Leading Indicators for a Decision
Choose leading indicators that appear before the outcome, connect through a credible mechanism, and can trigger a specific decision or correction.
Early is not the same as predictive
Activity metrics are easy to collect but may not lead to the intended outcome. A useful indicator changes before the result, has a stable relationship to it, and supports an action while there is still time.
Design an indicator
- Define the eventual outcome.
- Map the mechanism that produces it.
- Choose a measurable early signal.
- Set a threshold and response in advance.
Example: onboarding risk
Login count is weak, while completion of the first real workflow strongly precedes retention. The team uses workflow completion to trigger help, not as a vanity success metric.
Indicators can be gamed
When a measure becomes a target, people may improve the number without improving the outcome. Pair it with a quality or consequence check.
The decision rule
What to record before acting
Use an early measure only when its mechanism and response are clear enough to change action.
Apply the rule in context with record forecasts and review outcomes. Keep the evidence and assumptions with the decision so an outcome review can distinguish reasoning from luck.
Continue within evidence and uncertainty
Robust vs Optimal Decision
Prefer a robust option when it performs acceptably across plausible futures and the apparently optimal option depends on one fragile forecast.
Robust vs Optimal DecisionTail Risk When Comparing Two Options
Compare tail risk by identifying low-frequency severe outcomes, exposure limits, controls, recoverability, and who bears the loss.
Tail Risk When Comparing Two Options