Jason Puckett

May 15, 2016

One Multivariate Ad Test to Rule Them All

Multivariate ad testing is hard. Multivariate testing in AdWords requires spreadsheets, variable tracking, lots of excel knowledge, some Vlookups and a really good knowledge-base of ad testing fundamentals. Multiply that across a multitude of accounts, ad groups or campaigns and pretty soon a true multivariate test is unachievable within the confines of your day-to-day workload.

This article outlines a nice and easy multivariate test that can be executed in even the most slammed of schedules. It’s simple enough to understand, execute and iterate upon but complex enough to actually yield the fantastic data that multivariate tests typically do.

Dare we say it? The One Test to Rule Them All.

One Test to rule them all, One Test to find them, One Test to improve them all and reliably optimize them.

NERD ALERT! Pshhh. If you’re nerdy enough to think the above parody is funny, you’re probably nerdy enough to appreciate how awesome this test is. And that means you’re pretty cool in our book.

Brodo Swaggins

Ok, what is this test?

It contains six permutations of ad creative. 3x Description combinations and 2x of whatever other element you’d like to test.

Here’s how to execute:

  1. You will be testing 3 different versions of description line 1 and 2. This is not optional, this is a requirement, so go ahead and write them.

  2. Select whether you’d like to test your Headline or Display URL. This is essentially an A/B test paired with an A/B/C test (your descriptions). Sometimes Headline isn’t an option because it is unique to your ad group and you don’t want to lose the quality of your search term alignment. We get it. Go with Display URL.

  3. Create your 6 ads. Each of your Description combinations paired with both of the variables you’ve chosen between Headline and Display URL.

Does this not make sense? Here it is visually:

A/B Testing Headlines + A/B/C Testing 3 Versions of Description Line 1 & 2

Testing Headlines

A/B Testing Display URL + A/B/C Testing 3 Versions of Description Line 1 & 2

Testing Display URL

Ok, this is really simple - why is it so good?

It’s good because it’s simple! There are two main reasons why people are too afraid/lazy to execute multivariate tests on their search ads.

  • Too many ads per ad group - this is a legitimate fear. This 6 pronged test is only going to be yielding six new ad creative variations per ad group. From our experience (1000’s of ad tests, 6 new ads is not a bad thing). Don’t be scared. This is even great for low volume accounts

  • Too many elements to track - Six ad variations is not too many to track. Luckily you are only looking at 3 Description combinations and 2 variations of another element. It will be very easy to perform regression (which elements make the biggest difference) even just by eye-balling it.

Test Duration

This is dependent on the size of your account and any time constraints you are facing. We recommend running tests for between two and three weeks. If you can cross the 90% confidence threshold within that period, make a decision and move on to your next test.

If you’re making your decision based on CTR, we recommend an absolute minimum of 100 clicks per ad variation (you will need 600 clicks to execute this test). If you’re a larger account and get that per-variation volume in less time, go with the 2-3 week window.

Questions? Comments? Please let us know at Hello@AdBasis.com

Have a question for us? We'd love to hear from you!

comments powered by Disqus


Popular Articles

Understanding Conversion Layers and Executing Ad Experiments

August, 14th 2014 - Optimization

3 Steps for Building a Paid Acquisition A/B Testing Plan

October, 15th 2015 - Optimization

All Things Tested - 3Q Digital's Sean McEntee

August, 31st 2015 - "All Things Tested"