The RSA asset trap: when your top line is a double count

A Responsive Search Ad is not one piece of ad copy but a pool of headlines and descriptions that Google combines itself. Most reporting stops at the ad group level: how does this RSA perform as a whole. That misses exactly the level where the real question sits, namely which individual headline or description is doing the work.

Where the double count comes from

Go one layer deeper, to performance per asset, and there is a reason few specialists do this structurally: the same impression, click, or conversion counts toward every asset that was in that specific combination. A headline that happened to be shown often alongside a strong second headline looks better than it is on its own. Without a hierarchy that corrects for that, "our best-performing headline" ends up as a conclusion that mostly says something about who happened to stand next to it.

A worked example

Illustrative case: an ad group's top headline by raw conversions is "Free shipping, no minimum," credited with the same conversions as several other strong headlines because Google served it almost exclusively in combination with them. A second headline, "Same-day dispatch before 3pm," shows fewer raw conversions but was served far more often alone or alongside weaker headlines -- and converts at a comparably strong rate on its own. Read at the ad group level, the first headline is the winner. Read per combination, the second is doing more of the actual work; it just never got paired with help.

The fix is reading order, not statistics

The fix is not complicated statistics, but discipline in reading order: first keep the combinations that occurred most often separate from combinations rarely shown, and only then judge an asset on its own merits, not on the average of every combination it ever appeared in. Skip that hierarchy and you optimize on noise that looks like a pattern -- pausing a headline that was never actually weak, or protecting one that was only ever propped up.

Why this matters beyond one ad group

It is the same reason copy analysis at the asset level (the Meta equivalent is called creative fatigue, a different mechanism with the same symptom: a good number telling the wrong story) deserves its own kind of attention, separate from ordinary ad group reporting. It is not more work for its own sake, it is the place where the double count otherwise goes unnoticed, and where a confident-sounding "best headline" conclusion is most likely to be wrong.