AUTOMATED MULTIVARIATE OPTIMISATION

Direct marketers have been multivariate testing their promotional content for some time and measuring customer response rates.

Online conversion metrics can relate to any interaction from click through to a download or purchase. Content permutations are very numerous when combined and so there is a need for a solution that tests content simultaneously, intelligently and gets quick results. More tests in less time equals more revenue uplift!

How do we perform a multivariate test?

So a web page might have 6 content areas (maxyboxes) to test.....

...and each maxybox might have 6 content assets which could be assigned to them. Content assets can be entirely different from each other, for things like promotions, or could have only small changes in information, layout or design.

Maxymiser immediately starts to note which combinations perform in lower and upper values of the conversion metric and then focuses testing towards more positive results.

How our artificial intelligence is applied.

Random generation
As the names suggests, content samples are generated randomly and variants are shown in equal proportions to new visitors. This allows you to see statistical data for all Maxybox variants and to analyse the interaction between all different Maxyboxes and content variants.

Taguchi method
The Taguchi method focuses only on the most important areas and interactions in relation to conversion and enables testing of all variants without testing all of the combinations. A test can run with a smaller visitor sample than Random Generation and results are gained more quickly. As a result tests can be acted upon more quickly allowing more tests to be undertaken in a given timeframe. The Taguchi method gives valid test results in around a quarter of the time required for Random Generation. And time is money!

Continuous optimisation
This method allows Maxymiser to automatically optimise content to achieve the best metric uplift. Our machine learning algorithm acquires knowledge about what works, what doesn’t work and how to map variants to maximise an outcome. The algorithms discover which actions yield the highest reward through testing. Actions may not only affect the immediate test but also the next stages of any process. Our algorithms learn this and will sense the effects on the overall environment taking actions accordingly.

With a continuous optimisation methodology it is possible to see uplift in your metrics quickly and any underperforming combinations of content will ruled out to deliver quicker uplift.

Usability Optimisation

A/B testing simply resolves which of two versions of a page produce the best conversion. Instead of doing this manually Maxymiser will automate this testing process and produce a quick result. This works well for testing simple page form and structure variances or single content elements in two versions.