Seems like a pretty straight forward question, right? It would be, if service were free. But service is not free.
At each service level goal, there is an implied cost. And lowering cost is another of those goals. So our two goals are: lower costs and provide service. But those aren’t the only two goals! We also don’t want to burn out our agents with too high an occupancy, we want to provide training and professional growth for our employees, we want to sell product (for sales or collections centers), and so on…
A simple trade-off sensitivity curve, cost versus service level is below.
How we answer questions with competing objectives can be tricky, but what is certain is that the analyses required to make such trade-offs explicit is a longer term model of the contact enter network. The analyst needs to be able to determine cost, service, occupancy, training, overtime, and a bunch of other important metrics associated with any planning scenario. If you know the value of these metrics for a scenario, you can determine the trade-offs associated with competing scenarios.
By using your strategic planning system, you can find the answers to our multi-objective operation.
The idea being that executives should push for service consistency across intervals was one of the catchwords in our industry a couple years ago.
We began hearing of goals like “We want 90% of all intervals to have service levels greater than 80% within 20 seconds.” I may sound foolish here, but I don’t get this thinking.
In my humble opinion (Ric humble!?! HaHa), I would think that if you correctly measure service level as being across all of your customers, 80% of all customers being answered in 20 seconds or less should be good enough.
My guess as to why this concept got some traction is that maybe the way daily service was reported may have been wrong. Instead of volume-weighting service level across the day (service level as a percent of customers), service level may be interval-weighted in some reports (service level average across the hours, say). If you add up the service level of all of your intervals and divide by the number of intervals, then you are weighting your busiest period the same as you least busiest period, and that is likely a mistake.
In other words, service level reporting had a mistake in it, and so we had to create a new service discipline around our reports. That’s the only thing I can come up with to make sense of this other metric. But I can’t really get my arms around it. Please explain?
One thing we talk about during sales calls and during webinars is a different concept of service consistency. But we mean this in a very different way: we mean that we will make sure you have the exact right number of people available each week so that week over week- through all the seasonality- we maintain our service with little stress to the operation.
The idea being that the seasonality of all sorts of metrics: handle times, volumes, sick time, attrition, etc… make the development of week over week plans difficult. If you do that well, like CenterBridge does, then you will have service consistency.
There is a saying that goes something like: “Don’t make good the enemy of perfect”. What this means is that if something is not exactly right, it should not necessarily be discounted because it may be way better than the status quo.
A long time ago, I was presenting an ROI for a project I was working on. While presenting, we showed an honest to goodness, conservative improvement to the operation of 7 million dollars. The client told me that since the system (not CenterBridge) took 12 minutes to calculate the optimal answer (which only was a small part of the analyst’s time in the system), they were considering turning the system off.
My response to this person was “How much is your time worth?”
Would it be perfect that the optimizer returned a response in 5 seconds? Yes. Is it pretty darned good that it saves the company 7 million dollars a year even though the algorithms solved in 12 minutes? Of course it is. Don’t make good the enemy of perfect.
Did you ever think that shrinkage may affect your profitability? It can.
In this graph, we are holding everything constant, and only varying shrinkage (the same sort of graph that we saw with service level versus profit). In this case we hold all constant but shrink, and track the profitability of the sales center. Again we find a profit-maximization point!
My favorite graph is service level versus profitability. In this graph, we vary service level by adding staff, and track the resulting profitability. For most revenue producing centers, you should see a shape like the graph below.
This is a textbook, marginal profit curve. What we produce is a change in profitability by agent, meaning, if I hire one more person, how much will they sell, and what will they cost? If your goal was a 60% service level, you would be leaving revenue on the table. If your goal was 90%, you would be paying too much in agent costs relative to the revenue being produced. If you wanted to run the center as if it were a business, you would run a service level between 75% and 85%, in order to maximize profits.
There is a lot that goes into these graphs (and thank goodness I have CenterBridge to produce this one), but the concept is pretty straightforward. If the marginal value per call changes, these graphs shift substantially. I’ve got a few stories about the use of these in the real world, and I’ll try and post these soon.
Dr. Amit Garg of Operations Research Department has just made a successful world record breaking attempt for mentally solving 10 problems of division of 10 digit numbers by 5 digit numbers. He was able to beat the existing world record by 22 seconds (6:07 seconds to 5:45 seconds). Amit will send his result to Guinness Book of World Records’ office to officially be recognized as the new world record holder for this category.
Watch the Video
Check out other Memory and Mental Calculation World Records.
This next graph is pretty cool, and one which has been, prior to CenterBridge a very difficult piece of analysis to provide. It is simple: at each service level, what is the expected abandon rate?
For each contact type this graph will be very different- it depends on the customer patience of each type of caller.
This graph serves as a starting point to some very cool analyses. By knowing the expected abandon rate, you also can derive the expected revenue! You can get a very good picture of the number of customers you’ve ticked off (they hung up on you!). And you can get closer to the mother lode, for revenue centers: the relationship between staffing and profit (we’ll discuss that one soon).
About the last graph (pictured above): I think CenterBridge is still the only place in the whole workforce management space where abandons are forecasted and determined, as a function of the resource plan. In other words- we can figure out the marginal abandon rate: “if I hire one more person, how many fewer abandons would we have?”
THAT is very powerful, big picture stuff.
Here’s another cool graph. On the X-Axis, we are plotting shrinkage, while determining its affect on abandon rates. What this demonstrates is the cost- in revenue and abandons – of shrinkage. If your center is a sales, reservations, or collections center, then shrink has a lost revenue cost.
This is one of the real benefits to modeling abandons within CenterBridge. You can provide cost and revenue analyses associated with abandons, which is the next graph!
More fun with graphs. This next graph is a very cool graph, and it represents the cost at various call center service levels for a specific (and made up) call center. On the x-axis, we have service level, which is calculated by simulating the service different staffing levels (holding volume and handle times constant). But since we also know the cost at each staffing level, we can plot service versus cost.
This graph is instructive in drawing out the trade-off for service centers- it helps answer the question “what does one percent increase in service level cost me?”
If you are an outsourcer, this graph will help you price new business opportunities. If you are a customer service function, this graph helps determine an implied value of each call.
One other note: if you use an Erlang equation to develop this graph, it will be wildly off.
It’s a contact center cliché: The only constant is change. Whether the unexpected is a rebound of the economy, or a weather event, or some new edict from the government, we must somehow tame the effects of the heavens and Washington and the invisible hand of the economy.
It is not easy, and if we knew how to forecast any of these events early enough to be able to plan specifically for them, we would certainly be in another line of work.
But even if we don’t know exactly when these events will happen, we certainly know that they will happen. It doesn’t mean we cannot prepare for these events.
In this session, we will discuss mathematical methods for planning in the face of change and how to prepare your contact center for the unexpected.
Thursday, April 12th at 1pm EST