KPI benchmark
How accurate are AI forecasts?
Across 4 public AI deployments, the median forecast accuracy is 90%. Accuracy of a forecast against the actual outcome.
90%
Median forecast accuracy
4
Public deployments pooled
3 of the 4 values are claims by the organization itself, a regulator or an independent study; 1 come from the AI vendor. One value per organization; "up to" claims are listed but not pooled. Data as of 13 October 2026. Methodology
How to measure it
One minus the mean absolute percentage error over a stated horizon, on a holdout period.
Across the library
Median 90% across 4 deployments, reported range 83% to 96%, plus 1 "up to" value not pooled.
Higher is better. Unit: percent.
Reported values by use case
AI cash flow forecasting for corporate treasury
- Ant Internationalat least 90% · organization claim
- Capital A Berhad (AirAsia Group)90% · organization claim
AI demand forecasting and automated replenishment for retail
- Aeon Retailup to 40% · organization claim
AI for FX exposure forecasting and hedging support
- ASML96% · organization claim
Use cases that should track forecast accuracy
- AI cash flow forecasting for corporate treasury
- AI copilot for hotel revenue management
- AI demand forecasting and automated replenishment for retail
- AI demand forecasting for restaurant labor and prep scheduling
- AI dynamic pricing and markdown optimization for retail
- AI for clinical trial site selection and feasibility
- AI for fibre network route and build planning
- AI for FX exposure forecasting and hedging support
- AI forecasting of wind and solar generation for grid balancing and trading
- AI trade promotion optimization for consumer goods manufacturers
Frequently asked questions
- How accurate are AI forecasts?
- Across 4 public AI deployments, the median forecast accuracy is 90%. Accuracy of a forecast against the actual outcome.
- How do you measure forecast accuracy?
- One minus the mean absolute percentage error over a stated horizon, on a holdout period.
- Where do these forecast accuracy numbers come from?
- From public AI deployments in the Blits.ai AI Use Case Library. Every value is quoted from a named source, each organization counts once, and "up to" claims are listed but left out of the median. Data as of 13 October 2026.