It is often thought that Demand forecasting is an activity that can be undertaken separately from capacity planning and the influence this will have on customer queueing. This, however, is not realistic and how long customers have to queue, and their patience to queue might see customer abandon the queue and re-connect at a future point.
True inbound volume (we refer to it as the fresh volume from now on) is more appropriate to be used when one makes forecasts, since it is independent of the service levels in the contact centre. In contrast, the total inbound volumes are influenced by the service levels, skill level of the agent, and staffing decisions of the contact centre, due to the redial and reconnect customer behaviours.
On a much simpler basis it does not even require customers to abandon in order to influence future volume arrival. This is especially true in high touch (when the same customers need to regularly make contact) customer contact centres. For example, in a high touch contact centre if the contact centre is difficult to reach in mornings it is likely customers will shift contact arrival towards the afternoon/evenings. These types of gradual changes are easy to capture and normally the regular forecasting process will pick them up.
In a contact centre a significant fraction of the inbound contact volume involves customers reconnecting more than once. There are several reasons for customers to need to reconnect, including abandoned customers not getting their questions answered in their initial attempts, a customer checking what the status of their previous request or perhaps the quality of the agent who last handled their request not resulting in resolution.
It is therefore important to separate what is “true demand” (the number of unique customers contacts), what is reconnect as a result first contact resolution failure, and what is a reconnect as a result of a previous abandoned attempt. True demand data is best to use for trend and seasonality factoring whilst the reconnects probabilities are expected to be more stable over time, since they represent normal customer behaviour.
We conclude that the first call resolution and service levels are likely to be having a considerable impact on the number of contacts received the accuracy of the forecast derived and it is for this reason it might be interesting to see which part of the variability in the deseasonalized actuals can be explained by the service level or first contact resolution.
