Conducting the Decision Autopsy: A Framework for Learning From Failed Bets
Photo: Ash Carter, Public domain, via Wikimedia Commons
Every quarter, American companies quietly shelve failed initiatives — the product line that missed projections, the market entry that stalled, the platform migration that cost triple its estimate. The financial loss gets recorded. The reasoning that produced the loss rarely gets examined.
This is the pattern that separates organizations that compound their judgment from those that merely compound their expenses. When a bet goes sideways, most leadership teams perform a ritual: a tense retrospective, a few process tweaks, a slide about lessons learned, and then silence. Within eighteen months, the same decision-making patterns produce the same failure with different scenery.
The alternative is what we call the decision autopsy — a disciplined examination not of what happened, but of what the organization knew, believed, and chose at the moment of commitment.
Hindsight Is the Enemy of Understanding
The first obstacle to any useful post-mortem is hindsight bias. Once outcomes are known, our brains quietly rewrite history: the warning signs seem obvious, the optimism seems foolish, the people who dissented seem prescient. A retrospective conducted this way teaches nothing. It simply assigns narrative blame and moves on.
A proper autopsy works differently. It reconstructs the decision as it looked at the time — with the information available then, the constraints in place then, the incentives operating then. Only by standing in that moment can you determine whether the failure was a bad process, bad luck, or bad execution. These three causes demand entirely different remedies, and confusing them is how companies fix the wrong things.
Step One: Reconstruct the Original Thesis
Begin by writing down, in one paragraph, what the decision was actually premised on. Not the polished board narrative — the operating thesis. "We believed mid-market buyers would pay a 20% premium for this capability because our interviews showed X." Every failed bet rests on a chain of assumptions. Name them explicitly.
If the original thesis cannot be reconstructed, that itself is a finding. Organizations that commit capital without a falsifiable premise are not making bets; they are making wishes.
Step Two: Grade the Information, Not the Outcome
Next, assess what was knowable at decision time. Which assumptions were supported by real evidence? Which were inherited from a competitor's playbook, a consultant's framework, or a senior executive's intuition? Distinguish three tiers: verified, assumed, and hoped.
A sound decision built on verified assumptions that failed anyway is a luck problem — and luck problems should not change your process. A sound-seeming decision built on assumed assumptions is a process problem. Most strategic failures in US companies live in this second tier, and almost none get labeled honestly.
Step Three: Map the Dissent
Find out who saw trouble coming and what happened to their concerns. In failed initiatives, dissent is rarely absent — it is processed. It gets deferred to a follow-up meeting, softened into a footnote, or absorbed by a senior voice saying "let's not overthink this."
Trace the path of every early warning. Where did it enter the system? Where did it die? This map reveals more about your organization than any financial variance report, because it shows you the actual throughput of your decision-making machinery, not the official one.
Step Four: Separate Decision Errors From Execution Errors
A strategy can be sound and fail through poor execution. An execution plan can be flawless and fail because the strategy was wrong. Companies routinely conflate the two, which produces absurd remedies: reorganizing teams because the market thesis was wrong, or abandoning a viable strategy because the rollout was under-resourced.
Draw the line explicitly. Then evaluate each side on its own terms. Execution failures usually trace to capacity, sequencing, or ownership. Decision failures trace to information quality, incentive design, or authority structures. Different autopsies, different prescriptions.
Step Five: Change One Structural Thing
The final step is where most post-mortems collapse into theater. A list of lessons learned is not a lesson. An autopsy concludes with exactly one structural change — not five, not ten — chosen because it addresses the root cause identified above.
Structural means it outlives the individuals involved: a new gate in the capital allocation process, a standing rule that major bets require documented dissent, a requirement that every initiative above a threshold names its kill criteria in advance. If the change could be undone by one leadership transition, it is cultural decoration, not structure.
Why This Compounds
Individual executives have limited careers; institutional decision processes persist. A company that runs honest autopsies builds, over a decade, an increasingly calibrated sense of which kinds of bets it wins and which it loses. Its leaders develop pattern recognition that no outside adviser can sell them.
The companies that dominate their categories are rarely the ones that avoid failure. They are the ones that metabolize it — converting each failed bet into a permanently upgraded decision system. The autopsy is unglamorous, occasionally uncomfortable work. It is also the highest-leverage process most US companies have never institutionalized.
