Directional accuracy is the share of forecasts whose predicted direction — up or down over a defined horizon — matched what the price actually did. It’s a simpler, more procurement-relevant metric than absolute forecast error (MAE, RMSE), because procurement decisions are usually directional: should we fix now or wait? Should we lengthen or shorten the horizon? A forecast that is precisely wrong on direction is worse for a buyer than a forecast that’s directionally right but numerically off.
Directional accuracy is measured backward, on the realised price path, over a rolling window. INAYA reports it by commodity, by horizon (1, 3, 6, 12, 24 months), and by regime (trending, range-bound, volatile). The composite figure exceeds 90% on the 1–6 month horizon for liquid contracts; longer horizons and less liquid contracts degrade gracefully and the degradation itself is visible to the user, so a buyer never confuses confidence with reality.
The number matters because it converts directly into expected procurement value. A 90% directional read on a forward six-month curve means that nine times out of ten, a fixing committee acting on the central forecast is moving in the right direction — which is the bar for replacing a spreadsheet-driven process.
Related concepts: ensemble forecasting (how the forecast is built), commodity intelligence (what the forecast becomes when exposure is mapped to it).