Approach
Qualitative phase
The following actions were taken related to the forecasting process and -model:
-
Internal interviews with Proximus employees, e.g. forecast team employees, employees of impacted departments such as Capacity Management and Scheduling,... to gain insight into the current process and model and its main strengths and challenges
-
External interviews with employees of the external Contact Centers to know their feedback on the delivered forecasts, e.g. timing, accuracy, positive elements, improvement opportunities,...
-
External interviews with employees of other (telecom) companies, which were planned based on the network of both Proximus and Möbius. This provided an interesting basis for comparison and offered insights and ideas towards the future.
-
Internal brainstorming with Proximus employees to identify possible 'events' with an impact on call volumes, as well as the extent to which these are predictable in advance, e.g. weather conditions, commercial actions, roll-out of new devices, etc.
Statistical forecasting model
In a first phase, Möbius built a predictive statistical model from scratch, based on the received historical actual volumes. The obtained Weighted Mean Average Percentage Error (WMAPE) was calculated on a daily, weekly, monthly as well as annual basis and for the different workstreams as defined by Proximus, e.g. technical versus commercial, Dutch versus French, residential customers versus medium-sized enterprises, etc. Next, several iterations were used to test how the model could be optimised, e.g:
-
Calculating the correlation between each of the workstreams to know the grouping potential (cluster analysis)
-
Checking the impact of certain events on the achieved accuracy, e.g. outliers, Covid, deployment fibre, etc.
-
Detailed analysis for those events with impact, e.g. billing (receipt method, payment method, difference in amount versus previous invoice, etc.)
Results
The project led to the following results:
-
A better understanding of the role of 'Forecasting' in the wider Workforce Management Process
-
Improved cooperation between the various departments involved
-
A better understanding of which accuracy percentages are realistic
-
Approximately 20 concrete improvement ideas to work on
-
Clarity about which events need focus in the future
