Section 1
Most post-mortems blame the channel
Channel blame is attractive because it is nobody's fault and it suggests an inexpensive remedy: run the same material somewhere else. That remedy is why the second attempt so often fails in the same way. A more honest review starts from what the campaign asked the audience to believe and do. Did they believe the problem was theirs? Did they believe your evidence? Did they know what to do next, and was that step small enough to take? A failure at any of those points looks identical in the dashboard, and each requires a different repair. Getting to that level of specificity requires someone in the room who did not approve the original plan. Without that, the review produces a comfortable narrative and a repeat. Upstream errors are catalogued in [Common Storytelling Mistakes Entrepreneurs Make](/blog/common-storytelling-mistakes-entrepreneurs-make).
Section 2
Four ways a campaign fails, and only one is creative
Wrong audience. The story was competent and it reached people who do not have the problem. Nothing about the material can fix this. Right audience, wrong moment. They have the problem but not yet the pressure. The campaign is early rather than bad, and the correct response is to change the ask, not the copy. Right moment, no proof. They believe the problem and doubt you. This is the one that feels like a creative failure and is actually an evidence failure. Everything right, no distribution. The material was good and roughly nobody saw it. Common with high-effort assets that get a launch and no sustained placement. Patterns from the cases that did travel are in [How Storytelling Sparked Viral Marketing Campaigns](/blog/how-storytelling-sparked-viral-marketing-campaigns).
Section 3
A failure taxonomy
Once you can classify a failure, the remedy stops being a guess. The table below maps each failure type to the symptom in the data, the diagnostic question that confirms it, and the specific change that follows.
Section 4
Running a post-mortem people will tell the truth in
Set the rule before anyone speaks: the campaign is being examined, not the people. Then work from artefacts rather than memory. The original brief, the first draft, the approved version, and the numbers. Ask one question that reliably surfaces the real story. What did we assume that we did not test? Every failed campaign has one assumption nobody challenged, usually about the audience's level of awareness, and it is almost always identifiable in the brief. Write the conclusion as a change to a process rather than a lesson. We will validate the awareness assumption with five customer conversations before writing is a process. We will be more customer-focused is not, and it will be forgotten within a fortnight. Distinguishing the underlying mechanisms is covered in [AI vs Automation vs Machine Learning: The Business Difference](/blog/ai-vs-automation-vs-machine-learning-the-business-difference).
Section 5
What to stop doing
Stop reviving campaigns because of the sunk production cost. The money is spent regardless, and the second run costs attention you cannot recover. Stop reviewing only the failures. A campaign that worked for reasons nobody understands is a future failure with a delay on it. Stop measuring the wrong stage. Awareness campaigns judged on conversion get killed early; conversion campaigns judged on reach survive far longer than they deserve. And stop letting the person who commissioned the work run its review. Not because of dishonesty, but because they already know why it should have worked, and that belief is exactly what needs testing.
Section 6
What the evidence says about failure
Failed campaigns tend to break in the places the evidence would predict. Rational-only messaging underperforms: across more than 1,000 cases in the IPA Databank, campaigns built on emotional storytelling were nearly twice as likely to report very large profit gains as rational ones, 31 percent versus 16 percent (Binet and Field, IPA, 2013). Material that lists features without establishing a felt problem is arguing against that pattern from the first slide. Unanchored claims also do not stay in memory. In the well-known Stanford exercise, only 5 percent of listeners recalled an individual statistic minutes later, while 63 percent recalled the stories (Heath and Heath, 2007), which is a hard constraint on any campaign whose central asset is a number. Scepticism compounds the problem, so overclaiming is punished more than it once was. The share of consumers who trust online reviews as much as personal recommendations has fallen from a peak of 84 percent in 2016 to roughly 42 percent (BrightLocal, 2025), and Edelman finds trust now weighs as heavily as price and quality in brand purchase decisions (Edelman, 2025). A story that provokes doubt does not just fail on its own terms, it raises the cost of every claim that follows it. Usefulness, meanwhile, is the strongest predictor of commercial outcome: buyers who found supplier information genuinely helpful were 2.8 times more likely to experience purchase ease (Gartner, 2019). Most failed campaigns optimised for attention at a moment when the buyer needed clarity.