Pinning headlines in responsive search ads: does it help or just cap the test?

Pinning in RSAs splits people. One camp pins the brand name to position 1 and the CTA to position 3 and calls it control. The other says every pin removes combinations the system could otherwise test. I'd pin only when something is legally or brand-critical, like a required disclaimer or the brand name in a brand-heavy ad group. Pinning a generic benefit line to position 1 usually shrinks the set of headline pairs that get served, so the ad ends up leaning on whatever you locked in.

The cleaner test is an unpinned RSA against a pinned version in the same ad group, judged on conversions per impression rather than CTR. Has anyone run that split on a lead-gen account with real volume?

Agreed on pinning only what's legally or brand-critical, but the test design is where lead-gen accounts get stuck. Conversions per impression sounds clean, yet a local plumber or clinic might see 20 to 40 leads a month across the whole ad group. At that volume a 30% gap is often just noise, so you end up pinning or unpinning on a coin flip.

My rule of thumb for low volume is to pin only the brand name or a required line, leave everything else free, and judge the split on cost per qualified lead over four to six weeks, not CTR or the first week. If each arm isn't getting roughly 30 conversions in that window, the test won't tell you much, and the pin decision should come from what the business needs to say, not from the data.

Agreed on the 30-conversion floor, but per-arm volume is only half of it. RSA serving is uneven, so one arm can end up with far fewer impressions than the other even on a 50/50 split, and the low-spend headlines get read as if they were a fair sample. Check impression share per arm before you trust any cost-per-lead gap.

Two RSAs in one ad group won't give you a real 50/50 split. They compete in the same auctions, so the system picks whichever one it prefers and that one ends up with most of the impressions. If you want independent arms, you need a campaign experiment with a fixed traffic split. Are you running this as two RSAs in one ad group or as an experiment?

I don't think "won't give you a real split" is quite right. Uneven delivery between two RSAs in one ad group is normal, because the system favours whichever ad it predicts will perform. That makes it a measurement problem, not a dead end. Check the ad-level impressions and see how lopsided it is. If one ad is getting 80% of the traffic, the comparison is weak. If it's closer, the result may still be usable.

The bigger problem for this particular question is that a pinned and an unpinned RSA in the same ad group share most of their headlines, so they aren't really separate arms to begin with. For a clean answer on pinning, a campaign experiment with a fixed split is the right tool, and it's what I'd use if the question matters enough to wait six weeks.