Human Resource Allocation in a Call Centre

Abstract

A novel approach to the problem of resource allocation in telephone call demand satisfaction has been proposed by Prof David Lowe and Dr Ian Nabney in a feasibility study. The method, inspired by neural network approaches to combinatorial optimisation, has the advantage of allowing a flexible use of a multi-skilled workforce, proposing about25% fewer agents compared to figures produced by a traditional statistical method based on agents with single skills and separate queues.

The purpose of this thesis is to develop and investigate some of these assumptions. One of these assumptions is that there is an unlimited pool of people in each skills mix. As this may not be appropriate for more highly qualified profiles, a method to cap the number of agents is described and its effects are investigated. It has been implemented and added to the earlier algorithm.

To investigate the distribution of the calls across time, we developed a stochastic model of the queues based on the Erlang C model currently used in scheduling software. This model is more realistic than the assumption, made in the feasibility study, that the calls are uniformly distributed. In our new model, the arrivals are assumed to follow a Poisson process and the service time distribution is described by an hyper-exponential distribution, which takes into account the multi-skilled ability of each operator. This leads us to increase the number of agents required to achieve a proper service level, but nevertheless our approach still proposes 20% fewer agents compared to existing methods.

The reliability of our method is tested with different data sets. In addition, some simulations were carried out to validate our new approach. The results, which are quite encouraging, are presented in detail.

Another aspect of our approach was partially investigated. Instead of grouping people together in small numbers of groups with similar skills profiles, we grade each individual member of the workforce with his own unique skills profile. Consequently the incoming calls will be allocated to each individual and not to groups of agents. Some results and conclusions about this new issue are detailed.

Date of Award	1997
Original language	English
Awarding Institution	Aston University