan is finalized, not after it is already in motion.

For the technique to change decisions rather than merely produce a list, each generated failure must be converted into a falsifiable checkpoint — a dated tripwire with a named owner and a pre-decided response. Without that conversion, the pre-mortem remains theater: a productive conversation that leaves no trace in the plan itself. The failure register is the artifact that bridges analysis and action.

The Evidence: What Pre-Mortems Move

Adam Analytics reports a system implementing three estimation methods — reference class forecasting, Monte Carlo simulation, and pre-mortem adjustment — reduced schedule overruns from 67% to 28% by communicating uncertainty. That 39-percentage-point drop (67 minus 28, computed digit by digit) is the strongest published quantitative result in this grounding, and it warrants scrutiny before you cite it in your own planning documents.

The figure comes from Adam Analytics' account of deploying these methods in combination, not from a standalone controlled study of pre-mortems alone. Before adopting the 67-to-28 range as your benchmark, verify two things in the underlying source: the sample size of projects measured and the precise outcome definition — whether "schedule overrun" means missing the original baseline date, a revised date, or a tolerance band. A result drawn from a narrow project set in one organization may not transfer cleanly to yours.

The convergence check matters here. The Adam Analytics result aligns with the psychological mechanism that makes pre-mortems plausible: the act of imagining failure and working backward surfaces risks that optimism bias and status quo thinking otherwise suppress. The improvement the report documents is consistent with the practice producing genuine predictive value, not merely giving participants a sense of having "done something" about risk.

That said, the Adam Analytics figure should be treated as a reported claim, not a guaranteed return. Use it to establish that structured uncertainty-reduction methods correlate with better schedule performance — and to justify the modest investment a pre-mortem requires — rather than as a promise that your own overrun rate will fall by 39 points.

A practical threshold for your own team: if your current schedule-overrun baseline sits anywhere near that 67% mark, the question is not whether to adopt a pre-mortem, but whether you can afford not to. If your baseline is already well below 28%, the pre-mortem still functions as a falsification tool — it tells you which assumptions are load-bearing enough to warrant a named tripwire date and a pre-decided response.

Options Compared: Which Pre-Mortem Format Wins

Not all pre-mortem formats produce the same output, and the difference matters more than most teams realize. Three formats dominate practice, and they differ in whether they actually change a decision.

Format A — free-form brainstorm. A team gathers, imagines the project has failed, and calls out reasons. Theuncertaintyproject.org and systematyka.ireshev.com both describe the pre-mortem's core purpose as making risk and uncertainty explicit before commitments are made — and a free-form session does surface risks quickly. But the output is a list of concerns without owners, dates, or thresholds. When no one is named as responsible for watching a specific risk, and no date is set for checking whether it has materialized, the list rarely survives the first week of execution. The risks are identified but not converted into checkpoints.

Format B — facilitator-led meeting with template. Asana and rock.so both describe running a pre-mortem with a facilitator guiding the team through a structured template. This adds discipline: the facilitator keeps the session on track, and the template ensures consistent categories of risk are explored. But a template alone does not guarantee tripwires. A team can fill in every field of a well-designed template and still produce risks that are undated, unowned, and unconnected to a pre-decided response. The structure improves the conversation; it does not ensure the output changes behavior.

Format C — failure register with probability, tripwire date, owner, and pre-decided response. This format wins because it converts each named failure into a dated, owned, falsifiable checkpoint. Every entry in the register carries four fields: a probability estimate, a tripwire date by which the team will check whether the risk is materializing, a named owner responsible for monitoring it, and a pre-decided response that activates if the tripwire is triggered. The register transforms the pre-mortem from a one-time conversation into a living document that shapes decisions over time. When a tripwire date arrives, the owner either confirms the risk has not materialized or triggers the pre-agreed response — either way, the team acts on information rather than hope.

The test is simple: if your pre-mortem output cannot be checked on a specific date by a specific person, it is theater, not analysis. Format C is the only one that passes this test by design.

Costs and Numbers That Matter

Budget 90 minutes for the workshop itself. That is long enough for a facilitator-led team to generate failures, rank them, and assign each one an owner and a tripwire date, and short enough that a 2030 planning team will actually put it on the calendar instead of deferring it indefinitely. The facilitator-led format described by Asana and Rock.so assumes a single moderator who keeps the group imagining the plan has already failed and working backward; 90 minutes fits that structure without turning into a second planning meeting. If your team cannot hold 90 minutes, cut the attendee list, not the exercise.

Set a hard floor of 3 register entries before the plan can be funded. Fewer than 3 usually means the team is protecting the plan rather than stress-testing it — the failures named are the ones that are safe to name, not the ones that would actually kill the 2030 commitment. Three is the minimum that forces at least one uncomfortable entry onto the page. If the room produces only one or two, the facilitator should treat that as a signal to re-run the exercise with a different framing, not as evidence that the plan is robust.

Register field	What it must contain	Test it has to pass
Failure	A specific named way the 2030 plan dies	Can you picture the failure happening on a date?
Probability	A rough likelihood, stated by the owner	Would the owner bet on it at those odds?
Tripwire date	The date by which an early signal should appear	Is it before the point of no return?
Owner	One named person, not a committee	Can you say their name out loud?
Pre-decided response	The action taken if the tripwire fires	Is it decided now, not later?

The cost of skipping this is visible in the baseline. Adam Analytics reports that unadjusted estimates carried a 67% schedule-overrun rate, and that a system combining reference class forecasting, Monte Carlo simulation, and pre-mortem adjustment reduced that to 28% by communicating uncertainty. The 39-percentage-point gap is the size of the prize, but it only shows up when the pre-mortem output is a register with dates and owners — not a list of concerns. A register that names failures without tripwire dates cannot capture any of that adjustment, because there is nothing to adjust against.

The cost of running it badly is smaller but real: 90 minutes of a planning team's time, plus the follow-up needed to keep tripwires live. Weigh that against the alternative — funding a 2030 plan whose failure modes were never written down, never dated, and never assigned. The arithmetic favors the workshop every time, provided the output meets the 3-entry floor and every entry carries a tripwire date and a named owner.

Before you commit the 2030 plan, check three things: the workshop was scheduled for 90 minutes, the register holds at least 3 entries, and every entry has a probability, a tripwire date, a named owner, and a pre-decided response. If any of those is missing, you have theater, not analysis — and the plan should not be funded until it is fixed.

What the Evidence Does NOT Establish

Start with the attribution problem, because it is the one that should most change how you read every pre-mortem number you have seen. Adam Analytics reports that a system combining three estimation methods — reference class forecasting, Monte Carlo simulation, and pre-mortem adjustment — cut schedule overruns from 67% to 28%. That is a real result, but it is a bundled result. The pre-mortem was never run alone, so no part of that 39-percentage-point improvement can be assigned to it. If you are building a 2030 business case and someone cites that figure as proof that a pre-mortem works, the honest response is: which of the three methods moved the number, and how would we know? No grounding source isolates the pre-mortem's independent effect from Monte Carlo or reference class forecasting. Treat the figure as evidence that the package communicates uncertainty better than the baseline, not as a measured effect size for the pre-mortem itself.

The second gap is study design. None of the available sources sources reports a controlled trial of a pre-mortem against a placebo planning meeting — same attendees, same duration, same room, but no failure-imagining prompt. Without that comparison, the observed gains could come from the meeting itself: gathering the team, forcing a discussion, writing risks down. That does not make the pre-mortem useless; it makes the effect sizes directional rather than precise. The practical rule for your 2030 plan: use the pre-mortem to generate candidate failures, then judge it by whether the register changes a decision, not by whether the workshop felt productive. A session that produces a long risk list and no changed commitment has demonstrated nothing.

There is also an edge case where the method adds little. When the 2030 plan has a single dominant, already-known risk, the pre-mortem tends to return one entry — the risk everyone already named — and the exercise becomes a formality. The check is simple: before you schedule the session, ask whether the team can already name the top failure and its response. If yes, skip the workshop and spend the time on the response instead. The pre-mortem earns its cost when the failure space is genuinely uncertain, not when it is a single known hazard wearing a workshop's clothing.

What the evidence does establish is narrower and still useful. The sources agree on the mechanism: make risks, assumptions, scenarios, and uncertainty explicit before action, so surprises shrink and commitments get tested against named failure modes. That is a claim about process discipline, not about a measured causal effect. So the defensible position for a 2030 plan is this: run the pre-mortem because it forces explicitness, and hold it to the standard of a dated, owned, falsifiable register — not because a percentage proves it works.

Claim	What the sources support	What they do not
67% to 28% overrun reduction	A bundled three-method system improved outcomes	Any pre-mortem-only effect
Pre-mortem improves plans	Mechanism: explicit risks before commitment	Controlled trial vs. placebo meeting
Effect size	Directional signal	Precise, transferable estimate

2030 Plan Failure Register

A failure register is the only artifact from a 2030 pre-mortem that survives contact with the actual decade. Build it in a single 90-minute session with a named facilitator and a fixed workshop date on the calendar, because an unscheduled pre-mortem is a wish. The worked example below assumes a 2030 plan to deploy an AI forecasting tool across a 40-person team — a plan with enough surface area for adoption, data, and trust failures to compound quietly.

Checkpoint 1 is generation, and it is deliberately wasteful. Ask the team for more candidate failures than you intend to keep — the surplus forces people past the obvious ("the model is wrong") into the operational ("nobody trusts the output by month four"). Then rank each candidate on probability times impact and keep only the top three. The cut is the point: a register with a dozen entries gets read once, while a register with three gets defended. If your team cannot produce more than three, that is itself a finding — the plan has not been imagined as a failure yet.

Checkpoint 2 converts each surviving failure into a tripwire: a date plus a measurable signal. "Adoption stalls" is not a tripwire. "Adoption below 30% at 90 days post-rollout" is, because someone can check it without a meeting. Assign a review date well before 2030 to each tripwire so the register is tested while the plan is still cheap to change. Every tripwire needs a named owner — a person, not a function — and a pre-decided response written in advance, so the response is a decision already made rather than a debate held under pressure.

Three rules keep the register honest. First, no tripwire without a date — an undated risk is a mood. Second, no owner without a name — a shared owner is no owner. Third, no response that requires a new decision — write the response now, while nobody is embarrassed. The register is reviewed at each tripwire date, not at the next quarterly offsite.

Run this and the pre-mortem produces a document you can act on in 2027. Skip it and you have produced a conversation — pleasant, forgettable, and useless the moment the plan meets reality.

Decision Rules for 2030 Planning

The pre-mortem earns its keep only when it ends in a failure register that functions as a decision instrument. That means every entry carries four fields: a probability, a tripwire date, a named owner, and a pre-decided response. Strip any one of those and the entry stops being a risk and becomes a sentiment. The five rules below are the ones I apply before any 2030 plan gets funded, and they are deliberately mechanical — the point is to remove judgment from the moment when judgment is most likely to be wishful.

Rule 1: If a failure has no named owner, delete it. An unowned risk cannot change a decision, because no one is accountable for watching it or acting on it. When the register is reviewed, the facilitator reads each entry aloud and asks for the owner by name — not by function, not by "the team." If no name lands in the room within a few seconds, the entry comes out. This is not a punishment; it is a filter. A register padded with orphan risks looks thorough and functions as decoration.

Rule 2: If a failure has no tripwire date, it is a worry, not a risk. A tripwire is a specific date on which someone checks a specific signal and either escalates or stands down. "Supply chain disruption" is a worry. "By March 2027, the primary supplier's lead time exceeds the contracted ceiling — owner: procurement lead — response: qualify the secondary source" is a risk. Convert every undated entry into a dated signal or drop it. The register should be readable as a calendar, not as a mood board.

Rule 3: If the register has fewer than three entries, re-run the workshop with a different facilitator before funding the plan. A short register is almost always a facilitation failure rather than evidence of a robust plan. A different facilitator changes the room's dynamics: who speaks first, which assumptions get challenged, whether dissent is safe. Re-run before the money moves, not after.

Rule 4: If a tripwire date has passed with no response logged, the plan goes back to the owners, not to the next meeting. The register is a live document. A passed date with no recorded action means the checkpoint was theater. The remedy is a same-week review with the named owner, not a note in the minutes.

Rule 5: If the pre-decided response is "we'll decide then," rewrite it now. The value of a pre-mortem is that it forces the decision while the room is calm. A response field that defers the choice simply relocates the original problem to a worse moment. Write the action, the trigger, and the owner in the same sentence.

Rule	Condition	Action
1	No named owner	Delete the entry
2	No tripwire date	Convert to dated signal or drop
3	Fewer than three entries	Re-run with a different facilitator
4	Date passed, no response logged	Same-week owner review
5	Response defers the decision	Rewrite with action, trigger, owner

Run these five checks in order at the close of the workshop. What survives is a register you can actually manage against — and a 2030 plan that has earned the right to be funded.

Also worth reading: How Michael Rustom Turns Tech Investment Into Real World Results: How Michael Rustom Turns Tech · The Ancient Guild System How Medieval Craftsmen Created Innovation Ecosystems Through Knowledge, Collaboration, and Workshop Culture: Ancient Guild System How Medieval · The Evolution of African Political Philosophy How 2025's NEWLAMP Workshop Bridges Historical and Contemporary African Thought: Evolution of African Political Philosophy

What to do next

Step	Action	Why it matters
1	Book a 90-minute pre-mortem before any 2030 plan is committed, and state the output requirement up front: a failure register, not discussion notes.	Per the canonical decision rule, the session only counts if it produces the register — otherwise it is theater, not analysis.
2	For every named failure in the register, fill all four fields: a probability, a tripwire date, a named owner, and a pre-decided response.	Takeaway Detail is explicit: without all four elements the exercise changes no decisions.
3	Estimate each probability using the three methods — reference class forecasting, Monte Carlo, and pre-mortem adjustment — and communicate the uncertainty to the plan's sponsors.	These three methods cut schedule overruns from 67% to 28%, the Adam Analytics figure for communicating uncertainty.
4	Set tripwire dates so the earliest ones land inside the 4-week and 6-week checkpoints ahead of the 2027 milestone, not at the 2030 commitment date.	A tripwire that fires after the commitment cannot trigger a pre-decided response in time.
5	Make risks, assumptions, scenarios, and uncertainty explicit in the register before action, following the Systematyka purpose: reduce surprises before commitments are made.	Explicit uncertainty is what converts the pre-mortem from a discussion into a decision input.
6	Refuse to fund the 2030 plan until the failure register exists with all four fields populated for every failure.	This is the canonical decision rule — the register is the gate, not a deliverable to be completed later.

Frequently Asked Questions

What four elements must every named failure have for a pre-mortem to actually change decisions?

Every named failure needs a probability, a tripwire date, a named owner, and a pre-decided response.

How long should a pre-mortem session run before committing a 2030 plan?

The pre-mortem should be run in 90 minutes before committing any 2030 plan.

What must the 90-minute session produce instead of discussion notes?

The 90-minute session must output a failure register, not discussion notes.

By how much do the three estimation methods cut schedule overruns?

Reference class forecasting, Monte Carlo, and pre-mortem adjustment cut schedule overruns from 67% to 28%.

Who developed the pre-mortem and what did he call the underlying practice?

Psychologist Gary Klein developed the pre-mortem as a structured exercise in what he calls "prospective hindsight."

What is the stated purpose of a pre-mortem according to Systematyka?

Per Systematyka, the pre-mortem's purpose is to reduce surprises before commitments are made.

Quick answers

What four elements must every named failure get for a pre-mortem to change decisions?	Every named failure must get a probability, a tripwire date, a named owner, and a pre-decided response.
What is the thesis rule about a pre-mortem missing all four elements?	Without all four elements, the exercise is theater, not analysis.
How long should the pre-mortem run before committing any 2030 plan?	Run the pre-mortem in 90 minutes before committing any 2030 plan.
What must the 90-minute session output according to the reader rule?	The 90-minute session must output a failure register, not discussion notes.
Who developed the pre-mortem and what did he call it?	The pre-mortem was developed by psychologist Gary Klein as a structured exercise in what he calls "prospective hindsight."