Prepare
A/B testing is vital to your bottom line because it can have a huge impact on conversions. If a landing page works better as a result of an A/B test, you can generate more leads. If an ad is improved, new site visitors and transactions can increase. Not to mention, A/B tests are low-cost or free, depending on what you are testing. Before running the test, make sure you know what outcome you are looking for. Do you want more click-throughs? More subscribers? More revenue? If desired results are not identified in advance, there will not be a way to tell if the test was effective.
On your mark
Like the pentathlon, A/B testing is a process. A standard A/B test compares a control (A) with a new variation (B). When conducting an A/B test, only make one change and evaluate the results. If multiple changes are made, it is impossible to tell which of those changes impacted the results. Other types of tests can compare multiple variations at the same time, but that is a tale for another day. Before implementing an A/B test, make sure A/B will be the best way to accomplish your goal.
Get set
Now you're ready to start testing. First, use your analytics data to determine which areas could be improved upon. These areas could be on-page (a website), or off-page (an ad, email, etc.). Next, create a hypothesis about the variation. Would Content A lead to more conversions than Content B? Would the CTA be more effective if it was red or if it was blue? You can test content, landing pages, campaigns, ads, images, buttons, etc. There are so many different variations to test that it can be a bit overwhelming. A good game plan is to start by running tests at each stage of your marketing funnel.
Go!
Just like in swimming or sprinting, timing is everything in an A/B test. For example, when testing variations of an email, split your address book into two groups and send the two variations at the same time. You never know what could change for a user day-to-day or hour-to-hour. This split-test will help eliminate the possibility that other factors influenced the results. For other tests, like changes on a website, the test should run long enough to provide enough data to reach statistical significance. For accurate results, run both versions at the same time, for the same amount of time.
Practice makes perfect
No one makes it to the Olympics without thousands of hours of practice. Like those athletes, your testing will also improve with repetition. The more tests you run, the better you will get at predicting what changes may be most effective. If your B variation does not result in an improvement, use that data to formulate a new hypothesis to test. For more inspiration, check out this list of 50 A/B tests and the results.
Game rules
To help avoid a red card, Google Webmasters has some do’s and don’ts when testing variations of a website:
1. Don’t use cloaking: Show the same content to Google robots as you do to humans.
2. Do use rel=canonical: Use this on your variation URLs to show that the original is preferred.
3. Don’t use 301s: Use a 302 instead, making your redirect temporary instead of permanent.
4. Do keep it short: If Google thinks the test is taking an abnormally long time, the site may be penalized.
Constantly striving for improvement is key to success. ParadoxLabs is always here to provide development and support to make your business even better.