Why Every Ad Copy Test You’ve Been Running is Wrong

Richard Ingilby
Reading time: 7 minutes
25th April 2019

Now that I’ve personally insulted you and your profession I’ll be slightly less dramatic. Testing copy in PPC is a bit of a poisoned chalice, and it’s likely that you’re drawing false learnings from it. With a sample size of thousands, and even millions, of eyes up for grabs, the potential for discovery is huge.

However, it is for the most part, done very badly. Ad copy tests can inspire fear among PPC specialists because it’s often seen as a tedious task in this era of increasing click-and-go, dynamic optimisation search that we’re hitting. 270 characters (300 if you count paths, which isn’t even considering ad extensions) is a lot of room to play with.

We’ll start this article by running through the key difficulties and misuse of ad copy. We’ll then move on to how to test it effectively and accrue real business benefits.

How you may be doing it wrong:

Basing Ad Copy Testing on CTR only

If you’re only looking at click-through rates when testing ad copy, you’re doing it wrong. Anyone can serve an ad that will get people to click through more often, but does this really help your business objectives? If you’re serving ads that are fundamentally misleading—as in, they misrepresent your product or offering—you’re wasting money on worthless clicks.

Consider the advert below:

If I advertise that a product is “cheap” in my ads, then yes, it’s very likely to drive more clicks. However, my client is a family jeweller who value quality and handmade local products. Are people going to convert when they realise that my products are ten times more expensive than anyone else? Probably not.

If you’re a premium brand, let people know this before you’ve paid for their click. In the below advert, I’ve used language and price extensions to signal our higher quality and price range before we’ve paid for a click.

Whilst CTR is likely to be lower, I know I’m driving higher quality, relevant traffic that’s much more likely to convert. You should measure ad performance based on a variety of factors, including Conversion Rate, ROAS, Pages/Session and Bounce Rate. These are all good metrics to look at alongside CTR.

This is particularly true of Responsive Search Ads, which Google have even suggested often lead to a lower CTR. However, they also often obtain a higher number of incremental clicks because the dynamic ad copy makes you more eligible for a large number of extra auctions. CTR isn’t king anymore.

Drawing conclusions from account-wide analysis

Often, you’ll look at how ad copy performs at an account, or campaign level, such as the below test with three ad copy variants:

Variant Clicks Impressions CTR ROAS
Variant 1 2,024 44,955 4.50% 1.3
Variant 2 1,394 37,149 3.75% 1.2
Variant 3 5,300 122,324 4.33% 1.5

However, this overview is made completely meaningless by the fact that different ad groups, and different keywords within them, will all have completely different click-through rates. These have as much of an effect on the result as the copy itself:

Ad Group Variant Clicks Impressions CTR ROAS
Ad Group 1 Variant 1 810 14,038 5.77% 2.3
Variant 2 418 7,670 5.45% 2.7
Variant 3 4,240 82,892 5.12% 2.1
Ad Group 2 Variant 1 1,012 30,917 3.27% 0.9
Variant 2 976 29,479 3.31% 0.7
Variant 3 1,060 39,432 2.69% 0.4

In the above, we see that Variant 3 is among the top performers overall because it’s benefitting from the large majority of its clicks being in the naturally stronger ad group. However, it is actually the worst copy in both ad groups.

Dubious Results

Some PPC Analysts reject the above because it takes more time to do and often clouds results. Whilst there’s a clear winner in the top example, you frequently get a different winner for every ad group, with no recognisable patterns in the bottom example. Moreover, measurement is often unreliable because the winner keeps on changing.

There was a fantastic analysis at a PPC Hero talk in 2017 that tracked an ad copy test over time. It highlighted that one variant became a statistically significant winner for a few days before swinging back to dead even, to the other copy becoming visibly better before the whole process repeated itself. The results literally depended on when you measured it.

We’re lucky in that, at Ayima, we’ve built an ad analysis tool that uses Bayesian Statistics. It takes into account various factors such as time and measuring at an adgroup level to remove a lot of the uncertainty from ad copy testing. For other analysts, however, this may be a far more difficult ask.

Modest Incremental Gains

Part of this uncertainty comes from the very modest direct gains to be made on PPC performance from testing ad copy. As advertisers, we often believe our creativity and ad copy has more of an effect than it does. The fact is that people scan our copy for sheer milliseconds and judge relevance within a fraction of a second.

If you’re browsing search results, you’re going to be swayed less by a clever turn of phrase and more by sheer relevance to what it is you’re searching for. Since Googlers are generally making very rapid decisions based on perceived relevance, there’s very little in ad copy that gets noticed beyond Headline 1.

Don’t get lured into the trap of testing for the sake of testing. If there’s little likelihood that a change of adjective in description 2 is going to affect performance, push back and produce something else to test. For all of the above reasons, this approach is more likely to do more damage than good.

This isn’t to say that we shouldn’t test ad copy. Ad copy testing can be incredibly powerful if utilised effectively, and we’ve laid out our top tips for testing ad copy in 2019.

Our top 8 tips for testing your ad copy:

#1: Get your house in order. Make sure you understand the basics, then worry about the incrementals. Your headline 1s should be relevant to the specific keywords and work on quality score. Make sure there are three ads in each ad group. Have a relevant Path 1 and 2. Then, and only then, should you start worrying about testing.

#2: Have a plan. Don’t just test for the sake of running a test. It wastes your time and your client’s money. Know what you’re going to test and know what KPIs you’re using to measure them. Here at Ayima, we build testing dashboards to help us and clients plan and conduct tests that will help not just PPC, but all marketing channels

#3: Test things that will make a measurable impact. Try ad copy hacking using fundamental human biases such as:

  • Does placing more trust signals in ad copy work?
  • Test placing adjectives in headline 1, where people will actually read them.
  • Try implying something in your paths.
  • Use fewer characters. In a world of split-millisecond viewing, the whitespace around your ad could be more important than the words within them.

#4: Remember that PPC copy has slightly different principles to other channels- relevance to the search query is still absolutely vital, and it’s different to sell to someone already demonstrating purchase intent, but you can still test broad principles.

#5: Measure results at an ad group level, and look at all of the right metrics. If you can, develop a program to help you conduct tests, as we’ve built with Ayima Intelligence. If not, start looking more in depth manually. If not, your insights could be miles from the real picture.

#6: Test different formats. Pit Responsive Search Ads against the standard Expanded Ads. Use your time to add the highest value possible for your clients.

#7: Test ad extensions. Through 2018 and now into 2019, Google appears to be serving these more and more to give people additional information on the search page before clicking through. Also, sitelinks are far more noticeable than description 2 on a search page. With the exception of Headlines 1 and 2, this should probably be your next highest priority by value added.

#8: Use your learnings. Actually apply the ad copy tests. Work with the client and test the more effective messaging in other channels, on their website. PPC copy tests have fantastically large sample sizes, so it’s about time we start to use them effectively.

In Summary

There are fewer testing grounds more powerful than PPC for ad copy. With its millions of easily accessible and cheap impressions and clicks, plus the ability to monitor these very granularly, it’s a quick tool to effectively gather broad learnings on which indicators in your copy work for your brand.

However, there are several reasons why you might not be using this to its greatest possible effect, and even more pitfalls to avoid when measuring it. By following our 8 top tips for success set out above, you’re well on your way to getting the most out of your PPC efforts.

Written By Richard Ingilby
Asset 1 Asset 1 Asset 3