The Taguchi Files: How Multivariate Testing Boosts Email Marketing Results
Multivariate testing was pioneered in the late Fifties by a Japanese engineer and statistician named Genichi Taguchi who developed it to improve the manufacturing quality standards in textile mills.
Dr. Taguchi's calculations have found a fertile ground in the email marketing of the 21st century as the parallel multiple element testing procedures are proving to be undeniably effective in the optimization of every aspect of an email newsletter campaign.
A/B tests can take decades to exhaustively complete
Prior to the implementation of Taguchi's innovations, the standard marketing testing consisted of a simple A/B split: Two different versions of a marketing approach were tested at the same time and the one that performed best was used.
Although this strategy can lead to increasingly refined efficiency it requires continual, linear, and overwhelmingly long term testing to achieve precise results.
Interactions between elements are invisible to A/B
In split A/B testing the email marketer is limited in examining one element at a time. The same email is sent out but with two different subject lines; then the best subject line is used in another email which uses two different calls to action; and so on.
When the entire wealth of email factors are considered such as text content, graphics, colors, links, and presentation order you soon realize that conventional A/B testing could be infinite and never come up with definitive answers. Another weakness of A/B testing is that it cannot reasonably provide data as to how the combination of factors are affecting results.
Perhaps subject line A with graphics B provides a better click-through rate than content C with presentation order D. It is overwhelmingly difficult for A/B testing to provide that dimension of statistical data for your analysis.
Equivalent to running multiple A/B tests at one time
A Taguchi multivariate test is able display the interactions between disparate elements and be set up to provide that invaluable data to fine tune your marketing strategies. Think of it as running a number of A/B split tests but all at the same time.
By creating a number of different emails with a variety of different and fully independent elements it is possible to run a complete test in parallel which can then be interpreted by any of the various widely available multivariate statistics software packages.
List churn has to be taken into consideration
An important factor to consider is that any form of email marketing testing will vary according to the timing of each send.
Not only are there seasonal factors to consider but the majority of email subscriber lists have considerable churn, therefore the results which can be garnered from a list may vary considerably from the results of the same test from a list that has undergone a 40% turnover.
The key to achieving consistent and accurate multivariate testing results is to implement the procedures into each major campaign you send, and to keep testing in order to further refine the findings.
There are endless iterations of multivariate testing and you'll find an a number of options that adhere to Dr. Taguchi's precise calculations and others which do not.
Some mathematical experts even differentiate between multivariate testing and the Taguchi Method. From the standpoint of the average ecommerce email marketer, those finer points are best left to the theoreticians to debate.
The single most critical factor is that by implementing these parallel testing functions you will be able to achieve far more accurate results in a much shorter time than A/B split testing, and the incremental improvements you will be able to integrate into your marketing campaign will quickly show up in your bottom line.