01 June 2006

Marketing and mathematics or selling for engineers!

Debbie Jenkins has been promoting Lean Marketing for several years. That took the concepts of Lean Manufacturing and applied them in a Sales and Marketing environment. Multivariate testing takes the concept further. It has begun to show good results for simple campaign scenarios which would once have been slowly optimised by split testing. Most of these multivariate approaches seem to derive from Taguchi methods which have been used in manufacturing for many years - I first encountered them in the early 1970s but they have been in use since the second world war. It is only relatively recently that they have moved out of manufacturing and on to the marketeer's desk.

Taguchi methods, developed by Dr. Genichi Taguchi, refer to techniques of optimisation that embody both statistical process control (SPC) and experimental design. Most of the attention and discussion on Taguchi methods has been focused on the statistical aspects of the procedure, but it is actually a conceptual methodology for improving marketing processes.

The entire concept can be described in two basic ideas:

1. Performance should be measured by the deviation from a specified target value, rather than by conformance to specified norms
2. Performance must be built in through the appropriate design of the process and approach

Taguchi testing is more comprehensive than split testing, since it assesses the performance of a system while testing many variables simultaneously, and attempting to isolate the effect of changing any variable. In most split testing exercises it is the imagination of the team that is being tested - their ability to continue to improve the performance of a piece of copy or a site based on some gut instinct about what might be better than the status quo. In multivariate testing, all kinds of combinations can be tested within a relatively short period of time and it doesn't put a heavy load on the marketing team to come up with all those scenarios.

This kind of modelling can produce better and more comprehensive outputs than split testing, typically within half the time. The problem for the multivariate testing gurus is that I doubt that many people have the statistical capability to understand what is happening. Couple that with a lack of trust of black boxes, particularly when the approach comes up with a combination of variables which no-one in the organisation had thought of.

If there is a Business Case then people would be stupid to adopt it wouldn't they? Well, up to a point. Putting it back to a supply chain example, the mathematics of Economic Batch Quantities and Re-order points are well-known. It doesn't mean that they are rigidly applied in the majority of supply chains.

Taguchi is interesting - it deserves a wider audience in marketing.


Post a Comment

Links to this post:

Create a Link

<< Home