TPE #87: how to A/B test ad copy with Responsive Search Ads

Read time: 2 minutes.

GM and Happy Easter!

Ad copy A/B testing is one of the most impactful and important tasks for a Google Ads Specialist.

But it’s incredibly hard with Responsive Search Ads, automation, and limited insights.

What if I told you there is another way?

I’ll show it to you in this newsletter. 

Let’s dive in!

RSAs have made ad copy A/B testing impossible.

A few years ago, when (Expanded) Text Ads were still a thing, we’d run very granular ad copy A/B tests with small variations.

But that has become completely impossible with Responsive Search Ads (RSAs):

You add up to 15 headlines and 4 descriptions and let Google do the testing for you.

As a result:

• A few headlines get 70-90% of impressions.
• Most are left untested.
• You get no real performance data.

Here’s an example of the limited insights RSAs give us, even with 100K+ average impressions per Responsive Search Ad (everything just gets labeled as “good” or “learning” lol):

It’s even worse in small accounts.

You need WAY too much data to get statistically significant insights.

This makes ad copy A/B testing completely impossible.

But wait!!

Did you know there is a secret way that allows you to still A/B test copy?

I’ve been using it for years and thought everyone knew about it…

But I was wrong: I hardly ever see people actively using this feature in their accounts.

Let me show you!

How to run ad copy A/B tests with Ad Variations.

There’s a hidden feature in Google Ads called ‘Ad Variations’.

With Ad Variations, you can A/B test specific elements of your ads (headlines, descriptions, even landing pages).

And you get REAL data — plus an indicator if your test is statistically significant.

Here’s how to use Ad Variations.

Step 1: Navigate to Ad Variations.

1.1: Go to “Campaigns”, then click “Experiments”

1.2: Click on “Ad Variations”

1.3: Click the “+”

Step 2: Set up your Ad Variation.

2.1: Select Ads.

•  Choose the right campaigns.
•  Optional: set up a filter (if you only want to test ads with a specific headline, description, final URL or path.
•  Click “continue”.

2.2: Create variation.

•  Choose the type of variation you want to apply to your ads
•  Find and replace: finds and replaces a specific text (see example below where you could test “Buy now” vs. “Shop now”, or anything else like USPs, CTAs, keywords etc.)
•  Update URLs: sets a new URL to test with.
•  Update text: choose what you want to edit (headlines/descriptions) and what you want to test (add/remove/pin).

Find and replace is my personal favorite because it’s the easiest way to isolate ad copy and do proper A/B tests (which is impossible with regular RSAs).

2.3: Set variation details.

•  Add a name.
•  Set a start and end date (we used to be able to run tests without an end date but as of now the max duration is 84 days).
•  Set up your Experiment split (in most cases 50% is perfect).

2.4: Click “Create variation” to launch your Ad Variation.

Congrats, you’ve now launched your Ad Variation! 

All you have to do now is sit back, wait, and gather data.

In the meantime, you can analyze the performance of your Ad Variation test:

When you click on your Ad Variation name, you will get a deep-dive of the data and results of your experiment.

See below for 1: overall performance comparison between your control and experiment ads.

Hover over the data to see statistical significance (if you see a blue star, it means your test is statistically significant for that specific metric).

And 2: see the performance data by campaign/ad group/ad.

You can now A/B test ad copy with RSAs!

You now have a new powerful weapon in your arsenal.

Play around with Ad Variations and start A/B testing.

Again: my personal favorite is the find and replace function because it allows you to isolate tests (which is crucial for proper A/B testing).

Here are a few examples of what you can test now:

•  Landing page tests.
•  Lower case vs. Upper case.
•  USP vs USP.
•  CTA vs CTA.
•  Pinning vs. Unpinning.
•  Keywords vs. no keywords (or DKI).

These are just examples — test literally anything you like.

Hope it was helpful.

Enjoy the rest of your day and see you next week! :)

Cheers,

Miles (& Bob)