Forecast Accuracy: Definition, MAPE Formula, and Targets

Forecast accuracy is the measure of how close a workforce management forecast came to reality. The forecast might be of calls, chats, tickets, or handle time; the measurement compares the forecast value with the actual value for each period and summarizes the error. Because every staffing decision downstream, from hiring to the interval schedule, is built on the forecast, its accuracy is the first number a WFM team should know about itself.

What Is Forecast Accuracy?

A forecast is a prediction of workload by period. Accuracy is the gap between that prediction and what happened, expressed so that a 10 percent miss on a small interval and a 10 percent miss on a large one count the same. The standard way to do that is a percentage error per period, then an average of the absolute values across periods, which is mean absolute percentage error, or MAPE. Taking absolute values matters: without it, over-forecasting one interval and under-forecasting the next cancel out and the forecast looks better than it was.

The period matters as much as the formula. Call Centre Helper's guide notes that "across the industry, intervals of fifteen minutes are generally seen as the most desirable because they represent the most granular data it is practical to measure." A forecast can be within 2 percent for the day and off by 20 percent in individual half hours, and it is the half hours that the schedule was built on.

How to Calculate Forecast Accuracy

Percentage error for one interval

Error = (Actual − Forecast) ÷ Actual × 100

Mean absolute percentage error (MAPE)

MAPE = (Sum of |Actual − Forecast| ÷ Actual across all intervals) ÷ Number of intervals × 100

Weighted absolute percentage error (WAPE)

WAPE = Sum of |Actual − Forecast| ÷ Sum of Actual × 100

MAPE treats every interval equally, so a 50 percent miss on a 4:00 a.m. interval with 6 calls weighs as much as a 5 percent miss on a 10:00 a.m. interval with 400 calls. WAPE weights each interval by its volume, which usually reflects the staffing cost of the error better. Most teams report MAPE because it is the convention, and look at WAPE when overnight intervals are distorting the picture.

Worked example: one morning at a 50-agent inbound team

Interval	Forecast	Actual	Absolute error	Percentage error
8:00	120	131	11	8.4 percent
8:30	160	152	8	5.3 percent
9:00	210	238	28	11.8 percent
9:30	230	225	5	2.2 percent
10:00	240	262	22	8.4 percent
10:30	235	229	6	2.6 percent
Total	1,195	1,237	80

• Daily-level error: (1,237 − 1,195) ÷ 1,237 = 3.4 percent under-forecast. Looks acceptable.
• MAPE: (8.4 + 5.3 + 11.8 + 2.2 + 8.4 + 2.6) ÷ 6 = 6.45 percent.
• WAPE: 80 ÷ 1,237 = 6.5 percent.

The daily figure hides the fact that the 9:00 and 10:00 intervals were short by 28 and 22 calls, which at a 6.5-minute handle time is roughly 3 agents' worth of work in each half hour. That is the interval the service level was missed in, and the daily accuracy number would never have shown it. Chris Dealy makes the same point in Call Centre Helper's methods guide: "if you're just looking at daily or weekly totals your forecast will probably look more accurate than it is."

Forecast Accuracy Targets

The same guide gives the most widely cited targets: "if you've got a centre with at least 100 agents in, you should be aiming for within 5%, so a Mean Absolute Percentage Error of 5% or less," and for smaller centers, "because you haven't got the safety in numbers, then 10% is a reasonable target." The size dependence is statistical: small volumes are noisier, so the same forecasting skill produces a higher MAPE in a 30-agent center than in a 300-agent one.

Targets should also depend on the horizon. A forecast made six months out for a hiring plan will be less accurate than one made two weeks out for a schedule, and the two should be measured and targeted separately. Many teams track three: long-term (monthly volume, for capacity), short-term (daily volume, for scheduling), and intraday (interval volume, for real-time management).

Why Forecast Accuracy Matters

Forecast error becomes staffing error directly. Under-forecasting an interval by 10 percent means roughly 10 percent too few agents scheduled, which pushes occupancy up and service level down for that interval. Over-forecasting means agents idle and money spent on hours that were not needed. Because schedules are built weeks ahead, the error cannot be corrected on the day except with overtime, voluntary time off, or moving breaks, all of which have costs.

There is a second-order effect on the team. Persistent under-forecasting means persistent overload, which shows up in attrition and in schedule adherence, because agents stop respecting a schedule that is always wrong.

How to Improve Forecast Accuracy

• Measure at the interval, report by horizon. Keep a running MAPE for intraday, short-term, and long-term forecasts. Improvement work goes where the error is.
• Separate volume error from handle-time error. A workload forecast can be wrong because the calls were wrong or because the calls took longer. Track AHT accuracy alongside volume accuracy.
• Remove known causes before tuning the model. Marketing sends, outages, billing runs, and holidays explain most large misses. A calendar of drivers, kept by the forecaster, is worth more than a better algorithm.
• Reforecast frequently. A forecast made four weeks ago should be refreshed with the last two weeks of actuals before the schedule is published.
• Track bias. If the signed errors are consistently negative, the forecast is systematically low, and a simple uplift fixes it before the model does.

How to Misuse Forecast Accuracy

The most common misuse is measuring at the daily level and reporting a flattering number. The second is measuring only volume and ignoring handle time, which can swing workload by 10 percent while volume is exactly on forecast. The third is holding the forecaster accountable for accuracy while the business withholds the drivers, such as a campaign schedule, that would have made the forecast right. Accuracy is a property of the process, not the forecaster.

Forecast Accuracy in Contact Centers and Remote Teams

A 50-agent remote support team reported 96 percent daily forecast accuracy for a year while missing its service level target in about a third of intervals. When the WFM analyst recalculated at the half-hour level, MAPE was 14 percent, with the error concentrated in the first two hours of the day, where a Monday-morning pattern from an email campaign had never been built into the model. Adding the campaign calendar as a driver brought interval MAPE under 8 percent in a quarter, and the service level misses moved from a third of intervals to about one in ten, with no change in headcount.

Distributed teams add one twist: the actuals used to measure accuracy include only the work that was logged. If agents handle contacts outside the system, or the system's interval data is incomplete, the forecast is being measured against a partial record.

How to Track the Staffing Side

HiveDesk does not forecast contact volume or measure forecast accuracy; those come from the contact platform and the WFM model. What HiveDesk provides is the other half of the comparison: scheduling shows the staffed hours the forecast turned into, and clock-in, clock-out, and break records show the hours actually worked in each interval, so a service level miss can be attributed to forecast error, shrinkage, or adherence rather than guessed at. The plan is $5/user/month with a 14-day free trial and no credit card required. The workforce management guide covers how the three measurements fit together.

Related Terms

• Capacity planning: where the forecast becomes a headcount
• Shrinkage: the other input that determines whether staffing matches demand
• Service level: the outcome forecast error shows up in
• Occupancy rate: what rises when the forecast is low
• Average handle time: the second forecast that needs its own accuracy measure
• Workforce management: the discipline forecasting belongs to

Frequently Asked Questions

How is forecast accuracy calculated?

Most commonly as mean absolute percentage error: for each interval, take the absolute difference between actual and forecast divided by actual, then average across intervals and multiply by 100. Weighted absolute percentage error divides the total absolute error by the total actual instead.

What is a good forecast accuracy for a call center?

A MAPE of 5 percent or less is the commonly cited target for centers with 100 or more agents, and 10 percent for smaller centers, measured at the interval level. Longer-horizon forecasts have looser targets.

Why measure forecast accuracy at the interval level?

Because daily totals hide compensating errors. A forecast can be within 3 percent for the day while individual half hours are off by 10 to 20 percent, and staffing is built at the half hour.

What is the difference between MAPE and WAPE?

MAPE averages the percentage error of each interval equally, so low-volume intervals count as much as busy ones. WAPE weights by volume, dividing total absolute error by total actual, which better reflects the staffing cost of the error.

Should forecast accuracy include handle time?

Yes. Workload is volume times handle time, and an accurate volume forecast with a 10 percent handle-time miss produces a 10 percent staffing error. Track both.

Who is responsible for forecast accuracy?

The forecaster owns the model, but accuracy depends on the business sharing its drivers: campaigns, releases, billing cycles, and outages. Measuring the forecaster alone misattributes most large misses.

Browse more workforce management terms in the glossary.