Better RSAs Start With Ad Copy Testing
Let’s start with the numbers: the vast majority – 85%+ – of PPC accounts I’ve audited are not doing any structured, ongoing ad copy testing. Part of this is due to the (Google-perpetuated) notion that Responsive Search Ads (RSAs) do this for you (they don’t!), and part of it is that it’s usually one of the last things on any advertiser’s mind.
I think this is a HUGE mistake. Your ad is one of the most critical components of your ad account – it is the first thing a searcher sees from your brand, and as such, it plays a disproportionately large role in whether or not a searcher will turn into a site visitor. You can have the best product, landing page and user experience in the world, but if you can’t convince your audience to learn about it, you’re going to go the way of TiVO.
Put simply: poor ads result in lower CTRs, lower Quality Scores (QS), and higher CPAs/Lower ROAS – all of which negatively impact your business’s (or your client’s) bottom line.
The agenda for today’s issue is pretty simple:
- Ad Copy Testing in the Era of Automation
- My 3 Go-To RSA Testing Strategies
- Getting Started with RSA Ad Copy Testing
Ad Copy Testing In the Era of Automation
There is ample data that RSAs dramatically under-perform their predecessor Expanded Text Ads (ETAs) in most metrics – this was published in a 2022 study by Adalysis, and I can confirm that the results of this test hold true across our portfolio of accounts as well.
Part of this is that Google has a different set of objectives than advertisers – most evidence suggests that Google weighs CTR heavily in RSA optimization, which is why you might see an RSA with a higher CTR getting 5x or 10x more impressions than one with a lower CTR but a higher CVR / lower CPA.
Part of this is simply the number of possible combinations for an RSA: a 15-headline, 4-description, unpinned RSA contains up to 47,280 asset combinations – which means optimization can take a LONG time and a LOT of impressions – upwards of 4.4M if each combination is given 100 impressions (which is far fewer than the number required for statistical significance in most cases). The simple reality is that most ad accounts – let alone most ad groups – will not serve 4.4M search impressions in any reasonable period.
And part of this is that there are certain patterns to both ad copy (novelty + variety) and ad structure (Bold Statement – Supporting Stat/Benefit – CTA, for instance) that tend to perform better than others (for instance, a CTA – CTA – Benefit ad is unlikely to perform well for any brand).
Ultimately, if you want higher-performing search ads, you’re going to need to help the machines out by either (a) using pins to reduce the number of combinations possible or (b) running structured RSA tests.
First – Determine The Questions To Answer
One of the hidden benefits of paid search is that it can be a goldmine of insights into a given customer base – especially if you are consistently doing ad testing. While this won’t replace all third-party research or customer insights work, it can be an extremely valuable supplement – allowing you to generate quick wins for your entire marketing organization, while improving your overall PPC results.
Some questions you might want to answer:
- Which decision triggers (credibility, discovery, appeal, detail, price, scarcity, usability, quantity) are most important to your target audience?
- Does your audience have a local preference?
- Does shipping matter (free / rush / free over $)
- What CTAs perform best?
- Which benefits are most compelling to a specific type of searcher?
- Does comparative language perform better than brand-specific language?
- Does prequalification help final conversion rates (great for B2B)?
- Does price (discounts) or value (free gift with purchase, total value) matter more?
- What is the best way to enhance credibility – reviews, awards, press or ratings?
Once you’ve clarified which questions you want to ask, the next step is determining the right testing setup to get those answers.
RSA Testing Setups
Ad Copy Testing in Google Ads comes in two primary “setups” and three “types”:
Single Ad Group Testing – In general, most RSA testing falls into the “SAGT” setup: we run either multiple RSAs or a Variant Experiment within a single ad group. This is ideal for high-volume + branded ad groups, and is especially helpful if you’ve already done some multi ad group testing to surface account-wide insights. Most SAGT can be done using Variants (more on that below).
Multi Ad Group Testing – Better known as text pattern testing, multi ad group testing involves testing a pattern of text (for instance, a H1 or H2) across multiple ad groups within your account. In order for these tests to yield actual insights, it is imperative that you’re consistent in your setup – from using similar text patterns across accounts, to placing the “test” copy in the same position, with relatively similar surrounding text. You can then assess the results of the test by either (a) using a pivot table (if you have very similar text patterns) or (b) using labels (more on that below).
For each of these setups, there are three primary “types” of tests that can be run:
- Running multiple RSAs in a single Ad Group – this is the most basic type of RSA “test” – add another RSA to an existing ad group, wait a bit, then pause the “loser” ad + add another “test”. The issue with this approach is twofold: (1) you have limited control over ad serving (even with “rotate indefinitely” as your ad rotation setting) and (2) it’s incredibly difficult to isolate specific asset performance, so you’re never quite sure what’s working / what’s not. There may be parts of your “loser” ad that actually perform quite well, but (absent significant analysis), identifying that is going to be difficult.
- Testing RSAs via Google’s Experiments – Google Ad’s “Experiments” is one of the most under-utilized features in the platform. It’s wonderful for many things (bid strategy tests, match type tests, etc.), but ad copy testing isn’t one of them – primarily for the same reasons as adding multiple RSAs to the same ad group: (1) it’s difficult to control ad serving and (2) it’s near-impossible to isolate specific asset performance.
- Testing RSAs via Google’s Variant Testing – I absolutely LOVE Google Ads variant testing. It’s easily one of the best optimization features developed for any ad platform, ever. In essence, variant testing allows you to upload multiple versions of an asset (a headline, description, etc.), then have Google dynamically serve either a control or test variant. This enables you to have full control over asset serving AND actually understand performance differences between your original/control asset and your experimental asset.
In general, my strong preference is to test RSA copy using Variants – this provides the most scientific method, and Google’s platform automatically aggregates + analyzes the test data for you. The limitation here is the number of tests that can be run simultaneously, as well as the ability to test slight variations on a theme (Variants allows you to test one headline against another, but if your different ad groups need slightly different text, you’re likely better off running multiple RSAs with pinning and “rotate indefinitely”).
My 3 Go-To RSA Testing Strategies
Over the course of testing 1,000s of RSAs, I’ve developed a three-step strategy for testing RSAs. In many cases (especially for existing accounts with plenty of data), we can leverage existing data to skip step #1 (“Seed + Iterate”) and proceed to structural testing.
- Seed & Iterate – This is the most common starting point for RSA testing; we begin with an unpinned “seed” RSA that includes 12-15 headlines and 3-4 descriptions – including a mix of brand, credibility, value props, offers, CTAs, etc. As this serves, we identify the best-performing combinations (you can view this in the RSA combinations + assets report), and use that to create a 2nd RSA with those assets pinned in their top-serving positions.
- Structural – The next “phase” of testing we typically do is structural. In these tests, we create multiple RSAs, each of which uses pinning to establish a different “structure”:
For each structural variation, we’ll include multiple assets for each theme; once we find “winning” structures, we’ll then move into “specific” testing to determine the optimal copy.
- Specific – The final testing structure we use is specific message testing. These tests involve pinning specific headlines + descriptions, then using variant testing to determine which specific copy works best. These tests are ideal for determining (for instance) if a Geo or Non-Geo message works best, or if a $ discount vs. a % off is the proper way to frame an offer:
Tips for RSA Testing:
- Rate Metrics Are Your Friend: always, always, always use rate metrics when assessing RSA performance. As mentioned above, a “loser” RSA may receive 10x the impressions of a “winning” ad, due to how Google assesses RSA performance (heavily weighted toward CTR). Thus, it’s possible for a “loser” ad to show *more* conversions (by counting stats), simply by virtue of having so many more impressions. Instead, look at the following set of metrics for each RSA:
- Click-Through Rate
- Conversion Rate
- Impressions Per Conversion
- Cost Per Conversion
- Conversion Value Per Impression
- Conversion Value Per Impression / Cost Per Impression
Assuming the data you’re feeding into Google Ads is accurate (and if it isn’t, stop your ad testing + go fix this ASAP!), this metric set will give you a far more comprehensive picture of your RSA performance, and allow you to make far more informed decisions on what to do next.
- Constantly Look At Your SERPs: this should go without saying, but I’m constantly surprised by the number of advertisers (and agencies) that have very clearly never actually looked at the SERPs where you’re placing ads. You wouldn’t buy a billboard on a highway without reviewing the surrounding area, nor would you place a TV spot without looking at the channel lineup during the time slot. Why is it OK to buy an ad on a SERP without ever looking at it?
My recommendation is to search your top 50-100 keywords, in your targeted locations (use SE Ranking’s Location Changer), from a mix of devices (Mobile + Desktop primarily) each week. Yes, it’s going to take an hour. But it’s one of the best hours you can spend to improve your ad copy performance. Look for:
- What are our primary competitors saying?
- Are other groups advertising on our SERPs? If so, who? Why?
- Does our messaging stand out or do we sound the same as everyone else?
- Are our key points of differentiation relevant to this searcher?
- Do the organic results have more compelling headlines? If so, what in particular is compelling?
- Are there other SERP features that reduce the prominence of our ads?
- What questions are populated in the PAA section? Does our LP address these?
- Ad Strength =/= Quality Score: Ad Strength is easily one of the most misunderstood “metrics” in Google Ads – and one that frequently gets in the way of proper RSA testing. I’ve had Google Reps tell clients Ad Strength contributes to Quality Score (nope!), and that they should cease their ad testing + pinning because “Google knows best” (also nope!). Finally, there’s not a strong relationship (in any data set I’ve seen) between Ad Strength and Conversion Rate, ROAS or CPA.
At the end of the day, here’s what I advise every client to do:
- Ignore Ad Strength – it’s just a measure of Google’s optionality. All it does is indicate the relative diversity + possible number of combinations available to the machine. You are writing ads for *people*, not for machines.
- Do Right By Your Business – ultimately, your Google Ads Account needs to work for your business + your customers. If you’ve tested and found that pinning results in better outcomes, then pin stuff. If you’ve found that running a few “effective” ETAs (i.e. an RSA with 3 pinned assets) works better, then do that.
- Continue to Learn – even if you have a solid formula in place, continue to test your ad copy.
- Make An Ad Tracker: I am a HUGE fan of creating an “Ad Tracker” or “Ad Repository” where every single asset I have in an account can be tracked + assessed, as well as where new ideas generated from (2) above can be added for client approval. Here’s a screenshot from one Ad Repository Document (headlines only) – if there’s enough interest, I’ll put together a template + share it out with the next issue. Just let me know.
- Status = whether the asset is live, approved/rejected, in review, pending review
- Type = headline, description, long headline, etc.
- Theme = the general grouping of the content – branded, CTA, value prop, bold statement, competitive positioning, pricing/value, review, etc.
- Content = the actual content for the asset
- Length = pretty self-explanatory
- Series = is the asset part of a series of pinned assets?
- Strength = relative strength of the asset’s performance. This can be broken down into the metrics above, but I prefer to have a single value to make it easier to assess
- Notes/Comments = additional information that may be relevant – for instance, flight dates, inspiration for the asset, sources of information contained (helpful for legal review), etc.
- Use Labels For Your Ads: Labels make aggregating data from across your account remarkably easy – provided they are created in a thoughtful manner. You are free to add multiple labels to a given ad. A few helpful examples/systems I’ve used:
- Theme/Concept: if you are testing RSAs thematically, create a system (and include it in your ad tracker). This will allow you to aggregate + compare performance of ads in one theme vs. another theme.
- Inclusion: Related to theme, I often will add “geo” and “non-geo” to ads based on whether or not the RSA contains a geo-modified headline (for instance, “Voted Baltimore’s Best Plumber” or “10 Great Locations in NYC”).
- Persona: If you have ads tailored to specific audience segments (for instance, the “value seeker” vs. the “big spender”), include those as labels.
- Offer Type: We’ll frequently test different offer types – % off, $ discount, Free Gift with Purchase, Exclusive Access – across different campaigns or ad groups. Including relevant labels makes it easy to group ad performance by the offer type, which can then inform your offer strategy moving forward (for instance, if you see that $ off consistently outperforms % discount, you can adjust future promotions to be $ off)
- Winner / Loser: Pretty obvious, but if you’re doing structured testing, you should 100% have labels for “winner” / “loser” / “testing” – not only does this help you stay organized, but it also serves as a helpful reference point when someone (inevitably) asks you to test a specific ad concept – you can easily go back + find the ad.
- Your CTAs Should Be Better: This is one of my personal pet peeves – incredibly well-thought-out ad copy in the H1 + H2, followed by a generic “Shop Now” or “Click to Learn More” CTA in the H3. In general, be as specific, actionable and emotive as possible in your CTA:
- “Design Your Dream Kitchen Today” vs. “View Kitchen Models”
- “4 Spots Left – Don’t Miss Out” vs. “Register Today”
- “20% Off Ends Tomorrow” vs. “Shop Now”
- “Make A Difference – Volunteer!” vs. “Sign Up”
- “Claim Your Free EBook” vs. “Learn More”