Call centres and service centres are an essential resource of today’s businesses, from emergency response to financial services. With more than 70% of all customer and business interactions handled through call centres, they often represent the primary channel to provide information and assistance to customers.
Across industries and government, more than 3.5 million were employed in US call centres − more than 2.6% of the US workforce in 2006, making it an important and continuously growing sector of society. With human resources as its major cost driver, a major challenge to call centres is that of forecasting call centre demand of inbound calls, call durations and call types in order to schedule agents efficiently across different services within a week and during the day (operational planning), or to determine the quantity and timing of hiring and training (tactical planning), in order to balance service quality, costs and efficiency.
Although advances in information and communication technology (ICT) have helped identify the true historic call volumes, which are often masked by redials and reconnects, the challenge of forecasting inbound calls has increased with the increasingly competitive business environments. As a result, forecasting call centre demand in half-hours or even 15-minute intervals is often particularly challenging, as it is driven by multiple overlapping intraday, weekly and yearly seasonalities, combined with multiple exogenous drivers such as advertising, bills & letters sent out, and different bank holidays with lead and lag effects, and different responses by call types and across socio-demographics.
The Lancaster Centre for Forecasting has significant experience in meeting the challenges of forecasting call centre data which current commerical software do not address. We deliver tailored solutions employing the latest statistical methods, from triple Exponential Smoothing to ARIMAX and Regresison models to state-of-the-art nonlinear methods of Artificial Intelligence such as Neural Networks or Support Vector Regression.