Finding Problems In Your Experiment Installation - A/A Testing

Troubleshoot A/A Testing

A good way to understand if your site has all the tracking code installed properly is to run an A/B experiment on all traffic on all visitors for all pages but without any real variation.

The idea is to see if your total traffic and all goals match between you Convert Experiments installation and your analytics system.

When you include all pages and the entire audience the numbers between the systems should match up closely and it's a great way to setup and verify all goals.

We found most differences in numbers appear because of not tagging all pages, having a audience or site area definition that excludes some numbers of having the same custom field (slot) for multiple experiments.

So setting up a simple A/A experiment goes as follows:

  • Select and A/B experiment
  • As variation for example remove a space or dot in the footer where nobody sees it.
  • Continue to the summary
  • In Audience select all visitors (removing any targeting)
  • In Site Area target your entire site by changing the target group: Page URL, Starts with http:///www.domain.com for example
  • Run the experiment

This should be including 100% of all the visitors and will allow you to see if there are visitors missing (usually missing Convert Experiments tags).

Have more questions? Submit a request

Comments