A pre-mortem inverts the post-mortem. Instead of asking what could go wrong, you assume the campaign has already failed and ask the team to explain why. The reframing matters, because people who will not voice doubts about a plan will happily explain a failure.
Running it
Do it after the plan exists and before spend commits. Ninety minutes, everyone who touches the campaign in the room, including the people executing rather than only those who designed it.
Open with the framing, stated flatly: it is three months from now, the campaign delivered a fraction of its target, and leadership wants to know what happened. Then five minutes of silent individual writing before anyone speaks. Without the silent round the most senior voice anchors the discussion and you learn nothing.
Collect every reason, group them, and have the room vote on likelihood rather than severity. You are looking for what will probably happen, not what would be worst.
The questions that surface real problems
What did we assume about the audience that turned out to be wrong? What broke operationally that we treated as someone else problem? What did a competitor do that we did not plan for? What did we not measure, so we could not tell it was failing until too late?
That last one produces the most useful output in most sessions, because it exposes campaigns with no early warning signal.
Turning it into decisions
Take the top three by likelihood. For each, decide one of three things: prevent it, detect it early with a specific metric and threshold, or accept it and say so in writing.
The written acceptance matters. Three months later when it happens, the difference between a team that predicted and accepted a risk and one that was blindsided is substantial, both for the retro and for how leadership reads the outcome.
Why teams skip it
It feels negative right when everyone is committed to the plan. That is also exactly why it works. Enthusiasm is highest at the point where objections are cheapest to act on, and that combination is what buries campaigns.