In my last post, “The Importance of Model Validation in Contact Center Staffing“, I briefly discussed three types of computer/mathematical models that are commonly used in contact center what-if analyses. Let’s expand on these a bit.
Predictive modeling is probably the most common model discussed in contact centers—this is the model that forecasts call volumes, handle times, agent shrinkage (i.e. sick time), and agent attrition. There are a variety of modeling methods available, like Holt-Winters. The call volume forecast is a fairly high profile result of these models.
Here are three tips to help with your forecasting:
- Forecast everything. It is important that your volume forecast be correct, but it is equally important that attrition forecasts and sick time forecasts be accurate as well.
- Measure variance to forecast. Once your organization has produced their regular forecasts, the difference between what happened and what was forecast is often a very important bit of information. If there is significant variance it needs to be explained.
- Variance to forecast can be an early warning device. If the forecast is consistently off, it may mean that something is changing in the operating environment that needs to be addressed. Because of this shift, you may need more or less agents.
Descriptive models are techniques that simulate your environment to allow analysts to ask what-if questions. The most common descriptive model is the model that determines how many agents you require to hit service goals for either short term or long term planning. The most widely used descriptive model is the old Erlang C model.
Here are three tips:
- Ditch Erlang C. It is well-known to over-staff, and it does not work in a multi-skilled agent environment.
- Validate your models. For the how and why -read my previous blog post
- Look into discrete-event simulation modeling to perform long term planning. It is very accurate and it works for multi-skill, multi-channel, and multi-site call centers.
Prescriptive models are models that prescribe an optimal solution. The most common prescriptive models are ones that determine agent schedules or agent hiring and overtime plans. These models are very effective because they perform tasks that are very hard to perform well by hand (remember scheduling using a spreadsheet?).
Tips for using this model:
- Use integer programming. For long term planning, make sure the hiring, overtime, and training optimizer uses a form of integer programming. Integer programming will determine a just-in-time plan that hits service goals at least cost.
- Make sure your scheduler is integer programming, too. The most efficient schedules are integer programming-based.
- Use a prescriptive model. Make sure hiring and overtime planning is performed using a prescriptive model (some contact centers develop hiring plans by hand or with a spreadsheet). Using models to develop your capacity plans can save your organization 5-10% of its paid agent hours!
Finally, the most important thing I’d like to leave you with is that mathematical models, when properly developed and validated, are terrific tools to improve the efficiency of your operation!