The Anatomy of Landing Page Testing Using AdWords


Notice: Undefined variable: post_id in /home/foundco/public_html/wp-content/themes/pivot-child/inc/content-post-single-sidebar.php on line 48

Landing Page Testing from a PPC Perspective

(We previously covered A/B or multi-variant testing using website optimiser here)

Imagine you want to find the landing page with the highest conversion rates or EPCs (earnings per click)  from a website that you don’t have access to edit yourself. One option is to use AdWords Campaign Experiments (ACE). AdWords Experiments seems to have been in “Beta” ever since it was launched over 2 years ago, however that doesn’t mean it isn’t useful. Firstly the Don’ts:

The Don’ts:

  • Don’t pick two different keywords and give each one a different landing page URLs, then compare conversion or revenue data sometime later. You should not do this because different keywords have different conversion rates. How do you know whether the difference in conversion rate is because of the different landing page, or because of the different keyword?
  • Don’t split test landing pages over different periods of time. You should not do this because conversion rates and EPCs also vary over time depending on a multitude of known and unknown factors.

My Solution is as follows:

  • Pick an Adgroup you want to use for the experiment.
  • You’ll then need to make a complete copy of the Adgroup that you want to use, with the same ads and bids except with different destination URLs changed prospectively to the two landing pages you want to test.

(You now have a Control Landing Page on the original Adgroup, and the Experiment one on the copy).

  • Give the experiment Adgroup some traffic for one or two days so that all ads are approved and the quality scores are the same as the original ones, then set up the Adwords Campaign Experiment.
  • To set up the Experiment you’ll have to go into the Setting tab and expand the Experiment section under ‘Advanced Settings’ right at the bottom. Click on ‘Specific Experiment Settings’, and give the Experiment any name you want.
  • On the Control/Experiment Split ideally you would use 50%/50%, however if the test landing page is very experimental and you don’t want a risk a drop off in conversion you could make the experiment percentage smaller.
  • As for the period of time, it doesn’t matter as long as you have statistical significant data, set it to 30 days and extend or end depending on whether you have enough information.
  • Once the Experiment is activated you’ll have to set the Control Adgroup to “Control” and the test Adgroup to ’Experiment’. You can do this by going to the Adgroup tab and selecting “Control” or “Experiment” under the menu that you usually use to Enabled, Pause or Delete Adgroups.

Now all searches should be split between the Adgroups depending on your Control / Experiment split choice during set up. Note that all other Adgroups in the campaign should be set to “Control and Experiment”.

From there it’s up to you how and when you decide you have enough data to say one is better than the other. You can do it with a little help from some statistical hypothesis testing. For example, a Chi-squared test can be used to compare conversion rates as found here.