The Power of True Optimization

The Power of True Optimization

By Ric Kosiba, Vice President, Interactive Intelligence

The word “optimization” means a lot of things to different people, and I think it is a word that is a tad overused. We hear it in meetings, in product descriptions, and we even hear it on television. Most people use it as another term for “to improve.” Mathematicians view the word completely differently — to them it has a very specific meaning. To mathematicians and engineers, optimization means to achieve the mathematically provable best solution that is possible. And that is an important distinction, because “provable” is a very high bar. Mathematicians and operations researchers go to great lengths to prove that there is no other better solution to their optimal solutions.

Sixty years ago or so, a few mathematicians began developing algorithms that would solve optimization problems in a provably optimal way. The first widely known algorithm was called the Simplex Method and it solved a set of specific problems, called linear programs, optimally. This algorithm, developed by a gentleman named George Dantzig, has saved companies untold amounts of money by allowing them to develop optimal plans or schedules or processes and, hence, to operate more efficiently — provably optimally efficient.

There are all sorts of ways that mathematical optimization touches our lives today. Our GPS’s run on optimization. Our water resources are managed with optimization algorithms. Products coming to market are produced and delivered in cost-effective ways, using mathematical optimization.

In the contact center industry, mixed integer programs, a form of linear programs, have been used for many years to produce mathematically optimal agent schedules. I do not believe it is an exaggeration to say that George Dantzig and his Simplex Method has saved the contact center industry billions of dollars through the use of truly optimal scheduling algorithms. More on this in a bit.

When I was a young engineer, I developed a linear programming-based system to help schedule an airline’s ground personnel. It was a very important project at an important time; the airline was losing money hand over fist, and was well known to be operating in an extremely inefficient manner. The prior scheduling process was very low-tech — it involved an analyst who developed a scheduling spreadsheet (in Lotus 123!) and used the algorithm of “guess and test.”

Scheduling falls into the category of problems that are very difficult for human beings to answer well, because it is simply too complex a puzzle to solve efficiently without the use of algorithms. But thanks to Dr. Dantzig I had the tool, the Simplex Method, to solve this scheduling problem. The results from the algorithm were terrific — it scheduled airport agents to cover the work as near to exactly as possible, with as few extra agents as possible. Using this model saved the airline many millions of dollars, it provided consistent service, and it helped to bring the airline back toward profitability.

My boss asked me to explore whether we could bring the same technologies to call centers, and I went over to our reservations center to see what I could find (which is how I tripped into the contact center industry).

But I was told that the airline’s reservation centers used mathematical optimization already, and they had one of the earliest workforce management systems, TeleCenter System from TCS Management Group. My first business trip ever, was to fly out to Nashville to visit TCS to determine whether their scheduling algorithms were really, truly, mathematically optimal. They were, and I had to find other problems to solve.

The point to all of this though is to discuss the benefits that we saw by using mathematical optimization. In those days, switching from our spreadsheets to a workforce management system meant going from manual agent scheduling to mathematically optimized agent schedules. The application of mathematical optimization improved the efficiency and lowered the costs of our operations by as much as 15%, as I seem to remember.

Our most expensive resource is our workforce, and a reduction in agent costs, simply due to better scheduling, was a huge improvement to contact center operations that reverberated throughout our industry. But other than costs, there were all sorts of other benefits. Workforce management operations were able to test new schedule permutations that were hard to support in spreadsheets, like rotating shift schedules. They were able to perform scheduling what-ifs. They were able to provide more frequent shift schedules. They were able to show the operational costs of schedule non-adherence. Maybe more importantly, they were able to better control the consistency of service delivery. All of this was realized because of some cool math.

Now any term with the word “mathematical” is bound to scare many people. But, hey! We are workforce managers! We love math and problems, and linear programming is not really that hard. There are still many cool contact center problems that we still can solve using mathematical optimization. Look at your job — if there are any processes that use spreadsheets and “guess and test,” odds are an optimization algorithm will make that process quicker, faster, and more efficient.

Ric Kosiba, Ph.D., 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.