Fail Fast and Learn

Learning to Embrace the Times When A Beats B

If you want to increase your success rate, double your failure rate.

Thomas J. Watson, Former Chairman and CEO of IBM

After what we have been showing you, it may be tempting to think that every experiment you run will have a positive outcome. (Why, then, even run the experiment? Just keep changing the thing that you were going to test, and it'll be great!) The reality, of course, is that not every variation is going to beat the control.

Even “failed” experiments have their silver linings, however: recognizing that a particular change will harm your goals is inarguably better than simply making that change, and as an added benefit experiments like this are often the ones that teach us the most about our visitors and what drives them. Indeed, experiments that fail tend to contradict some assumption held by the tester, and the results can point to reasons why that assumption is wrong.

Prime Real Estate versus Familiar Real Estate: IGN

Gaming website IGN wanted to encourage more visitors to the video site that brings them a big portion of their ad revenue. So they tried running an A/B test where they moved the “Videos” link over to the left of the main navigation (Figure 6.1).


FIGURE 6.1 Original IGN navigation versus variation with repositioned “Videos” button.

There are plenty of organizations in which ...

Get A/B Testing: The Most Powerful Way to Turn Clicks Into Customers now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.