How to Ensure Service Consistency
By Ric Kosiba, Ph.D., Interactive Intelligence
Let’s start with the obvious. Contact centers are seasonal operations. We all know that volumes vary by time of year, and most of us go to great lengths to employ sophisticated forecasting models to predict the seasonality of contact volumes.
However, volume is not the only seasonal metric that is important to forecast correctly. Handle times vary by year as customers call us for different reasons throughout the year. Outbound contact rates vary seasonally (try getting hold of a customer during vacation time!). Agent attrition, agent absence, and agent sick time vary by season and center location. Interestingly enough, customer experience scores (customer satisfaction, net promoter score, etc…) often vary seasonally and by center location, too.
An aside: Using sensitivity analysis, it is easy to demonstrate the effect on service delivery of staffing to a poor call volume forecast or the wrong shrinkage forecast. We’ve done this analysis and it shows that accurately forecasting shrinkage is at least as important as forecasting volumes accurately. But how many of us take as much time to forecast sick rates (by center and staff group) as we do volume forecasts?
Consistent Service is Important
I think if we were to ask our senior managers what was most important for workforce management to do well, I expect “consistent service delivery” would be high on their list. We all know that delivering variable service levels week in and week out is not a path to job security for either our executives or ourselves.
Consistent service delivery is important to our businesses. Given that the call center is often the primary touch point between our company and our customers, meeting service expectations is critical to our company’s mission. So, we have our own charge — we need to tame this seasonality and provide consistent service.
It is well known that traditional workforce management software focuses primarily on short-term operational management, today to a few weeks out. Because of this, most of us have built our own capacity planning spreadsheets or bought a strategic planning system, in order to plan for those things that are not short term in nature, like hiring, vacation planning, training planning, etc… What happens to the workforce management process if our capacity plans are wrong?
Capacity/strategic planning systems serve to ensure that when the call center opens Monday morning, the exact right number of agents sit down at their desk and are available to work. If the planning process is off and we have the wrong number of agents available, either we will have a “day-of” overstaffing problem or an “emergency” understaffing problem in the call center. We will be scrambling to send agents home early or scrambling to find more resources. It makes our real-time workforce management job much harder, and it also makes delivering consistent service much harder.
What Do We Need To Forecast?
Most of us spend a fair amount of time working on our volume forecasts and handle time forecasts. However, any historical item that exhibits seasonality can be forecasted. But should we forecast an item simply because we can? We only need to ask ourselves whether that specific metric affects our resourcing decisions and if it does affect our decisions, we should take the time to forecast it!
Like we mentioned earlier, all of our shrinkage items are critical to forecast well, because they affect staffing significantly. Outbound shops need to forecast contact rates and right party/wrong party handle times. Chat operations need to forecast handle times by concurrent sessions seasonally. Email and back office functions need to forecast handle times by work step, as well as the process flow and how the process changes over time.
How about customer experience scores? Certainly there is seasonality to these sorts of metrics but do they affect decisionmaking? Of course they do. Customer experience forecasts help us prepare our executives for seasonal trends in customer satisfaction. They also help target training programs for seasons when our operation historically performs poorer. Customer experience trends and forecasts can also help to show how different contact centers perform relative to each other. It is perfectly reasonable to staff centers differently based upon their customer experience performance.
“Random” variability is the second cousin to seasonality. It is when the operation has events that significantly affect service delivery, but the events occur randomly without any discernable seasonality. For example, insurance companies or power companies see spikes in calls when the weather gets bad. We know these events will happen, but we don’t necessarily know exactly when they will happen.
The first step to taming these random events is simply to tag when they happen in our history and approximate the magnitude of the spike relative to a “smoothed out” forecast. We want to know how big each event is and the frequency of these events by season.
How to Manage Seasonality and Variability
The true key to managing seasonality is to have a robust and automated capacity/strategic planning process, simply because it allows an analyst to evaluate how changes in forecasts affect costs, service delivery, customer experience scores, revenues, and profit.
There are five major functions to a robust capacity planning process.
First, it needs to gather, store, clean, and validate historical time-series ACD, workforce management and variable labor cost data. Second, it needs to forecast contacts, handle times, and all the other important metrics quickly and accurately. Next, it needs to accurately produce week over week staffing requirements. Fourth, it must automatically produce efficient capacity/staffing plans. Finally, it needs to automatically convert these plans into budgets, including both costs and revenues (e.g., sales functions).
This process has three benchmarks:
It needs to be fast. Developing a forecast, determining staff requirements, developing a capacity plan (hiring, overtime, undertime, and controllable shrinkage plan), and producing a budget needs to take a few minutes rather than hours or days.
It needs to be accurate. Forecasts need to be tight but more importantly the model which translates volumes, handle times, and staff available into service levels, abandons, occupancy, and ASA needs to be very accurate. Erlang is not.
It needs to be mathematically optimal. The capacity plan should be developed using optimization technologies such as integer programming. These algorithms should optimally determine when and where to hire, when to offer overtime or undertime, and the best time for controllable shrink (i.e., training).Multi-site and multi-skill centers are complex and difficult to evaluate by hand. Staff/capacity planning should use the power of algorithms to ensure just-in-time staffing.
If you have these three things – speed, accuracy, and optimality – your analyses will be consistent and flexible. Developing such a process will result in plans that always produce consistent service delivery at least cost.
What about volume (and other metric) variability? By producing several plans, the smart analyst can develop a breakdown of risk versus certainty of service delivery versus cost for any forecast scenario; enabling them to evaluate options for taming “random” variability. For example, the analyst will know how many extraordinary events happen during the summer season historically, so they can plan for the “average extraordinary” event, or the “90th percentile extraordinary” event.
By using a robust planning process, the costs and operational risk of each plan is well understood and known within minutes!
As much as we would like for forecasting to be something we do once a year, we know that all of the items we need to forecast can trend, change, and shift. Our forecasts are not always stable, so forecasting and planning needs to be a discipline; monthly reforecasts of all of our metrics seem to be an industry best practice.
By actively managing our hiring plans, our overtime plans, our controllable shrinkage plans with a process that is fast, accurate, and optimal, we will ensure that the number of agents available each day is exactly the number we needed. We are glossing over many details here, but feel free to reach out to me if you would like to discuss this in much more detail.
Ric Kosiba, PhD is a charter member of SWPP and vice president of Interactive Intelligence’s Decisions Group. He can be reached at Ric.Kosiba@InIn.com or (410) 224-9883.