Skip to main content
TurnellaBeta

Forecast accuracy calculator

Paste your actual and forecast contact volumes to compute WAPE, MAPE, and bias in one click. Supports any interval length — 15-minute, 30-minute, hourly, daily, or weekly.

Format: actual, forecast — one row = one interval (15 min, 30 min, hourly, daily, or weekly).

WAPE

5.8%

Good

MAPE

5.5%

mean absolute %

Bias

-0.5%

balanced

Intervals

10

data points

How to interpret your WAPE of 5.8%

< 5%
Excellent
5–10%
Good
10–15%
Acceptable
15–20%
Poor
> 20%
Very poor

Interval breakdown

#ActualForecastAbs error% errorDirection
112011554.2%↓ under
21451601510.3%↑ over
39810022.0%↑ over
4210195157.1%↓ under
517518052.9%↑ over
6130140107.7%↑ over
7889022.3%↑ over
8220205156.8%↓ under
916517053.0%↑ over
10140128128.6%↓ under
Total1,4911,48386WAPE 5.8%

WAPE, MAPE, and bias — the three forecast accuracy metrics

WAPE

Weighted Absolute Percentage Error

Σ|A − F| / ΣA × 100

The primary WFM accuracy metric. Weights high-volume intervals more — quiet overnight periods don't distort the score. Use WAPE as your headline number.

MAPE

Mean Absolute Percentage Error

(1/n) × Σ(|A − F| / A) × 100

Simple average of interval-level % errors. Sensitive to low-volume intervals where a small absolute error creates a huge percentage error. Use MAPE as a secondary check.

Bias

Mean Forecast Error (direction)

Σ(F − A) / ΣA × 100

Positive = over-forecast (overstaffing risk). Negative = under-forecast (understaffing risk). A small bias is normal; above ±5% signals a structural problem.

Why WAPE beats MAPE for WFM: A 03:00–03:30 interval with 2 actual contacts and 3 forecast contacts has a 50% MAPE contribution — but an absolute error of only 1 contact with zero staffing impact. WAPE weights this tiny interval at only 2/total-volume, so it barely affects the score. Use WAPE as your primary accuracy measure.

WAPE benchmarks by forecast level

Accuracy benchmarks vary by interval length — daily forecasts are more accurate than 30-minute interval forecasts on the same data, simply because averaging smooths out noise.

LevelExcellentGoodPoor
Daily volume< 3%3–8%> 15%
30-minute interval< 7%7–15%> 25%
15-minute interval< 10%10–20%> 30%
Weekly volume< 2%2–5%> 10%

Benchmarks are for stable demand patterns. Operations with high volatility (utilities, insurance) will typically have higher WAPEs — compare against your own historical baseline, not just industry norms.

How to improve forecast accuracy

Separate signal from noise

If WAPE is poor on individual intervals but good at the day level, the issue is intraday distribution — the day total is accurate but the pattern is wrong. Check for ACD data integrity issues (login events creating artificial peaks) and verify your intraday template is up to date.

Chase the bias first

A persistent positive bias (over-forecasting) often comes from a baseline that was set during a high-volume period. A persistent negative bias suggests your seasonal index is not capturing a trend. Fix the structural bias before optimising interval-level accuracy — a biased forecast is wrong in the same direction every day.

Account for special events

A single missed special event (bank holiday, product launch, media event) can drag your monthly WAPE from 8% to 15%. Maintain a forward event calendar and apply volume multipliers for each event type. Post-event, recalibrate your multipliers from the actual data.

More history is not always better

Using 3 years of history to forecast a business that changed its channel mix 18 months ago will produce a biased forecast. Weight recent history more heavily, or use a shorter window that matches the current operating model.

Forecast accuracy questions

What is WAPE and how is it calculated?

WAPE stands for Weighted Absolute Percentage Error: WAPE = Σ|Actual − Forecast| / ΣActual × 100. Unlike MAPE, WAPE weights each interval by its actual volume, so high-volume intervals have a larger impact on the score. This prevents quiet overnight intervals with large percentage errors from distorting the overall accuracy measure. WAPE is the standard accuracy metric in workforce management.

What is a good WAPE for a contact centre forecast?

WAPE benchmarks: under 5% is excellent; 5–10% is good (typical for well-run WFM operations); 10–15% is acceptable; 15–20% is poor; above 20% is very poor. These benchmarks apply to 30-minute interval forecasts. Daily-level WAPE is typically 3–5 percentage points better than interval-level WAPE for the same operation.

What is the difference between WAPE and MAPE?

WAPE weights each interval by its actual volume. MAPE is a simple average of percentage errors. For WFM purposes, WAPE is almost always preferred because quiet intervals can have extreme percentage errors (e.g. 50% error on 2 actual contacts) that inflate MAPE dramatically but have no operational impact. WAPE correctly down-weights these low-volume intervals.

What does forecast bias mean in WFM?

Bias = Σ(Forecast − Actual) / ΣActual × 100. Positive bias means you consistently over-forecast (overstaffing risk). Negative bias means you consistently under-forecast (understaffing risk). A small bias under ±3% is normal. A persistent bias above ±5% signals a structural problem — likely an incorrect baseline, missed trend, or mis-calibrated seasonal index.

Related calculators and guides

Track forecast accuracy over time

The free calculator scores a single forecast period. Turnella tracks WAPE and bias across every forecast period and connects accuracy outcomes to staffing requirements automatically.

Open the full app →