Say the word “pre-mortem” in a team meeting and you will usually get one of two reactions: a raised eyebrow or a nervous laugh. Some people think it sounds morbid, others assume it is just risk management with a cleverer name. Both reactions come from a few persistent myths. Let’s go through six of them, because understanding what the technique is not makes it much easier to use well on an operations or engineering team.
Myth 1: “It’s just negativity with a fancy label”
Here is the thing: a good pre-mortem is not about being gloomy. It is about being specific. Teams that plan with enthusiasm tend to describe the path they hope for in vivid detail and the path they fear in vague terms. A pre-mortem simply gives the fears the same level of detail. The tone of the room is usually brisk and a bit playful, not dark. People leave with a short list of fixes, and most leave feeling more confident in the project, not less.
Myth 2: “We already have a risk register, so we’re covered”
A register is useful, but it contains the risks someone thought of. It rarely contains the awkward, interpersonal or unglamorous ones, such as “the on-call rota has a gap in the first week” or “the person who knows the integration is leaving.” A pre-mortem works as the discovery step that feeds the register. Think of it as the brainstorm that comes first, with the register as the tidy record afterwards.
Myth 3: “It only works for big, formal projects”
Not at all. The sweet spot is actually small and medium work: a migration, a new alerting setup, a change to how incidents are handled. For teams exploring how automation can reshape operations, a piece on what AIOps involves for operations teams shows how many moving parts even a modest rollout touches. Each moving part is a candidate for an imagined failure, and a thirty-minute session is plenty.
Myth 4: “The loudest, most experienced person will see all the risks”
Experience helps, but it also builds blind spots. The person who has run ten similar projects can miss what is different this time, and the newest team member often spots the thing everyone else has stopped noticing. That is why the format asks for silent, written notes before any discussion. It levels the field. A junior engineer writing “I don’t know where the deployment keys are stored” is giving the team valuable information, and the written format makes it easy to say.
Myth 5: “Security risks are for the security team to imagine”
This one is tempting, particularly in busy operations groups. In practice, the people who build and deliver software often have the best view of where things could go wrong in the delivery path, and the security specialists benefit from hearing it. A pre-mortem that includes both sides tends to surface issues neither would find alone. If you want a source to browse for the vocabulary, DevSecOps Now keeps a blog that focuses on securing CI/CD pipelines, which is a useful reference for the delivery-path part of the exercise. Keep the discussion at the level of “what could break or be misused here?” and leave the technical depth to the specialists afterwards.
Myth 6: “If we find too many problems, we will never launch”
Teams worry that opening this door releases a flood. What usually happens is the opposite. The list gets ranked, most items turn out to be cheap or acceptable, and a handful become real actions. The exercise converts a vague sense of unease into a finite to-do list, and a finite list is calming. Projects rarely stall because of the pre-mortem; they stall because of problems that nobody named until too late.
It helps to be explicit about what is allowed to remain a risk. Some items are simply accepted, and that is fine as long as the acceptance is a decision rather than an oversight. Write “accepted, because…” next to those lines so future readers know it was considered.
A bonus myth: “Once is enough”
A single session at the start of a project is better than none, yet the imagined failures change as the work progresses. Early on, the notes are about scope and people. Later, they are about integration, timing and handover. For machine-learning-heavy projects the failure picture shifts again once real data arrives; an explanation of keeping machine-learning products dependable at scale hints at how many new questions appear after launch. Short repeat sessions at milestones keep the list honest.
So what does a good session look like?
- A clear premise. “The project failed. We are writing the review.”
- Silent writing. Five to ten minutes, one reason per note.
- Grouping and voting. Cluster similar notes, then pick the top few.
- Owners and dates. Each chosen item gets both.
- A follow-up. Put a check on the calendar so the list is read again.
The honest limits
Debunking myths should not turn into overselling. The pre-mortem will not find every problem, and it cannot replace testing, careful design or experience. The research behind the underlying idea, that imagining an outcome as already real prompts more concrete explanations, is encouraging, but most real-world evidence is from practitioner reports. Treat it as a good habit rather than a law of nature.
It also asks something of the team: the willingness to hear uncomfortable notes without defending the plan. If that willingness is missing, no technique will help. Starting with a small, low-stakes project is a gentle way to build it.
Try it this week
Pick something modest that is about to start: a new alert routing rule, a tooling change, a handover of a service to another group. Gather four or five people, set a timer, and say the premise out loud. You may be surprised how quickly the room fills with useful, specific notes. And if someone tells you it sounds negative, you can smile and say it is the most optimistic thing you will do all week, because it is how you make sure the good outcome actually happens.
