This practice is part of: Product Ownership.
What is it?
Multivari, aka Multivariate or Multi-variable testing, is a product testing approach, which allow teams to understand influence of multiple different factors on the product performance.
If you have many factors (potentially) influencing the performance of a product, an A/B testing approach will take enormous amount of tests and time, because you can only test the difference of the product performance with two different values of a particular factor (A and B) at a time. An A/B test will also not tell you of potential reinforcement between factors.
This is where Multivari testing comes in. Unlike A/B testing, Multivari testing will allow you to design product versions which differ from each other on multiple different factors. As in A/B testing all of the different product versions will be available in production and customers will be randomly redirected to different versions. The results from the customer interactions with the different versions will be collected as raw data.
You can then understand the impact of each factor as well as the interplay of those factors using a statistical analysis of variance (ANOVA or MANOVA) over the collected data.
It is important to note that multivari testing requires more transactions/interactions with the product in order to achieve a statistically significant amount of data. If the data is insufficient the results will not be representative of the whole population, i.e. not meaningful.
Several variations of Multivari Testing to consider:
- Full factorial testing - testing of all possible combinations of the factors
- Fractional factorial testing - only partial combination of the factors is tested
- Taguchi testing - a partial combination defined with use of heuristics or other methods
Why use it?
This practice can save significant efforts and time when assessing the impacts of different factors on the performance of a product/service. When multiple factors are at play, an A/B test becomes hard to impossible to execute and as pointed an A/B test will not really provide answers on the interplay fo factors.
The results of the testing will ultimately provide answers to the question which combination of factors (e.g. features) achieves the best performance.
This is why this practice is one of the important practices in the pivot phase of product development.
Image credit: Photo by Steve Harvey on Unsplash