Why you’re doing conversion testing all wrong

If there’s one thing marketers have plenty of, it’s ideas. You’ve got slogans and copy you want to try, new images you’d like to test, a design you think is perfect, parts of your site you’re not sure are working, and on and on. But how do you know which of those ideas are actually good? How do you know which will actually resonate with your customers?

Conversion rate optimization (CRO for short) has long been the way we’ve tried to answer exactly those questions. Typically, that comes in the form of A/B testing, where you compare your current site (the control, or A) with your new idea (the test, or B). The problem is, while you’re testing that one idea, all your other ideas are just sitting there, gathering cobwebs. And since it’s hard to guess which ideas will drive more revenue–in fact, only 1 in 6 are successful at all–every test that fails is essentially lost money.

Confirmation vs. Exploration

See, a major problem with A/B testing is it’s confirmatory. You’re confirming whether a single idea is better than what you have. You’re, in essence, flipping a coin. And since those tests can take weeks or months, you’re not moving quickly as a department.

What if, instead of confirming you have a good idea, you could explore all your ideas, even across your entire funnel? What if instead of guessing what your customers wanted, you could actually have them tell you? What if you approached testing in a whole new way?

With artificial intelligence, you can. AI opens up an entirely new way to test. It’s exploratory not confirmatory. In other words, instead of confirming you have a good idea, you’re exploring all your ideas. Instead of trying a new button color versus an old one, with AI, you can try five button colors, a handful of images, new headlines, removing elements, switching fonts, changing header size, and a whole lot more, all at once. Genetic algorithms test each idea, find the promising ones, and then evolve your site by combining the best performing changes with each other and testing for generation after generation. In other words, AI allows you to ditch your bad ideas faster and find not just your best ideas, but the best combination of your smartest ideas.

And exploration beats confirmation every time. After all, if you only have one good idea, the AI will find that. But chances are, you have tons. Using our Sentient Ascend product, we’ve seen customers discover double-digit improvements while testing thousands–or even hundreds of thousands–of potential sites at once.

The question for marketers then becomes: do you know everything about your customers? And if you don’t, how fast are learning what they want? Are you exploring a vast search space of good ideas or just confirming one or two? And what might happen if you decide to make artificial intelligence a part of your strategy? If folks like ABUV Media and Cosabella are an indicator, you could be seeing 30-40% conversion increases. And that’s the sort of thing that makes a real difference.