Demand Forecasting Diagnostic Tool

Free Demand Forecast Accuracy Audit — MAPE, Bias & Model Selection

Upload sales history or enter it manually, and Mathnal's forecasting diagnostic detects your demand pattern, recommends the right model, and computes MAPE, wMAPE, MAE, MASE, bias and tracking signal — instantly and free.

What is forecast accuracy and how is it measured?

Forecast accuracy measures how close your demand forecast is to actual sales. The most-used metric is MAPE (Mean Absolute Percentage Error) — below 10% is excellent, below 20% is good, above 30% needs intervention. But MAPE alone hides systematic error, so this tool also reports bias (chronic over- or under-forecasting) and the tracking signal that flags when a model has drifted.

How do you choose a forecasting model?

The right model depends on your demand pattern. Stable demand suits moving averages and exponential smoothing; trended or seasonal demand suits ARIMA or ETS; intermittent demand suits Croston's method; complex patterns may suit Prophet or XGBoost. This tool auto-detects trend, seasonality, intermittence and level shifts, then recommends an appropriate model so you don't guess.

Why does forecast bias matter more than error?

A forecast can have low average error yet still be consistently wrong in one direction. Positive bias (under-forecasting) drives stockouts; negative bias (over-forecasting) drives excess inventory and write-offs. Because bias compounds period after period, detecting and correcting it usually delivers more value than chasing a slightly lower MAPE.

Who is this forecast audit for?

Demand planners, S&OP teams, category managers and analysts who want a quick, objective read on forecast quality and a defensible model recommendation — without setting up Python or paying for forecasting software.