Spoiler alert: though it might seem magical, it's just pure math and science.

When people hear that website experimentation and optimization can be done up to 10x faster, their first reaction is astonishment. When the shock goes away, they start to question our math.

In this article, we'll explain that there's no magic behind Juo's ultra-fast decision-making engine. Only science (which we believe is much better).

Need for speed

Time is money, and in the ecommerce world, both time and money fly fast. Very few can afford the luxury of waiting months to gather enough data to make a decision, as the standard A/B testing process takes up to several months (depending on the traffic size). That's ages!

Speed aside, think about what it means for business: so much can happen in these months. Seasons change; different things might work on Black Friday and on Christmas, but you're still running the same test. You might even get a better idea to test, but until the current one is completed, your hands are tied. That's frustrating.

Running a CRO agency, we conducted a lot of traditional A/B tests for our clients, and guess what? Most of these tests didn't end when they were supposed to end. Business owners were getting impatient and wanted to see the results right away: if one of the tested page versions was performing better; they called it a day didn't want to wait for the test to conclude.

They wanted faster results that would bring them profit right away and didn't care much that terminating the test prematurely would mess up the statistical significance. It's hard to blame them: after all, they want what's best for their business and their customers. If something works, it works - no need to get into scientific debates.

The holy grail of experimentation

We asked ourselves, what if we could give them both? Quick results while maintaining the statistical significance.

We looked for a method that would allow quicker testing and delivering business value faster while paying respect to math principles. And there it was, sequential analysis. And that's how Juo's engine was born.

We had a holy grail of experimentation: much faster results that are still accurate.

So how it works?

The basic principle remains the same as in traditional A/B testing: you take two versions of the given page, split traffic between them, and see which one performs better. The main difference is under the hood.