Unbounce Answers/FAQ/A/B Testing

How do I run an A/B test?

Oli Gardner
posted this on October 21, 2010, 5:22 PM


Check out our video in the Getting Started Guide on Running A/B Tests »


A/B testing has the ability to make marketing super fun. Instead of arguing over messaging or creative – try more than one idea simultaneously and turn it into a contest.


Step 1 – Create a Page Variant


Starting with a copy of your existing page is the easiest way to start a test.

The basics of A/B testing are that you run different versions of your page – called variants – side by side with a series of changes in each variant. The idea is to see which one converts the best.

You can run as many variants in your test as you like (e.g. an A/B/C/D/E/F/G… test) but the more variants you have, the more traffic you need to send to your campaign to be confident that the statistics are not a coincidence.

To create a variant:

  • Open a page from your Pages tab
  • Click one of the two big buttons in the A/B Test Centre to duplicate an existing variant (easiest method) or create a variant from a blank page

That’s it! You now have 2 page variants  that you’ll use as the basis for your test.


Step 2 – Decide WHAT to Test

Now we need to decide what you’re going to change on your new page variant. Here are a few suggestions of page elements that are worth testing:

  • The primary value proposition or campaign message
  • Your hero shot – the main photo or product shot
  • The body text that describes the details of your offer
  • Your bullet points (features & benefits)
  • Change the wording of your primary Call To Action (CTA)
  • Change your lead gen form

You can change just one thing or change many at once to create a completely different page.

Changing many things at once might produce a more dramatic change in conversion, but will provide you with less causal information than just trying one thing at a time. It’s totally up to you.


Step 3 – Decide HOW to Change Your Test Elements

Now that you’ve decided which page elements you want to test, you need to think about *how* to change them to give you a conversion lift.

Looking at the list of elements in Step 2, you could try some of the following things:

  • Make your lead gen form shorter (ask for less information)
  • Give your lead gen form a more prominent position (higher up the page and try it on the right-hand side)
  • Simplify the page with less text, or separate it into bullet points for greater clarity (as an exercise – try removing 50% of your page copy)
  • Make the CTA bigger
  • Add a directional cue (such as an arrow) that points to the CTA to draw attention to it

Successful manipulation of emotional and physical response is your key to conversion success. By utilizing some principles of design you can start refining your pages.

For more design ideas check out this blog post: Designing for Conversion – 8 Visual Design Techniques to Focus Attention on Your Landing Pages


Step 4 – Watch the Stats and Promote a Champion



Now that you’ve created one or more new page variants you need to start the test. To begin your A/B test follow these steps:

  • Assign each variant some traffic (weight). To do this, click on the 0% beside your variant (in the weight column) in the A/B Test Centre, input the % weight you want to assign that variant, then press enter to confirm the changes. You can assign as little or as much weight as you like. Note: the more traffic you give it the faster you’ll gain confidence that your test results are valid. For a test with one Champion page (the original one) and one variant, you can just try doing a 50/50 split.
  • Publish (or re-publish) your landing page
  • Start driving traffic to your page. Note that you will only ever have one URL for this page – Unbounce randomly selects which page a visitor sees based on your traffic weighting.  

    Note: When a visitor first accesses your page, we deposit a cookie on their machine to track the view and the variant they saw.  As long as that cookie is present when the visitor returns, they will continue to see the same variant.  The same goes for you visiting your own page.

  • Watch the confidence column to see how statistically valid the test is. The more traffic you drive to your page, the faster this number will climb. Once the confidence is above about 90% you can be more confident that the conversion results you’re looking at are valid.
  • At this point if your test variant is outperforming the Champion, (higher conversion rate) you can promote the variant to become the new Champion page! From the Page Overview screen, promote a challenger variant by clicking on the gear next to that variant and then choosing "Promote to Champion" from the drop down menu.

    Note: your old Champion will now be discarded (but not lost) and you will have to begin the testing process over again to see if you can make it even better.
  • champion_variant.png


User photo
Brad Lakin

In my Actos campaign I have 4 variants but for some reason the Actos Final 4 variant is not getting its 1/4th share on the impressions.

December 29, 2011, 10:03 AM
User photo
Dan Croxen-John

In setting up a test of a landing page (existing vs. new), how do I measure the difference between the performance of the two pages. Can I set up a conversion trigger, like in Google/Visual Website Optimiser? What could this conversion trigger be? A click-though to the next page, an order, a completed enquiry page, etc. Would be good to understand how you can compare the performance of your two (or more) landing pages?



January 24, 2012, 6:41 AM
User photo
Matt Roberts

Is it possible to let unbounce automatically optimise the display based on the best performing variant?

August 20, 2014, 10:08 AM
User photo
Quinn Omori

Hi Matt - The app can't promote a challenger variant to the new champion automatically. It's unlikely that one variant will get all the way to a 100% confidence score and while anything above 90% should be more than enough to justify promoting a challenger variant, we erred on the side of caution and made the actual promotion a manual process, rather than having it trigger after a certain threshold.

August 20, 2014, 3:28 PM
User photo
Thom Duncan

It would be helpful to add instructions or a link to the instructions about how to promote a challenger to champion. It's the last bullet point on this page.

October 16, 2014, 9:31 AM
User photo
Quinn Omori

Hi Thom - Thanks for the feedback. This article needs a larger overhaul, but in the interim, I've added a note on how to promote a challenger variant and added a screen cap to help clarify.

October 16, 2014, 11:21 AM