One dropdown decides if your A/B test runs at all.
If you run two ads per ad group, and you should, then somewhere in that campaign’s settings there’s a dropdown that decides whether you’re actually testing those two ads or just feeding whichever one got off to a faster start. It’s called ad rotation, and Google ships it set the wrong way by default.
I still see this one on accounts I take over almost every time. The setting sits under campaign settings, easy to miss, and the default is “Optimize: prefer best performing ads.” Sounds great. Who doesn’t want the best ad to show more? The problem is what “best” means and when Google decides it.
Under Optimize, Google looks at early click-through data and starts serving the apparent leader disproportionately. Early. As in, before either ad has anything close to enough impressions to call a winner. So the ad that happened to catch the first good run of auctions gets the volume, and the other one gets starved. You wrote two ads to compare them, and the platform has quietly decided the comparison is over before it began. The “loser” never gets a fair shot, which means the test you think is running isn’t really running at all.
I watched this play out on a brand-campaign launch recently. Two ads in the same ad group, same audience, same everything, and within a few days the impressions had split roughly 65/35. That’s not a verdict on the copy, that’s the rotation setting picking a favourite on thin data and then compounding it. The 35% ad could have been the better performer over a full month, and I’d never have known.
The fix is one dropdown: set rotation to “Do not optimize: rotate ads indefinitely.”
That’s it, as long as your bid strategy honours the setting (more on that below). Then both ads get an even shot, the auction data accumulates honestly, and after a couple of weeks you’re looking at a comparison you can actually trust. I set this on every campaign at creation, and on every account I audit that’s still sitting on Optimize, flipping it is one of the first things I do.
A couple of things worth saying, because they’re the objections I hear. First, you might worry that rotating evenly means showing a “weaker” ad more often and leaving money on the table. But you only know an ad is weaker after you’ve measured it, and you can’t measure it if the platform won’t serve it. Rotating indefinitely is the cost of getting a real answer, and that answer pays for itself across every future ad you write for that account. Second, there’s an old belief that certain bid strategies force you onto Optimize. I called that a myth in an earlier version of this piece, and I have to correct myself: for conversion-based Smart Bidding it is real. On Target CPA, Target ROAS and Maximize Conversions, Google picks the ad per auction and quietly ignores the rotation dropdown, whatever you set it to. The even rotation you flip this setting for only truly runs under Manual CPC, and in my own testing under Target Impression Share. So check two things, the dropdown and whether your bid strategy honours it at all. That second check is the one I skipped for years.
None of this contradicts how I judge the ads themselves, by the way. I rank ad quality on impressions, then click-through rate, then conversion rate and volume, then cost per conversion. Rotation doesn’t change any of that. It just makes sure both ads earn enough impressions for those numbers to mean something.
So if you’ve got two-ad ad groups (and again, you should), go check that dropdown before you trust a single ad-test readout. The setting is free to change, takes ten seconds, and decides whether the last month of “testing” told you anything or nothing.
If you find yours on Optimize, you’re not behind, you just found the leak. Go flip it, and let your next test actually be a test.