Multi ad group testing using text patterns

Modified on Fri, 27 Dec, 2024 at 10:49 AM

Multi ad group testing using text patterns helps you to split-test sets of ads across multiple ad groups or campaigns. Each set shares a common text patterns in headline 1, headline 2 and so on, which is matched in the same location for all ads.

This approach can be useful for testing how well specific messages perform. It's also useful for low-volume accounts that don’t accrue enough data to reach statistical significance with single ad group testing.

How to run multi ad group tests with labels

Manual tests

You can run multi ad group tests as often as you need. Go to Ad testing > Multi ad group > Test by text patterns, then click Find current winners.

Select your campaigns and the date range for this test. A common date range means Adalysis will calculate a date range during which all ads within the test were active.
Specify the location within the ad where the pattern should be matched. One location applies to the whole test.
Specify the pattern you want to test. Patterns can be whole text, or use wildcards (e.g. Find a plumber in *). Click the folder icon to see the most common patterns across your chosen set of campaigns and locatio
Click Find current winners.
If you want Adalysis to run this test automatically, click Save & run this test daily.

Important: Please be careful when using a common test date range. If one matching ad out of hundreds was enabled only a few days ago, it will cause the whole test to run using only the last few days of data.

To familiarize yourself with this approach, click Run a sample multi ad group test for a sample test result from one of your campaigns.

Daily test runs

Unlike single ad group testing, you'll need to define your multi ad group test before Adalysis can automate it for you. You can save your manual tests by clicking Save & run this test daily or choose Define an automated test.

Understanding the test result data

Once you open the test result details, you will see the aggregate performance of each pattern (over the date range used) and the confidence the algorithm has in each of the metrics. You'll also see performance boost projections based on pausing all ads with the losing pattern(s).

Bulk actions

Pausing losing ads

Click the Pause button to pause all ads with the losing pattern for a specific metric. The test is then marked as archived and you can always find it under the Archived test results tab.

Pausing and replacing losing ads

Click Pause & create new ads to pause the losing ads and automatically create replacements. You can specify a new pattern to replace the losing text in your replacement ads. Preview new ads lets you check your ads before going ahead.

Please don't use wildcards in the replacement patterns. For example:

The original test pattern includes a wildcard, e.g. Find a plumber in *. This will match ads like Find a plumber in Florida or Find a plumber in LA.
Your replacement pattern should match Find a plumber in, e.g. Call a plumber in.
The new ads will be created using the full text Call a plumber in Florida or Call a plumber in LA. The wildcard text in the old ads is preserved.

How pattern matching works in multi ad group testing

Pattern How ads are matched to the pattern
abc abc exists exactly as is (with no additional text).
* abc abc exists at the end of the text.
abc * abc exists at the beginning of the text.
abc * xyz abc exists at beginning and xyz at end, with any text in between.
* abc * xyz abc exists anywhere before xyz. xyz exists at end.
abc * xyz * abc exists at beginning. xyz anywhere after abc.
!!abc abc doesn't exist as is (or use a wildcard * before/after abc)

Related content

History of multi ad group test result changes

Pattern	How ads are matched to the pattern
abc	abc exists exactly as is (with no additional text).
* abc	abc exists at the end of the text.
abc *	abc exists at the beginning of the text.
abc * xyz	abc exists at beginning and xyz at end, with any text in between.
* abc * xyz	abc exists anywhere before xyz. xyz exists at end.
abc * xyz *	abc exists at beginning. xyz anywhere after abc.
!!abc	abc doesn't exist as is (or use a wildcard * before/after abc)