Radical transparency · updates live
Our forecasts, graded.
Every tool makes savings promises. Nobody audits their own. Below is the aggregated, anonymized accuracy of every CARTIEAI decision forecast that customers graded against reality — published from day one, whether the number flatters us or not.
No graded forecasts yet — and that's the point.
This page went live before we had numbers to brag about. As customers approve decisions in the Decision Ledger and record real outcomes, the aggregate accuracy appears here automatically — unedited, uncurated.
Methodology — nothing up our sleeve
- Accuracy = 100 − |actual − predicted| ÷ |predicted| × 100, floored at 0, per decision.
- Only approved decisions where the customer recorded a real-world outcome are counted.
- Aggregated across every CARTIEAI workspace. Anonymized by design — no customer names, no decision titles, no per-decision amounts ever leave a workspace.
- Demo-workspace and internal test decisions are excluded — the number is customer-only.
- We cannot edit or exclude individual grades. Bad forecasts count exactly like good ones.