Trust Center
How accurate the platform's predictions are, how much promised decision value was actually realized, and the verified savings behind it — all auditable.
Prediction calibration scoreboard
No prediction outcomes tracked yet.
Calibration curve
Claimed confidence against verified outcomes. A well-calibrated model tracks the diagonal — above it is underconfident, below it is overconfident. Scored on verified outcomes only (correct = 1, partial = 0.5, incorrect = 0).
Not enough verified outcomes to plot calibration yet.
Accuracy by model version
Whether each released version actually improved on the last, on verified outcomes.
No predictions recorded yet.
Dated failure forecasts
Forecasts whose predicted date has passed must be verified by the Reliability Engineer before they can count toward calibration.
0
Past date, unverified
0
Date still ahead
0
Verified
Prediction ledger
Every forecast, its claimed confidence, the verified outcome and who verified it.
No predictions match this filter.
Decision value realization
No decisions verified against outcomes yet.
Verified avoided-failure value
$0
verified (0 interventions)
$0
projected pipeline (0)
No savings verified yet — verified entries appear here with evidence.
Calibration by accountable role
No verified decisions with both expected and actual value yet. Calibration can only be scored once outcomes are verified.
Realisation = Σ actual ÷ Σ expected across verified decisions. Approver names are as recorded at approval and are not authenticated sign-offs.