Halfway through the year, most of the accounts I'm reviewing are celebrating the same number: click-through rate is up compared to January. It should be. Every account's CTR improves as targeting and creative testing mature over six months. The number nobody's watching in the same meeting is conversion rate on that traffic, and in three accounts I've looked at this month, it's quietly down over the same period. Net those two moves against each other and two of the three accounts are paying more per actual customer than they were in January, inside a reporting narrative that says everything improved. That's not a data problem. It's a slide-selection problem, and it costs real money in the H2 allocation it's about to inform.
The mechanism is straightforward once you look for it, and it's almost never discussed in a mid-year readout. As a campaign matures, the algorithm gets better at finding people who will click. It does not automatically get better at finding people who will buy, and those are not the same skill or the same population. Six months of optimization toward CTR as a proxy metric can produce an audience that's genuinely more click-happy and no more likely to convert, sometimes less likely, because the algorithm found the cheapest clicks available rather than the best buyers. This is Goodhart's law running quietly inside an ad account: the proxy became the target, the system got extremely good at the proxy, and the thing the proxy was supposed to stand for wandered off unmeasured.
Creative fatigue makes this worse in a specific way. The same three ad concepts running since January are still generating clicks, often at a lower cost as the algorithm gets efficient at serving them, but the audience seeing them for the fifteenth time is disproportionately people who click out of familiarity or habit rather than genuine intent. CTR holds up fine. It's a comfortable, familiar-looking number. Conversion rate on that same traffic erodes quietly underneath it, and because the CTR chart is the one going up and to the right, it's the one that gets screenshotted into the mid-year deck.
The general principle is worth naming, because CTR and conversion rate are just this month's example: paired metrics have to be read as pairs, and every flattering number in an account has a counterpart that can hollow it out. CPL pairs with lead quality. ROAS pairs with new-versus-returning mix. Reach pairs with audience qualification. Impression cost pairs with placement quality. An account review that reads any of these solo is not reading performance; it's reading whichever half of the story the presenter chose to bring, and presenters bring the ascending line. The reporting rule that fixes it is almost embarrassingly simple: no leading indicator appears on a slide without its trailing partner on the same slide. One rule, applied ruthlessly, and the flattering-screenshot genre of reporting dies in a month.
This is exactly the kind of gap a mid-year check exists to catch, before H2 budget gets allocated based on the metric that's telling the more flattering story. Pull conversion rate specifically for the creative that's been running longest and compare it against creative launched in the last thirty days. If the pattern I'm seeing elsewhere holds, the newer creative is converting better even with a less impressive CTR, and that's the number that should be driving the refresh decision, not the one that looks best in a screenshot. While you're in there, segment the long-running creative by frequency band; the high-frequency slice is usually where the habit-clickers live, and excluding or capping it recovers conversion rate without touching anything else.
None of this means CTR is a useless metric. It means it's a leading indicator that needs a trailing one sitting right next to it on the same slide, permanently, by rule rather than by someone's monthly vigilance. Right now, at the midpoint of the year, is exactly when that gap is cheapest to close, because the H2 budget hasn't followed the pretty chart yet. In six months it will have, and the same review will be explaining a miss instead of preventing one.