Seeking advice for using Google Analytics to analyze and test LP variations

  • 13 March 2018
  • 1 reply
  • 5 views

My A/B tests are unlikely to achieve statistical significance in a reasonable amount of time if I am only looking at conversions, so I would like to take engagement metrics like avg. duration and bounce rate into consideration. I’m seeking advice on the best way to do this.

My thought is to create a custom report in GA, but I’ve never done this and I’m having difficulty with it. If anyone could point me towards a helpful article or something, that would be great!

Also, any other tips or ideas on how to

  • determine a winner when you have relatively low traffic

  • set up automation so that you receive a report on the regular

would be greatly appreciated.!


1 reply

Userlevel 7
Badge +3

Hi @baltzer,

You should try to always calculate the needed sample size before starting a test. Statistical significance shouldn’t be used as a stopping rule.

Otherwise, you can end up in your current situation “fishing” for successful metrics after the fact.

Changing your KPIs after a test is not advisable. Even more so with engagement metrics like avg. duration and bounce rate.

I can show you numerous landing pages with a really high bounce rate (ex. 80%+) that still convert really well. Also, there is the issue of how would you calculate statistical significance for a metric like bounce rate or avg. duration?

If you don’t have enough traffic, I would concentrate on gathering more data and forming better hypothesis for your tests.
Another option would be testing bigger/bolder changes on the page. The more you increase the expected lift in CR, the less traffic you would need in general.

Now finally to answer your original question… if you are using the native GA integration in Unbounce, you can create a custom report and add a few segments. The segments should be setup to look for your landing page variations. 1 segment per variation.

Hope this points you in the right direction.

Best,
Hristian

Reply