Fifteen unpinned headlines and an Excellent Ad Strength rating is how most B2B accounts run responsive search ads. It is also why their headline tests never produce an answer.
RSA headline testing in B2B works when you control one slot and test ideas in the rest. Pin headline position one to the statement that tells the searcher what you sell and who it is for. Group the remaining headlines into a small number of themes, each one a different reason to buy. Then read the result by comparing whole ads, not by staring at the asset report.
This is for marketing leaders and paid search managers at mid-market and enterprise B2B companies who run Google Ads on a thin monthly count of qualified conversions and want their copy tests to settle something.
Why the default RSA setup teaches a B2B account nothing
The default setup optimizes delivery, and delivery is not the same thing as learning. Google's own documentation describes the mechanism plainly. You can provide up to 15 headlines and 4 descriptions, the system assembles them into combinations, and over time it learns which combinations are most relevant for different queries (Google Ads Help, About responsive search ads).
That is a fine arrangement if your goal is the best-performing ad this week. It is a poor one if your goal is to know which argument your buyers respond to, so you can put it on the landing page, in the LinkedIn ads and in the sales deck.
Run the arithmetic on a typical mid-market ad group. Say it produces 12 qualified conversions a month. Each impression shows up to three headlines, and Google credits every asset in the ad with that impression. So those 12 conversions get smeared across 15 headlines, each one sharing credit with two others it happened to appear beside. On average each headline carries two or three overlapping conversion credits a month, and none of those numbers is independent of the others.
Google says as much in the fine print of its asset reporting. Asset-level ratios such as CTR, CPA and ROAS "should be used as directional indicators only," because they "are influenced by the combination of assets served together" (Google Ads Help, About the ad-level asset report).
I read that as Google telling you, politely, that the asset report cannot referee a headline test. It can tell you which assets the system chose to show. It cannot tell you which idea sold.
Ready for paid ads that pay off?
Book your free auditWhat to pin in a responsive search ad: headline one, and only headline one
Pin headline position one to your category-and-buyer statement, and leave everything else unpinned. Headline one is the part of the ad that decides whether the right person clicks. It is where the searcher checks that you sell the thing they typed and that it is built for a company like theirs. If that line rotates, every other result in the test is contaminated by the question of who clicked.
Pin two or three distinct versions of that statement to position one, not a single line. For a composite contract management company, the pinned set might be:
→ Contract Management Software
→ Contract Tracking for Legal
→ Built for Legal Teams of 20+
All three do the same job, which is to name the category and signal fit, while giving the system a little room to choose. Our post on writing Google Ads copy that repels wrong buyers covers what belongs in this slot in more detail. The short version is that a qualifier in headline one costs you some clicks you should not want anyway.
Plenty of PPC managers treat any pin as a mistake. Google's own guidance is closer to this position than they assume. The RSA help page does say pinning "isn't recommended for most advertisers and can affect ad strength." The same page, and the Ad Strength page beside it, also recommend a hybrid approach of pinning one or two essential headlines, pinning two or three unique versions to a position, and letting the rest rotate (Google Ads Help, About Ad Strength).
Your Ad Strength will probably drop a notch when you do this. Let it. Google's documentation states that Ad Strength "isn't used to calculate Ad Rank, Quality Score, or auction wins" and does not determine whether an ad is eligible to serve. Ad Strength is a good checklist for coverage and a bad referee for a test. I would take a readable test on an ad rated Average over an unreadable one rated Excellent every time.
Write headline themes, not fifteen headlines
A theme is one reason to buy, written three or four ways. A lot of the RSAs we open in audits have fifteen headlines that are really five ideas and ten rewordings, with no way to tell which idea did the work. Build the unpinned pool from two or three themes on purpose instead.
The strongest themes in B2B search tend to come from a short list.
→ Speed or time to value. How fast the buyer gets to the outcome, stated as a period of time.
→ Ease. No IT project, no code, no migration weekend. This one works best when it names the specific pain it removes.
→ One sharp capability. The single thing you do that the searcher's current tool does not. Not a feature list.
→ Proof. A number, a customer count, a result, but only a claim you can back up on the landing page.
Source the raw material from places where buyers have already spoken. Sales call recordings are the best source by a distance, because they hold the words prospects actually use to describe the problem. Customer reviews and support tickets come next, then internal positioning documents and your own website. Pull competitor ads too, mainly so you can see which theme every rival already owns and avoid writing the tenth version of it.
The most useful test inside a theme is specific against generic. "Live in 14 Days, Not 6 Months" and "Fast, Simple Setup" make the same promise. One of them gives a B2B buyer something to repeat to their boss. I expect the specific version to win in most B2B accounts. Test it anyway, because a hard number also repels the people who needed it faster, and in some accounts those were good buyers.
Keep a list of themes that lost. A theme that failed a year ago without evidence behind it deserves a retest once you can attach a real proof point to it. The earlier result may have been a verdict on the missing proof, not on the idea.
Read the test at the ad level, not in the asset report
The unit of the test is the whole ad. On the Ads page, Google reports each RSA's stats as the totals for all the combinations built from that ad's headlines and descriptions. That is a clean number, and it is the one to compare.
The setup that works in most B2B ad groups is two RSAs side by side:
→ Both ads carry the identical pinned headline one set, the same descriptions and the same final URL.
→ Ad A's unpinned headlines come only from theme one, for example speed with proof.
→ Ad B's unpinned headlines come only from theme two, for example ease.
Now the only difference between the two ads is the argument. When one ad produces a better qualified conversion rate, you know which argument to scale.
One honest caveat. If the campaign runs a Smart Bidding strategy, Google automatically uses the Optimize ad rotation setting, which serves the ad it expects to perform better more often (Google Ads Help, Use ad rotation). The two ads will not get equal impressions. Compare rates, not totals, and do not read much into it when one ad gets 70% of the traffic in week two. That imbalance is itself a weak early signal about which theme the system favors.
When an ad group is too thin to test alone
Pool the test across campaigns with ad variations. Many B2B ad groups will never reach a readable conversion count on their own within a quarter. Google's ad variations tool exists for this case. It lets you test one change across multiple campaigns or your entire account, and you choose the percentage of traffic that sees the modified ads (Google Ads Help, About ad variations).
Google positions ad variations for testing one change at scale and custom experiments for testing several changes on a smaller scope. For headline themes that is the right split. Swap one generic headline for its specific counterpart across every non-brand campaign at once, and you get a verdict from pooled volume that no single ad group could give you.
What counts as a winner
Judge themes on qualified conversion rate, and read click-through rate only as the early signal. A theme that lifts CTR while qualified conversions per click fall has attracted curious browsers. In B2B that is a loss dressed up as a win.
These are the rules I use.
1.) Read CTR once each ad has roughly 1,000 impressions. It tells you whether the theme earns attention at all.
2.) Do not call a winner on conversions until each ad has around 15 qualified conversions. Below that, a single good week decides the result.
3.) If an ad group cannot get there within a quarter, move the test into ad variations rather than waiting.
4.) Keep the winning theme, retire the losing one, and bring in the next theme from your list. Change one theme at a time so each test answers one question.
This is the search-specific version of the argument in our post on B2B ad creative testing. Test different ideas, not rewordings of the same idea, and give each test enough volume to finish.
One exception to pinning: AI Max and final URL expansion
If final URL expansion is on, your pins are not respected. Google's AI Max documentation states that with final URL expansion enabled, "pinning of RSA assets will not be respected," and that pinned assets won't be used when a more relevant URL is chosen (Google Ads Help, How AI Max for Search campaigns works).
So run headline tests in campaigns where final URL expansion is off. A campaign that picks its own landing page and rewrites its own copy is not a place to learn which headline works. Our post on fencing Google AI Max in B2B covers how to keep those controls in place.
How to set up an RSA headline test, step by step
1.) Write your category-and-buyer statement three ways and pin all three to headline position one.
2.) List every candidate reason to buy from sales calls, reviews, internal documents and your site, and group them into themes.
3.) Pick the two themes you most need an answer on. Write three or four headlines for each, including at least one specific, provable claim.
4.) Build two RSAs per ad group, identical except for the theme in the unpinned pool. Keep the descriptions and final URL the same.
5.) Make sure the landing page carries the same promise as both themes, or the page becomes the variable. Our post on message match in B2B paid search covers why.
6.) Read CTR at around 1,000 impressions per ad, decide on qualified conversion rate at around 15 qualified conversions per ad, and pool thin ad groups through ad variations.
7.) Write the winning theme down and send it to whoever owns the landing page and the LinkedIn creative. A headline test that never leaves Google Ads has done a fraction of its job.
If you want a second opinion on how your current RSAs are set up, that is part of the account review in our B2B Google Ads management work. Open your top non-brand ad group first. If it has fifteen headlines, no pins and no themes, you have found the place to start.
