The Ad That's Dying Might Just Be Losing the Coin Flip on Who Gets Shown Last.
A conversion count crashing week over week on one ad usually reads as a creative that stopped working. Often the platform reshuffled credit inside the same ad set, not a real performance change.
The panic moment: one ad's numbers fall off a cliff
An advertiser watching a specific ad's conversion count crater week over week has an obvious read available: the creative wore out, the audience saturated, something broke. The instinct is to pause it, replace it, or dig into the creative for what went wrong. None of those diagnoses may be true.
What's actually moving is which ad gets the credit
Inside a single ad set, several ads compete for the same audience, and the platform's last-touch logic assigns each conversion to whichever ad the person happened to interact with last before buying. When delivery shifts, even slightly, toward a different ad in the same set, conversions that used to land on one ad start landing on another instead. The first ad's count drops. The second ad's count rises by roughly the same amount. Nothing about either ad's actual pull on the audience changed. The credit did.
Judge at the level the platform can't reshuffle
A documented case moved 8% on total account conversions and $4 on cost per action while one specific ad's individual numbers appeared to collapse, because the account-level total nets out the reallocation and the single-ad view does not. Reading performance at the ad set or campaign level, not ad by ad, avoids mistaking a credit shuffle for a real decline. In-between views, checking a single ad against last week's version of itself, read like spaghetti code: technically accurate, structurally meaningless.