Forecast accuracy calculator
A forecast that is right in total can be wrong in every interval, and it is the intervals that staff the day. Paste the forecast and the actuals for a day or a week and this calculator gives the volume-weighted error, the interval-level error, which way the forecast leans, and where it missed by most.
Σ|forecast − actual| ÷ Σ actual. Weighted by volume, so the busy intervals count most.
- Actual
- Forecast
- MAPE
- 1.1%
- Bias
- +0.1%
- Intervals within 10%
- 100.0%
- Worst over-forecast
- 08:30, +4
- Worst under-forecast
- 10:30, −4
- Totals
- 3005 vs 3001
Mean of each interval's error; treats every interval equally.
Over-forecast overall.
Under-forecasts cost service level; over-forecasts cost money.
Forecast vs actual.
Show the working
- 24 intervals: forecast total 3005, actual total 3001.
- WAPE = Σ|forecast − actual| ÷ Σ actual = 30 ÷ 3001 = 1.0%.
- MAPE = mean of |forecast − actual| ÷ actual over intervals with actuals = 1.1%.
- Bias = (Σ forecast − Σ actual) ÷ Σ actual = 0.1%.
Accuracy is measured after the fact. Pebble WFM forecasts every interval from your history and tracks accuracy as the actuals arrive. Free month, no card needed.
How forecast accuracy is calculated
WAPE, weighted absolute percentage error, is the sum of the absolute errors divided by the sum of the actuals. A miss of ten contacts in a busy interval and ten in a quiet one count the same, which is right for staffing, because ten contacts is ten contacts of work wherever they land.
MAPE, mean absolute percentage error, averages each interval's percentage error. It treats a 50 per cent miss on a quiet interval as worse than a 5 per cent miss on the peak, which is how it ends up flattering a forecast that is bad at the peak, and why it is reported alongside WAPE rather than instead of it.
Bias is the signed total error. A forecast can have a modest WAPE and a consistent lean; over-forecasting costs money in every interval, under-forecasting costs service level in every interval, and the fix is different. The worst intervals show whether the misses cluster at a particular time of day, which usually points to the intraday shape rather than the total.
Frequently asked questions
- What is a good forecast accuracy?
- At interval level, a WAPE under 10 per cent is good and under 5 per cent is excellent for a stable inbound queue; daily totals should be tighter. Measure at the level you staff at: a daily figure of 3 per cent can hide interval errors of 20.
- Should I measure against the original forecast or the last reforecast?
- Both, for different reasons. The forecast that was used to build the roster is the one that explains service level; the last reforecast is the one that explains intraday decisions. Report the roster forecast as the headline.
- Why is my MAPE much higher than my WAPE?
- Quiet intervals. A forecast of 8 against an actual of 5 is a 60 per cent error on three contacts. MAPE counts it in full; WAPE weights it by the three contacts. If the two diverge, the forecast is fine where it matters and noisy at the edges of the day.
Stop doing this one interval at a time
Pebble WFM forecasts your demand, computes the staffing requirement for every interval, builds the roster and publishes it, with self-service for agents and a copilot that can do what a planner can. Explore a sample organisation on day one.