principle · not checked is the only worthless outcome
Not checked is the only worthless outcome
Right is comfortable. Wrong is uncomfortable. Both are valuable. Only the cycle that never ran the check is worthless.
Predictions land four ways. One has no value.
| Outcome | What it means | What it produces |
|---|---|---|
| Right | The prediction was accurate. | The model is updated with confidence — this is a thing we now know how to predict. |
| Too conservative | The change was bigger than predicted. | The model is updated with calibration — we underestimate this kind of change. |
| Wrong | The prediction did not come true. | The most valuable update — the model has a specific gap. The next brief inherits the lesson. |
| Not checked | Nobody ran the check. The date passed. | Nothing. The model cannot update. The cycle ran blind. |
The asymmetry is the principle. Three of the four outcomes are gold. Only one is worthless — and worthless is generous. Not checked is the silent corruption of the chain. The cycle that never ran the check has learned nothing, but the team often feels like it learned something. That feeling is the danger.
Why teams skip the check
Three reasons. All real. None acceptable.
The check is uncomfortable. A wrong prediction is uncomfortable to read. The team's instinct is to defer, to say we'll check next month, to find a reason the date was wrong. The defer-once becomes the defer-forever. The model never updates.
The check has no instrumentation. The brief named a method that requires a metric the team didn't build. By the time the date arrives, the team realises they cannot run the check. The fix is in Before We Build — instrument the check method as a story in the cycle, not as an afterthought.
The check has no owner. The brief said PO will check. No calendar entry. No commitment. The date arrives and the PO is in a different conversation. The check doesn't run. With a named owner, the check happens on time about 95% of cycles. With PO in the abstract, about 60%.
The discipline
A short rule:
A claim without a check date is a guess. A check date without a calendar commitment is a wish. A check that didn't run is the only outcome with no value.
Schedule the check at brief-time, in the named owner's calendar. Hold the date even when the team is busy. Run the check even when you suspect it will be uncomfortable. Write the signal reading the same week. The model updates because someone arranged for it to.
What changes when the principle holds
Across enough cycles, two things compound. First, the team's calibration improves — the predictions get closer to reality, because each wrong prediction surfaced a gap the next brief closed. Second, the team's honesty with itself stabilises — wrong predictions stop feeling personal because they are routinely the most valuable result.
A team that runs the check every cycle is a team whose third year is meaningfully different from its first. A team that skips the check is one running its first cycle's model in its tenth year of operation.
See also
- Canon — After We Build · Signal and the Prediction
- Canon — After We Build · The Model Update
- Principle — Predictions over plans
- Practice — Writing predictions