That seems backwards, right? Haven't we always assumed that the smart way was to isolate each change you want to make to a process so you can control the experiment and determine the value, success or failure, of the change?

Apparently, there's a way to assess an entire battery of changes all at once, using statistical sampling techniques. It was developed by a scientist named Charles Holland for use in nuclear weapons plants. In a newly published book called Breakthrough Business Results with MVT, he and his co-author have outlined the kinds of effects this kind of analysis can have through name-brand case studies.

The call center implications are marvelous, and detailed. One of the reasons why MVT has such potential in call centers is because the center is an environment in which every action is measured (and many are scripted). It makes changes and substitutions easy to articulate.

In MVT, you identify the factors (up to forty of them) that you'd consider altering. These can be simple changes: the wording of a greeting; the particular mix of products in an upsell offer; the addition of specific training to reps; the list is potentially endless. Once you identify the possibilities, MVT analysis creates a matrix of "recipes" that allow you to test all the considered changes in combination with the other changes. You run the "recipes" on a series of limited target agent groups, and voila, you can see definitively which changes help, which hurt, and which have no effect. In the example I was shown, detailing a wireless phone call center, there were 24 factors identified, which were broken down into 28 statistically balanced "recipes" - each of those recipes was run by a selected team.

One factor, or variable, was whether or not to set a revenue target. Does setting a revenue target for reps increase sales? No. It was found that it hurt sales by a fairly large margin.. Providing sales training also hurt sales results. Offering upsell features on every call? Made no difference at all in sales numbers. Surprising, a little counter-intuitive, but definitive.

MVT doesn't get into the question of why things work or don't work, but it does crunch out quick answers to the what works question. In the wireless center, having regular department meetings for the reps improved sales more than almost anything else.

MVT essentially shortcuts you towards fast fixes - the ones that are lowest cost, highest value within your center. It doesn't direct you to complex process improvements; it encourages you to throw all your ideas into the hopper and then it weighs the relative strength against the cost-effectiveness of each idea.

The hitch in this methodology is that not everyone can do it. Designing the test sequence - the mix of recipes that will give you a statistically valid answer to your multvariable questions, is a specialized skill that's not widely available. One of the reasons that QualPro has made a business out of MVT consulting is because it's so ridiculously complex that it takes actual scientists to design properly (there's that math thing again). My sense, though is, that it's a legitimate - if not outright brilliant - solution to the problems raised by call center operations. Call centers are nothing if not cauldrons of multivariability.

Sometimes it really does take complexity and higher-order expertise to arrive at the simplest of solutions.