Forecasting is a critical component of a call center’s workforce management program. The accuracy of your forecast has a ripple effect on almost all performance metrics and business outcomes. A bad forecast can directly lead to under- or overstaffing, which then has cascade effects on average handle time, CSAT, labor waste… the list goes on and on! Forecasting has often been described as part science and part art, and there are many factors that impact forecast accuracy. However, one critical component is the accuracy and reliability of the historical contact volume data from which your forecast is generated.
Why does historical data matter?
You know the expression – garbage in, garbage out. An athlete can’t eat fast food every day and then expect to place first in the big race. Similarly, you can’t feed your forecast incomplete, inaccurate data and expect to get a solid, accurate forecast. The historical dataset used has a huge impact on forecasting. Forecast accuracy is dependent on the historical data fed into the software. The best WFM technology and forecasting algorithms in the market won’t help if you’re historical data set is not appropriate, complete, or accurate! The more historical data you have available, the more accurate the forecast can become. For example, if you want to see annual trending, you need least 2 years of data. On the other hand, if you want an identifiable pattern for interval forecasting you might not need 2 years, but providing 2 years – or as much as you can! — will allow for stronger confidence as you have additional data points to validate the forecast.
What if I don’t have historical data?
The reality is that sometimes you just don’t have great historical data, especially in the case of young contact centers, or those that are just newly implementing contact routing or workforce management software. There is no doubt forecasting with limited or incomplete data is challenging – but it is still doable and valuable. Although 2 years data is ideal, even a couple months of data is better than no data at all!
Although it doesn’t happen too often, there are instances where there is absolutely no historical data to use, like when a contact center implements both ACD and WFM for the first time simultaneously. In this case, the efforts are tied around developing a “contact rate” using a different metric, such as revenue forecast, number of items sold, expected number of visitor to a site, pieces of mailings be sent etc. For example, if in the past you sold 500 nick-nacks, and you received 100 calls, you could use this assumption in future months. If the business expects nick-nack sales to increase throughout the year, you can assume to same for calls.
Typically this process is used for a minimum of twelve (12) months, to ensure you are adequately capturing trends based on seasonality. While assumptions based on other metrics aren’t perfect, they are far better than nothing at all! You can’t expect your workforce management software to make something (a great forecast) from nothing (bad historical data). So feed your forecast well!