How to run a premortem before a major business commitment

Run a business premortem before authorizing a major commitment by asking participants to imagine that it has failed and explain how. Collect separate contributions before discussion, then turn plausible failure routes into evidence checks and mitigation decisions. An imagined failure is a prompt for investigation, not a forecast of what will happen.

A premortem is an exercise that assumes a proposed project has failed and works backward to possible explanations before the project proceeds. Gary Klein’s 2007 article describes that method. The workshop below is an advisory adaptation for a business commitment; its preparation format and risk register are not presented as an experimentally validated package.

Choose a commitment the group can still change

Composite scene: You review an expansion proposal alone after a family evening. The presentation has a confident sponsor and a detailed launch narrative. You can identify who will explain the upside at the board meeting. You cannot identify who has been asked to describe a credible failure route before the commitment is made.

Use a premortem while a material choice remains open. If the commitment has already been made, identify what can still change: scope, sequence, operating safeguards, or the next approval. Do not describe an exercise as open consultation when management is unwilling or unable to consider any response.

Name the decision owner. Participants can supply evidence and challenge assumptions, but someone must decide what the findings mean for authorization. Keep legal, safety, finance, and specialist approvals within their proper authority.

Define failure in work terms

Give the group a specific hypothetical outcome. For an implementation, failure might mean that the system cannot support the intended customer process after the planned transition. Describe the work consequence without adding a dramatic catastrophe merely to generate energy.

Use a time horizon that matches the commitment and actual planning context. Distinguish failing to deliver the project from delivering a project that does not solve the customer problem. Both may matter, but they lead to different questions.

Avoid an outcome so broad that every concern becomes relevant. “The company failed” can produce an unmanageable discussion. A bounded commitment makes it easier to trace a proposed failure route to a current assumption, responsibility, or dependency.

The independent-risk premortem method

1. Give participants a common factual brief

State the proposed commitment, intended outcome, current evidence, constraints, and decisions still open. Mark estimates and unresolved assumptions. Include the operating people who understand important dependencies, not only the sponsors of the proposal.

Explain the process: the group will imagine a failure, collect possible reasons, examine evidence, and return to the decision. The exercise is not an accusation that the sponsor is incompetent or a contest to produce the most alarming scenario.

2. Collect separate failure explanations

Ask each contributor to write plausible reasons for the defined failure before hearing the senior sponsor’s account. Separate preparation is a proposed design choice intended to preserve distinct inputs. It is not a claim that independence eliminates influence or bias.

Ask for a mechanism. “Poor execution” is too broad. “The migration requires a data owner who has not been assigned” gives the group something to examine. Let contributors identify an uncertainty without inventing a probability or pretending they know the future.

3. Turn scenarios into testable questions

Read the explanations and group genuine duplicates. Preserve different routes that happen to lead to the same outcome. Ask what current evidence supports each concern and what information would weaken it.

For the hypothetical migration, a missing owner becomes a question about responsibility and capacity. A concern about customer adoption becomes a question about how the intended workflow has been tested. A vivid story with no supporting information stays a hypothesis until investigated.

4. Assign evidence checks and authorized mitigations

For each material concern, identify the next action, responsible person, and decision affected. An evidence gap may require verification. A verified dependency may require an authorized change to scope, timing, or resources. A specialist risk may need qualified assessment.

Do not turn every item into a general instruction to monitor closely. State what the owner will inspect and what result would trigger further action. If a proposed mitigation creates a new obligation or risk, obtain the appropriate approval before treating it as part of the plan.

5. Return to the commitment explicitly

The decision owner should state whether the proposal proceeds, changes, waits for information, or stops under the proper authority. Record the rationale and unresolved exposure. Explain what the group contributed and which issues need further work.

A completed workshop is not an approval. Keep formal review requirements separate. If management proceeds despite an unresolved concern, identify who accepted that exposure within their authority rather than implying the workshop proved the project safe.

Keep a risk register that changes the plan

Use fields for the imagined failure route, current assumption, supporting evidence, missing information, responsible owner, proposed response, and review trigger. Keep the working document short enough that a decision owner can use it.

In the migration example, a row might describe an unassigned data owner. The current assumption is that someone will provide source records and resolve discrepancies. The evidence check is to identify the authorized owner and confirm their capacity. The response depends on that answer; do not write “appoint an owner” as though the role and resource are already available.

Another row might concern a mismatch between the new workflow and a real customer task. The appropriate next step could be a permitted test or direct user inquiry. It should not become an invented testimonial that customers support the plan.

These examples are hypothetical. The actual register needs confirmed details from the relevant operational and specialist owners.

Prevent seniority from deciding which risks exist

Edmondson’s 1999 psychological-safety research concerns interpersonal risk-taking and team learning. A practical question for the workshop is whether a contributor can identify a weakness in the sponsor’s plan without being asked to defend their loyalty.

Set the conduct expectation before collecting input. Ask the sponsor to address the proposed mechanism rather than the person’s motives. Let participants correct facts. Preserve a material concern even when only one contributor raises it, while examining the evidence with the same standard used for other claims.

The research does not establish that a premortem creates a safe environment automatically. Look at how the concern is handled during and after the exercise. A person who supplies an inconvenient fact should not be made responsible for keeping the proposal comfortable.

Do not mistake a vivid scenario for a likely one

Tversky and Kahneman’s 1974 paper examines judgmental heuristics, including availability. It provides a caution about using ease of imagination as evidence. It does not assign probabilities to the risks generated in your workshop.

Ask what makes a failure route plausible in the actual context. Distinguish direct evidence, a reasonable mechanism, and a speculative possibility. Use qualified analytical judgment where likelihood or consequence estimates matter. Avoid adding precise percentages because the table has an empty probability column.

A severe but unsupported story may justify asking a question. It does not automatically justify stopping a project. A less dramatic, well-supported dependency may require immediate attention. Keep those differences visible in the recommendation.

Handle a concern that is outside the workshop’s scope

A participant may identify a legal, safety, employment, security, or financial issue that the group cannot assess properly. Preserve the concern and route it to the appropriate qualified owner. Decide what interim instruction is authorized while the assessment occurs.

Do not ask the room to vote on whether a legal obligation applies or whether a safety concern can be ignored. Likewise, do not let a specialist label end all discussion without clarifying the decision question the specialist must answer.

If the issue involves an allegation about a person, use the relevant formal process and restrict private information. The premortem is a planning exercise, not an improvised investigation.

Review the workshop after evidence arrives

Fischhoff’s 1975 experiments show how outcome knowledge can alter judgments about earlier uncertainty. Preserve the original risk notes and distinguish them from later discoveries. Do not rewrite a vague warning as a precise prediction after a problem occurs.

For the migration, compare the recorded ownership question with the actual handoff. If the check found capacity and the transition later failed for another reason, do not claim that the workshop anticipated the failure simply because it mentioned execution.

If a concern was raised but the assigned check never happened, identify that follow-through failure. If a check happened but the assumption later changed, identify the change. These are different lessons for the next commitment.

Industrial/organizational psychology, the psychology of work and organizations, connects individual judgment with the way teams gather and use information. The value of this exercise depends on that work, not on the dramatic quality of the imagined ending.

End with a usable commitment record

Before closing, confirm which plan changed, which information remains due, and who owns the next authorization. Give contributors an appropriate account of how their concerns were handled. Do not imply that every suggestion was adopted or that every possible risk has been discovered.

At the next approval point, reopen the register alongside the current proposal. Retire resolved items with their evidence. Keep unresolved material dependencies visible. The workshop earns its place when it changes an assumption, a check, or an authorized decision before the company makes a commitment it cannot easily undo.