Multivariate testing of websites has come to the fore amongst many companies over the last six months, with many digital teams doing multivariate testing to see which different version of website pages perform best. Powerful stuff and clearly a great way to tease out performance improvements. But where do the different pages emerge from?

There’s a whole process of audience research, persona and scenario creation, customer journey analysis et al before you come up with a site design that is likely to serve your customers well and deliver good conversion rates. And with existing sites, customer experience testing laid over web traffic analytics gives a powerful “why” over the “what” of the analytics.

The danger with multivariate tools is that much of this is forgotten – if it’s easy to create different treatments and see which is best, why not just try all sorts. Play with the copy, change the colours, and see which works best.

This is akin to lining up your cannon, firing out two random shells and then celebrating the news that one was closer to the target than the other. You could do this for years without getting near your target. Better to remember the basic disciplines of website success – understand your target audience, the journeys they take from interest to sale, and how they experience your brand offering online. Get the basic offering right, then use multivariate testing to tweak the offering for continuous improvement.

There is a valuable role for multivariate testing, but uninformed site design is not it.