Skip to main content

What is a good benchmark for forecast accuracy in contact centers?

This is a question we get from our clients quite often. According to Calabrio a 5% error is typically the industry standard. This corresponds to the COPC high performance benchmark of 95% forecast accuracy on the forecast that is used for scheduling. However, Injixo writes in its 2019 contact center benchmark report that 43% of respondents reported a lower accuracy or don’t measure accuracy at all, and that is a self-reported number. Also, at least 10% of respondents is lying, indicating that error is less than 2% on average.

So what is the actual ‘Industry Standard’? And are these kind of benchmarks of any use? Well… yes and no. Industry benchmarks are useful to e.g. software creators, to estimate how much potential there is for improvement across an industry. But when it comes to measuring the success of the forecasting practice within your team, that’s a different story!

Volume matters!

First of all: volume matters! Customers will contact you at random moments which translates to a poisson-distribution. This means the bigger the number of customers calling, the more stable the distribution becomes and therefore easier to predict. This also means you’re forecasting with a handicap when you are dealing with lower volumes.

Different behavior

Secondly: Different types of customers behave differently. Customers of energy companies tend to get in touch more on cold days, especially the first cold days of the winter season. As the weather is notoriously hard to predict that means a lower benchmark can be justified.

Also: how far into the future do you need to forecast? Some contact centers generate schedules 2 weeks ahead, while others schedule up to 6 weeks ahead. For the latter it is much harder to achieve an accurate (scheduling) forecast.

So if you’re looking for a benchmark to evaluate your forecasting performance, make sure to factor in these variables. The good news is that a maximum forecast accuracy can be calculated for every unique set of historical data.


If you’d like our help in setting your internal benchmarks, get in touch!