Simulation modeling and Madden: Understanding “What if” analyses through football

Many years ago, I got my first promotion for drawing a simple graph.

I had been asked a question: Given changes we were expecting at our airline, what should be our servicing strategy?

That question is the sort of question that you would ask a $500/hour consultant, and I certainly wasn’t one of those. But my boss knew that I had built a discrete-event simulation model of the contact center network, and he felt it was time to take the model out for a spin.

Many of you haven’t heard of discrete-event simulation modeling, but if your home is like mine, you have several modeling experts living right in front of your television; the best example of simulation models are computer games.

My son is addicted to the Madden Football games, and they are a perfect example of a discrete-event simulation model. In this game, when my son pushes a button for the quarterback to pass the football, the game console suddenly calculates the odds of the scenario. Is the quarterback being chased?  Is the pass short or long? Is the receiver covered closely?  Does the receiver have great hands?  All of these factors go into the computer in order to calculate, using probabilities, the odds of the pass being completed. When the odds are determined, the computer virtually “rolls the dice” to determine if this specific ball is caught. Throughout the game the console is calculating the odds of all sorts of events and rolling a lot of dice.

Now contact center simulation models do the same thing, just without the cheering. Instead of calculating the odds of gaining 5 yards or fumbling the football, the contact center simulation is determining the odds of an agent being available, the odds of a customer abandoning the call, the likely handle time, the likely call routing, the odds of a customer buying product, and more. Given that these models can calculate the expected results (costs, service provided, sales, agent occupancy, customer experience scores) of a specific scenario, it is possible for the system to evaluate all sorts of possible scenarios.

For instance, a contact center model can simulate the performance of a contact center as staffing increases or decreases. It can determine service levels as handle times increase or agent shrinkage decreases. It can accurately estimate the effect on sales as agent attrition changes. This makes the simulation model the perfect “what if” engine, and honestly, the purpose of simulation models is to provide fast and accurate “what if”s.

So back to my first promotion. The airline was expecting new competition at their hub, and they knew that their prices would drop. They also knew that price drops meant more phone calls in their centers (they were going to market these as fare sales), so they wanted to know how many agents they would have to hire. We developed a few scenarios to evaluate, with different ticket prices and call volumes. For each of these, I developed a marginal profit curve:


To create this curve, we simulated the contact center over and over again, adding staff between each iteration, and holding all other performance drives (like volumes and handle times) constant. We noted the service levels at each staffing scenario, as well as the number of abandons. As staffing increased, service levels increased and abandons decreased. Applying a variable labor cost model and noting the sales per call, we can draw the marginal profit curve.

This curve is extremely interesting in that it shows us, clear as day, the expected profitability of our contact center network at varying service level goals. In the picture above, we would want to run a service level goal of 65% of calls handled within 20 seconds service goal in order to maximize the profitability of the operation. If we ran our center network as though it was a business, this is the goal we would choose.

Set up properly, simulation and “what if” analyses are very powerful tools. They can help decision-makers answer very tough questions; $500 per hour consultants need not apply.

Find out more about what-if analyses by downloading our eBook, “Get Easy Answers to Hard Questions.”

Ric Kosiba

Ric Kosiba

I joined Interactive Intelligence in August 2012 as part of the Bay Bridge Decision Technologies acquisition. I helped found that company back in 2000 and thoroughly enjoyed working with our brilliant development and operations research team, which helped us become the leading U.S. supplier of long-term forecasting and planning solutions. In my current role as vice president of the Bay Bridge Decisions Group, I’m responsible for the development and enhancement of our contact center capacity planning and analysis product line. I tripped into the call center industry about 22 years ago and can honestly say that I still love it. I hold an M.S.C.E., B.S.C.E., and Ph. D in Operations Research and Engineering from Purdue University (go Boilers!). I reside in Maryland with my wife and four children. I love being a dad and enjoy coaching kid’s football, basketball and lacrosse.