8 questions you can answer from memory. 2 minutes. You get a ranked shortlist of what likely broke, why your symptoms point there, and the concrete fix. No email, no signup.
This shortlist comes from 8 answers, not from your account data. It narrows the search, it does not replace looking at the numbers. Confirm the top diagnosis before you change anything.
Facebook ads rarely stop converting for mysterious reasons. Across the accounts we audit, the same seven causes explain nearly every drop, and each one leaves a distinct fingerprint in the metrics. The order below reflects how often each cause turns out to be the real one once we dig in: tracking problems and creative fatigue lead by a wide margin, genuine market shifts trail far behind.
Work out which fingerprint matches before you touch the account. Half the damage we see comes from the response rather than the drop itself: founders kill healthy campaigns and reset learning phases to fix a problem that lived in the checkout the whole time.
The most common cause and the least suspected. A theme update, a new app, a consent banner, a checkout migration, or an expired Conversions API token silently stops purchase events from reaching Meta. Two things then happen at once: your reporting shows a collapse, and the algorithm loses the signal it optimizes on, so real delivery degrades too.
The signal: reported conversions fall off a cliff within a day or two while clicks and CPMs stay normal. Events Manager shows the Purchase event flatlining or losing match quality. The drop often starts the same week as a site change nobody connected to it.
The fix: run Test Events in Events Manager against a real purchase and verify the pixel and CAPI both fire with deduplication intact. Or skip the manual work: our free tracking scan checks all of this in about 60 seconds.
Ads wear out. The audience that clicked in week one has seen your ad a dozen times by week six, and response decays with each exposure. Nothing in the account changed. The audience simply stopped responding.
The signal: a gradual slide over 2 to 6 weeks. Frequency climbing above 3 on prospecting. CTR falling week over week while CPM stays roughly flat. New creative briefly restores performance, then follows the same curve.
The fix: replace concepts, not colors. A new hook, a new format, a different angle. Small recuts of the winner inherit its fatigue. We ship 4 to 8 genuinely new creatives per two week cycle on most DTC accounts, and the refresh cadence matters more than any single ad.
Meta relearns delivery after significant edits: budget changes beyond roughly 20%, swapped creative, changed targeting, pauses and relaunches. During relearning, performance dips and volatility spikes. Founders who edit daily keep their ad sets in permanent learning and never see the account's real baseline.
The signal: the drop starts within days of edits. Ads Manager shows Learning or Learning Limited on your main ad sets. Performance whipsaws instead of trending.
The fix: batch edits into one session per week. Raise budgets about 20% at a time, every 2 to 3 days. Test new creative in separate ad sets instead of editing winners. Then leave it alone for 72 hours minimum before judging any change.
The ad's job ends at the click. When clicks hold and purchases vanish, the problem lives on your site: a theme update that broke the mobile variant picker, an app conflict that kills the cart, a shipping rate that doubled, a discount code that expired mid campaign.
The signal: CTR fine, traffic fine, on-site conversion rate down. The drop coincides with a site change. Abandoned carts tick up.
The fix: buy from your own store on a phone, on cellular data, starting from the actual ad. Watch every step for friction: load speed, variant picker, cart, discount field, payment. Roll back the most recent change and watch for recovery. Then confirm the same change did not also break your pixel.
Budget raised faster than the audience and creative can absorb. At higher spend Meta buys deeper into the auction, reaching people less likely to convert at higher CPMs. CPA rises with spend, and doubling the budget can more than double the cost per purchase.
The signal: performance held until budgets jumped. CPA rises roughly in step with spend. CPM climbs while CTR holds steady.
The fix: return to the last budget that hit your target, then scale in 20% steps with new creative feeding each one. Broad audiences absorb increases better than narrow interest stacks. If you need the volume regardless, accept the higher marginal CPA knowingly instead of discovering it in the monthly numbers.
Apple's App Tracking Transparency limits what Meta can observe on iOS devices. Ads Manager undercounts conversions for iOS heavy audiences, and the gap widens wherever tracking infrastructure is weak. The account looks broken while the bank account disagrees.
The signal: Ads Manager sags while Shopify or Stripe revenue holds. The gap grows when your audience skews toward Apple devices. Blended revenue over ad spend stays steady.
The fix: judge the account on blended numbers first: total revenue divided by total ad spend for the period. Then improve what Meta can see: Conversions API alongside the pixel, with strong event match quality on purchase events. Our free scan grades exactly this.
Sometimes nothing broke. Post-holiday lulls, summer slowdowns, and Q4 CPM inflation move whole categories at the same time. Your funnel converts the same fraction of the same intent. There is simply less of it, or it costs more to reach.
The signal: the drop matches the same weeks last year. CPMs rose across every campaign at once. Competitors in your category went quiet or loud together.
The fix: check year over year in Ads Manager before changing anything. If the dip repeats annually, plan for it: pull spend to your seasonal floor, keep creative testing alive through the trough, and rebuild spend ahead of the recovery instead of after it.
Diagnose before you edit. Every change made while guessing resets learning, muddies attribution, and stacks a second problem on top of the first. Get the diagnosis, make one fix, wait 72 hours, measure. The quiz above gives you the shortlist in 2 minutes, and the tracking scan confirms the most common culprit in about 60 seconds.
An overnight collapse is almost never creative or audience. Ads decay gradually; infrastructure breaks instantly. Check three things in order: tracking (did the pixel or Conversions API stop firing after a site change), delivery (did a campaign get paused, rejected, or budget capped), and learning resets (did a big edit push ad sets back into learning). The Test Events tool in Events Manager answers the first in minutes, or our free tracking scan checks pixel, CAPI, and event match quality in about 60 seconds.
Three numbers tell you. Frequency climbing above 3 on prospecting, CTR falling week over week, and CPA rising while nothing else in the account changed. All three together points to fatigue with high confidence. The fix is new concepts with different hooks and formats, not small recuts of the same winning ad, which inherit its fatigue.
For cold prospecting we treat 3 as the line: above that, most audiences have seen the ad enough times that response decays. Retargeting tolerates more, often 5 to 8, because warm audiences need repetition before they buy. Judge frequency over the ad's lifetime, not just the last 7 days, and always read it next to CTR: high frequency with a stable CTR can keep running.
Yes. App Tracking Transparency limits what Meta can observe on iOS devices, so Ads Manager undercounts conversions, especially for iOS heavy audiences. The test: compare platform reported revenue against your store's actual revenue for the same period. If the store looks fine while Ads Manager sags, you have a reporting gap, not a performance problem. Conversions API with strong event match quality closes much of the gap.
72 hours minimum for a normal dip, a full week if the ad sets recently exited learning. Meta's delivery has natural variance, and reported conversions trail clicks by up to a week because of attribution windows. The exceptions are structural breaks: a pixel that stopped firing or a checkout that errors should be fixed the moment you find it. Diagnose first, then either fix immediately or wait deliberately.
A drop you can diagnose in 2 minutes usually points at a system that let it happen. We run Meta accounts with tracking verified weekly, creative refreshed on a fixed cadence, and edits batched so learning never resets by accident.