The Austrian Automobile, Motorcycle and Touring Club ÖAMTC sees itself as a point of contact and promoter of the interests of its members in all questions relating to mobility. Under this premise, the Yellow Angels on the street and in the call centre focus their daily activities on the service concept. As part of the telephone and information service, call centre employees are the first point of contact for members and customers.
The expected call volume plays a major role in optimal personnel deployment planning in the call centre. The wrong number of call centre employees can lead to long customer waiting times during peak periods or waste staff time during quiet periods. However, the very complex Excel-based model previously used for personnel deployment planning did not allow for dynamic changes and reached its limits due to the multi-dimensionality of the process. The algorithms necessary for flexible adaptation for processing data, for example on weather development and seasonal holiday behaviour, could not be used.
The ÖAMTC saw the solution to the problem in the use of machine learning (ML). The core functions of this technology include recognising patterns of behaviour among callers and comparing this data with a variety of different parameters. As experienced digitalisation experts, the Mobility Club commissioned A1 Digital to develop a proposal for the innovative use of ML technology for personnel planning.
Based on historical data on calls (including length of calls, waiting times, etc.) A1 Digital iteratively developed a machine learning model to predict call volume. In the first step, a use case was specified in joint workshops. A1 Digital then used its own self-parameterising ML platform to test different algorithms as well as different models (e.g. a neural network and the Random Forest classification method) and their configurations. The result was a solid model that was able to forecast the probable call volume and thus the personnel requirements over certain periods based on various input parameters such as the number of calls in the last week, the weather in the previous year's period and the holiday periods. With an extensive proof-of-concept, A1 Digital then demonstrated that the developed ML model can actually accurately predict the number of calls in the desired time horizon.
The range of predictions ranges from very accurate intraday forecasting, which reveals how many calls will be made in the next three hours, to monthly forecasts. With the knowledge that, for example, a public holiday in March leads to a highly probable number of calls in 15 min increments, the ÖAMTC can now plan and coordinate the deployment of personnel much more precisely over a longer time horizon. The ML model developed by A1 Digital enables the mobility club not only to predict the number of calls, but also the content of the calls in a later step, in order to further optimise subsequent actions, e.g. the towing service, with the input.
“The proof-of-concept has shown us that the use of machine learning can change our work in a sustainable and positive way. The way A1 Digital's AI experts have condensed the multitude of data into a reliable, expandable ML solution convinced us to make the right decision when we got A1 Digital on board for this ML project.”