Pre-Mortem Template

Why pre-mortems beat risk brainstorms

A standard kickoff risk session goes like this: someone asks "any risks we should flag?", three people contribute generic risks ("resource constraints", "shifting priorities"), the team nods, the risks get logged, nobody believes them.

A pre-mortem inverts the framing. You tell the team: "It's six months from now. The program failed. We're doing the post-mortem. What happened?" Suddenly people are specific. They name the executive who lost interest in month three. They name the workstream that depended on a team that's already overloaded. They name the assumption that's load-bearing and untested.

"Optimism is the enemy of useful risk identification. Pre-mortems are how you make the team safe enough to be honest."

This template structures the workshop so the framing is right and the output is usable. The cheat sheet covers the facilitation moves that surface the real concerns instead of the polite ones.

When to run a pre-mortem

The default answer is "at kickoff." That's right but incomplete. The three best moments:

• Kickoff. Before the team is invested in the plan, before optimism calcifies. Highest-leverage moment.
• After a major scope change. The original pre-mortem covered the original program. New scope = new failure modes worth surfacing.
• Before a high-stakes phase. Going into beta, into a regulatory review, into an exec demo: a 60-minute pre-mortem can catch the failure mode that nobody wanted to be the one to name.

The output of the pre-mortem feeds your RAID log (the surfaced risks), your program charter (the constraints and assumptions), and your communication plan (the stakeholders who need previewing on the top risks).

Building a pre-mortem worksheet in Excel

A pre-mortem does not fail in the first fifteen minutes. It fails in the last fifteen. The room is good at imagining failure, and a well-framed session produces twenty scenarios without much effort. What the room is bad at is deciding which four of those twenty are worth spending money on, because twenty items in an unordered list all feel equally urgent while somebody is describing one of them out loud. The worksheet has to do the ranking, and it has to do it live, on the screen, while people are still in the room. Here is how the pre-mortem template Excel worksheet is built.

Nine columns, filled left to right during the session.

1. ID. So a failure can be referenced in the charter, the RAID log, and later the post-mortem without anyone retyping it.
2. The imagined failure. Past tense, as something that already happened. "The data team never started our migration" is a pre-mortem entry. "Resourcing risk" is a category pretending to be one.
3. Category. Six values, dropdown, no free text.
4. Likelihood, 1 to 5. Dropdown.
5. Severity, 1 to 5. Dropdown.
6. Score. Computed. Nobody types in this column.
7. Mitigation. Something a person does by a date, not something a person watches.
8. Owner. One name.
9. Due date. One date. Required on anything that scores high.

Two more columns sit to the right of those nine, J and K, and both are computed: the rank and a checker. They are the two that make the session work.

Make it a Table first. Select the header row and one row of data, press Ctrl and T. In a live workshop that step is worth more than it looks. When someone calls out failure number nineteen and you type it at the bottom, the score, the rank, and the flags fill themselves in, and you never break the rhythm of the room to drag a formula down. Freeze the header with View › Freeze Panes so row 3 is still visible when you are on row nineteen.

The ID. With the header on row 3:

="PMF-"&TEXT(ROW()-3,"00")

PMF-01 on the first entry. Two digits, because a session that produces more than ninety-nine imagined failures has a problem this sheet cannot fix.

Dropdowns on the three typed-value columns. Select the likelihood column, Data › Data Validation › Allow: List, Source: 1,2,3,4,5. Same for severity. Then set the Error Alert style to Stop, not Warning. That is not fussiness. In a live session somebody will type "4-5" or "high" into a scoring column, and one text value in a numeric column turns the ranking into an error at the exact moment you need it on screen. The category column gets its own list: Scope,Dependency,People,Technical,Sponsor,External. Six buckets is enough to cluster with, and few enough that the room does not spend ten minutes arguing about which bucket a failure belongs in.

The score. Likelihood times severity, blank until both are filled:

=IF(OR($D4="",$E4=""),"",$D4*$E4)

Guarding on the blanks matters more than it looks. Without the guard, an unscored row reads as a zero, and zero sorts to the bottom, which is exactly where a failure nobody has scored yet does not belong.

The rank, which is the column the session actually runs on. Put this in J:

=IF($F4="","",RANK.EQ($F4,$F$4:$F$40)+COUNTIF($F$4:$F4,$F4)-1)

RANK.EQ on its own gives tied scores the same number, so three failures scored 16 all come back as rank 2 and there is no third place. The COUNTIF tail breaks the tie by row order: the range $F$4:$F4 is anchored at the top and open at the bottom, so it counts how many times this score has already appeared above. First occurrence adds nothing, second adds one, third adds two. The result is a clean 1, 2, 3 with no duplicates, which is what lets you say "we are discussing the top three" and have that phrase mean something specific.

Conditional formatting so the top of the list is visible rather than calculated. Select A4:K40, then Home › Conditional Formatting › New Rule › Use a formula, red fill:

=AND($J4<>"",$J4<=3)

Dollar signs on the column letters, bare row number, so the one rule colors whole rows across the block. The top three go red as soon as the scores land, and they re-color themselves the moment someone argues severity up from 3 to 5 in the last ten minutes. That live re-ranking is the reason to score in the room instead of after it.

The COUNTIF that catches what pre-mortems are famous for. A pre-mortem produces a good list and no owners. Everyone leaves agreeing that the migration dependency is the scary one, nobody's name is next to it, and eleven weeks later it is the reason the program is late. Column K:

=IF(AND($E4>=4,$H4=""),"High severity, no owner","")

And one count in a cell at the top of the sheet, the cell the facilitator watches in the last five minutes:

=COUNTIFS($E$4:$E$40,">=4",$H$4:$H$40,"")

The rule I hold to: that cell reads zero before anyone leaves the room. Not before the summary goes out. Before anyone leaves the room. A name assigned in the meeting sticks. A name assigned by email on Thursday is a name somebody else volunteered you for.

Excel and Google Sheets. The worksheet is an .xlsx and runs in both, which matters more here than on most templates, because half of these sessions are remote and a shared Sheets tab is the fastest way to let eight people watch the ranking move. IF, AND, OR, COUNTIF, COUNTIFS, RANK.EQ, TEXT, and ROW behave identically. Three honest differences. The Excel Table imports as a plain filtered range, so rows typed at the bottom during the session do not inherit the score and rank formulas: fill those formulas down to row 40 before the meeting starts and the problem goes away. The Stop-style error alert on the 1-to-5 dropdowns arrives in Sheets as "Show warning," which lets "4-5" straight through, so re-select "Reject input" after import or the guardrail is decorative. And Sheets has no "Stop If True" on conditional formatting, so if you add rules beyond the top-three one, order them deliberately: when two rules set the same property on the same cell, the higher rule in the list wins and there is no switch to halt the rest.

A worked pre-mortem example

Sixty minutes, eight people, a payments and checkout program six weeks out from kickoff. The framing was the standard one: it is now November, the program has failed, we are sitting in the post-mortem, what happened? Six of the entries that survived the clustering pass, scored in the room.

Twenty-one entries came out of that hour. These six are what was left after clustering, and the ranking is what made the last twenty minutes worth attending. The room wanted to spend its energy on PMF-04, because duplicated build work is the failure engineers find most vivid and the easiest one to argue about in detail. The sheet said PMF-01, at a score of 20, and PMF-01 was the one nobody had a mitigation for yet. That is the trade the ranking buys you. Not a better list. A better argument.

PMF-05 is the row worth looking at twice. Severity 5, likelihood 2, no owner, so the checker column fired and the row went amber on the shared screen with four minutes left. It got a name before the meeting ended. It was also the only one of the six that nobody had raised in the standard kickoff risk round two weeks earlier, because "legal might be slow" is not a thing people volunteer while the sponsor is on the call and the mood is good.

Six rows. Sixty minutes. The next artifact is the post-mortem, and its job is to tell you which of these six you got right.

Questions PMs ask

How long does a pre-mortem workshop take?

60 minutes is the sweet spot. The template's facilitation guide is structured around a 60-minute session: 5 min framing, 15 min individual brainstorm, 20 min cluster and discuss, 15 min severity score and ownership, 5 min wrap-up.

How is this different from a risk brainstorm?

Framing. A risk brainstorm asks "what could go wrong?", and produces generic answers. A pre-mortem asks "the program has failed, why?", which forces specificity and gives social cover for surfacing concerns people don't want to be the first to name.

What format is the template?

Excel (.xlsx) facilitation worksheet plus a PDF cheat sheet for the facilitator. Also opens in Google Sheets if you prefer to run it as a shared doc during the workshop.