How fast the org catches its own mistakes and reverses them. Mistakes aren't the problem. Mistakes that compound for years are.
ERROR → DETECT → CORRECT
Errors surface fast because there's no cost to surfacing them. Wrong turns get reversed, not buried. Post-mortems lead to actual changes. Being wrong fast is fine.
The org learns to defend what isn't working rather than fix it, because fixing it means admitting it was wrong. The same problems recur. People know about them. Nobody fixes them.
The cost of a mistake isn't the mistake. It's the duration. A bad call caught in a week costs a week. The same call running uncorrected for three years costs three years of compounded dysfunction.
When failure is career-ending, people hide failure. Hidden failures compound. The org develops antibodies against surfacing bad news, which means leadership runs on filtered information. The errors don't get smaller. They just get invisible.
Sunk cost has veto power. Once an org has invested enough in a direction, changing course reads as admitting the original call was wrong. The people who made that call are often the same people who'd have to authorize the reversal. So they don't. Not because they're bad at their jobs. Because the incentive structure rewards defending the decision over correcting it.
Failed experiments get killed in days. Every product decision has pre-defined kill criteria. Being wrong fast is the goal.
Two-pizza teams can reverse decisions without executive permission. Reversibility is a design principle.
Streaming strategy reversed direction multiple times. Each reversal took 12+ months to execute. The org can't correct at market speed.
The Versant spinoff decision took years to execute after the signal was clear. Structural complexity extends every correction cycle.