When Meta ads stop converting overnight, establish which of three things happened before you touch the campaign: the conversions stopped, the reporting of them stopped, or the campaign was reset and is learning again. They look identical in the dashboard and they need opposite responses, and rebuilding a campaign that was merely re learning destroys the only asset you had.
Each of these takes a few minutes to rule out. None of them requires changing a setting.
One: did the orders stop, or only the reported ones?
Open your store admin and count actual orders for the last seven days against the seven before it. If store orders held up while reported conversions fell, your advertising is fine and your measurement broke. That happens after a theme update, a consent banner change, an app install or removal, or a change to how the Conversions API is being fed.
This is the check to do first because it is the only outcome where the correct action is to change nothing about the campaign. The gap between platform reported revenue and store revenue is large even when everything is working, which is measured here on an account we owned.
Two: was there a significant edit?
A significant edit sends an ad set back into the learning phase, where delivery is deliberately less efficient until it has gathered around fifty optimisation events. Performance dropping right after a change to targeting, creative, optimisation event, or budget is not the campaign failing. It is the campaign relearning, and the worst possible response is another edit.
Check the edit history before you form a theory. If someone changed something two days before the drop, you already have your answer, and the fix is to wait rather than to act. The mechanics of that threshold are in the note on ad sets that will not spend.
Most accounts that are permanently in learning are there because somebody is trying to fix them.
Three: did the auction change, or did the creative wear out?
If orders genuinely fell and nothing was edited, look at cost per thousand impressions and frequency next to each other over the same period. Rising cost per thousand with steady frequency means the auction got more expensive, which is a seasonal and competitive fact rather than something you did. Steady cost with rising frequency means you are showing the same people the same thing repeatedly, and the creative has worn out.
Those two look the same in a results column and they need different work. One is a bidding and budget conversation. The other is a creative conversation, and on creative we advise rather than produce, which is set out on the creative page.
What if all three come back clean?
Then look outside the ad account, because something changed between the click and the order. Stock going out on your best selling variant, a price rise, a shipping cost change, a checkout app update, or a payment method quietly failing will all show up as ads that stopped converting while the ads carry on doing exactly what they did last month.
Buy something from your own store, on a phone, starting from one of your own ads. Most of the time this ends the investigation. The wider version of that sequence is in the note on traffic without sales.
What we do with it
We run those three checks and say which one it was, in writing, before proposing anything. A drop that turns out to be measurement gets a measurement fix and no campaign changes at all, which is a cheaper week for you and a duller one for us. How that gets reported is on the reporting page.