BPO being a service industry is always grappling with deploying right quantum of FTEs on the production floor. Efficient management of FTEs on the floor on a daily basis is a formidable challenge to managements in any BPO Process. While a surplus workforce would put pressure on profit margins, a deficient workforce would lead to long queues, longer resolution time and ultimately dissatisfied customers which could hamper its business. Efficient deployment of FTE could be only possible when reliable prediction of the Cases/calls volumes on a daily basis is made. This forecast would provide granular visibility to management on the requisite FTEs on a daily basis and facilitates them on taking view on intra week deployment of FTEs.
The forecast of daily quantum of Calls/Cases is a serious challenge to the Call Centre’s as: firstly, there is great deal of volatility as usually these are sourced globally for international clients; secondly, Call Centre’s do not have any control on the quantum of Calls/ Cases as they are exogenously determined; thirdly, there are seasonality trends which have to be captured; and lastly, forecast should be accurate (<=10 % Error Margin).
Understanding the Model Design:
With the available historic time series data on the inflow Calls /Cases, semi-log regression prototype model is structured and cross validated for few of weeks with the actual inflow data. The Model is also adjusted for holidays, yearend slums, and weekends as surge follow after such breaks. This fine tuning of the model matures it to give mandated forecast accuracy.
How the Model meets the Challenge:
The granular Visibility possible on the cases/calls inflows, when leveraged with the daily closure or productivity trends gives forecast for the requisite FTEs on a daily basis to the management.
BPO industry in India have been generally relying on the monthly forecast of Cases/Calls. Even couple of companies who have been forecasting the daily volumes of Calls/Cases are not high on accuracy. Our Model forecast have been tested for considerable period and is statistically high on accuracy (<=10% Error Margin). For one of the BPO process our Model has suggested a surplus of FTE of about 25% on week days and 150% deficiency on weekends for one of current CRM process managed across Geos. Insights from the forecast suggested, rescheduling of weekly offs for few FTEs and reallocation of few FTEs to other processes would have significant positive impact on the profit margins. The granular accurate visibility of inflow Cases/Calls to management could make this happen. In similar way, this could be replicated across various BPO processes.
Try deep learning using MATLAB