Over the last ten years, "new" and distinct mathematical modeling technologies, coupled with improvements in computer processing speed, have enabled more sophisticated contact center analytics. While most of these modeling methods have grown up independently of each other, they derive their "super model" power by working in conjunction; it is these technologies that enable huge improvements in decision-making for contact centers. These include data warehousing, forecasting, discrete-event simulation modeling, and mathematical optimization (integer programming) technologies.

These technologies, used together with a strong business process improvement orientation, enable the development of an Enterprise Analytics business process. There are four such technologies/processes:

1. Automated Forecasting and its Appropriate Role

In many, if not most, organizations, the role of forecasting is both simple and very narrow: to determine the expected contact volumes and handle times accurately. It is narrow because contact volumes are neither the only nor the most important business driver to forecast. It is simplistic because the true value of a forecasting team is not a single forecast, but is a part of a contact center monitoring system and a larger planning process. Leading analytic organizations recognize this. They view forecasting differently:

*They view forecasts as the baseline and variance to forecast as a warning indicator to understand. They worry less about forecasting "error"; they instead assume that any variance to forecast is either natural variability of the business or a change in the environment that needs to be explored. For example, if absenteeism is higher than expected, it is a warning that something else has changed in the operation.

*They automate forecasting so they can apply forecasting expertise to every important contact center metric. Call volume forecasts are important, but so are handle time forecasts, attrition forecasts, sick time forecasts, training plans/forecasts, vacation plans/forecasts, wage rate forecasts, etc... It is essential to make this forecasting process as easy as possible, to allow for the use of sophisticated forecasting methodologies of these other important items.

*The best forecast does not necessarily mean the lowest "error." Standard methods of determining forecast error are helpful, but not necessarily the best way to judge between competing methodologies. The final downstream product of the forecast is a set of decisions and the forecast that produces the best decision is the better forecast. When viewing competing forecasting methodologies against hold-out data, the best forecasters take the next logical step and ask, "which methodology poses the most operational risk to the organization?" Oftentimes, it is not the methodology with the lowest Root Mean Squared Error or absolute error.

There are a myriad of mathematical technologies available to forecasting analysts; however, the most important item to consider is that the data stream being forecasted matches well with the mathematical methodology chosen.

But often, contact center executives can improve their forecasting process by simply reminding their forecasters that the purpose of the forecast is to make decisions, and to focus their analytic team on that direct purpose. Which forecasting methodology will yield the staff plan with the least amount of risk?

2. Automated Variance Analysis

Variance analysis is usually used in the context of budget analysis. That is, variance to budget is an item explored but only if line item costs are too high. This is clearly short-sighted. In contact center operations, variance to plan should be regularly analyzed to certainly include costs, but also to include all major assumptions associated with the strategic operating plan. This includes (by center and staff group), wage rates, handle times, volumes, vacation plan, employee attrition etc.

3. Developing Response Plans

Enterprise Analytics require two key planning capabilities. The first is the ability to simulate the operational performance of contact center environments quickly and accurately. The second is to automatically and optimally develop best response business plans given the appropriate business constraints. The mathematics associated with modeling these two functions have been available for decades, but it is only recently that computer speed has allowed these to operate fast enough for contact center business use.

4. Enterprise Performance-Risk Outcome Matrix (EProm)

The final step in the Strategic Analytic process may be the most important. Because mathematical technologies automate much of the strategic planning process, the time required to forecast, build what-if scenarios, and determine the best business response to every scenario is surprisingly short. More importantly, it is also of a significantly higher quality (i.e., more accurate and comprehensive) than manual or spreadsheet planning processes. These technologies allow a different take on the planning process – they allow us to monitor and optimally plan for business uncertainty.

By developing an Enterprise Analytics process, businesses can make strategic decisions almost casually, as a matter of course. No longer do "strategic initiative" processes require expensive consultants and several months of strategy meetings. Strategic decisions happen as part of the normal course of making business decisions. This is very powerful.

What do contact center executives want? They want answers to their business problems in a timely manner with a real expectation of accuracy. Mathematical modeling technologies, through an Enterprise Analytics process, fulfill the promise of real decision support for contact centers.