Why Your Meta Ads Stopped Working — And How to Figure Out Which Problem You Have

Your ads worked. Now they don't. Same account, same offer, roughly the same creative, and the numbers fell off a cliff — or bled out slowly enough that you didn't notice until the month closed ugly. The panic response is to change everything at once: new creative, new audiences, bigger budget, Advantage+ on, attribution window widened, all on the same afternoon. Two days later the number moves and you have no idea why, which means you can't reproduce it and you've learned nothing.

After managing north of $100M in Meta spend, I can tell you that "my ads stopped working" is almost never one problem with one obvious cause. But it's also not the twenty-cause guessing game most articles turn it into. Four causes account for the overwhelming majority of sudden performance drops. This is how to figure out which one is yours before you spend another dollar reacting to it.

First: Separate "Suddenly Worse" From "Gradually Worse"

Before you diagnose the cause, classify the shape of the decline, because the two shapes point at completely different suspects.

A sudden drop — performance was fine, then over a few days it wasn't — is almost always an external event. Something broke or changed at a specific point in time. A creative crossed into fatigue, tracking snapped, a competitor launched, a promo expired, an ad set got reset into learning. Sudden drops have a date. Your job is to find what happened on or around that date.

A gradual decline — a slow erosion over weeks where each week is a little worse than the last — is almost always structural. Audience saturation, a slowly weakening offer against improving competition, creative going stale across the whole account. Gradual declines don't have a date; they have a trend. Your job is to find what's been quietly eroding.

Pull a 60- or 90-day chart of your core metric before you do anything else. The shape of the line tells you which diagnostic path to walk down and eliminates half the possible causes in thirty seconds.

Cause 1 — Creative Fatigue (Most Common)

The mechanism has changed, and understanding the new version is what makes this diagnosable. Fatigue used to mean one thing: the same people saw your ad too many times and tuned it out. That's still true. But under Meta's current delivery system, fatigue also means the system itself has down-ranked your creative's expected performance based on declining engagement. It's not just that your audience is tired of the ad — it's that the delivery model has learned to expect less from it and is pricing and placing it accordingly. That compounds the decline.

How to identify it: frequency climbing past roughly 3–4 in a 7-day window, CTR declining while CPM holds steady or rises, and hook rate (3-second video views over impressions) collapsing. That specific combination — people seeing it more, clicking it less, and abandoning it in the first frame — is the fingerprint of fatigue. If your hook rate is falling, the audience is deciding in frame one that they've seen this before.

The fix isn't a color variant of the same ad. It's a genuinely different angle — a new hook, a new problem framing, a new format — because the system needs a distinct signal to work with, not a reskin of the thing it already down-ranked.

Cause 2 — Tracking or Signal Loss

This is the one that hides, because the ads look fine while the data underneath them rots. What happens is that Meta starts optimizing toward a phantom version of your buyer, because the conversion signals it learns from stopped flowing accurately. The pixel fired on the wrong page after a site update. CAPI stopped syncing after a platform change. Event match quality quietly dropped. The algorithm is still optimizing hard — it's just optimizing toward a target that no longer reflects who actually buys.

How to identify it: open Events Manager and look for recent errors, deduplication warnings, or a drop in event match quality. Then do the crude but decisive check — compare the purchase events Meta is recording against your actual orders in Shopify or Stripe over the same window. If Meta thinks you got 40 purchases and your backend shows 70, the system is flying on partial data, and its optimization is degrading in proportion to the gap.

Most operators skip this layer until everything else looks fine and they're out of other explanations. By then the signal loss has been quietly degrading delivery for weeks. Check it early, especially if your drop lines up with any site change, theme update, or checkout edit.

Cause 3 — Audience Saturation or Retargeting Pool Depletion

Sometimes the account didn't break — it ran out of room. Your warm audience is exhausted, or your retargeting pool is too small to sustain the spend level you've pushed it to. You're asking the same finite group of people to convert again and again, and the well is dry.

How to identify it: CPMs rising sharply while reach stays flat — you're paying more to talk to the same people, not new ones. Frequency very high specifically on retargeting ad sets. A retargeting pool that hasn't grown because top-of-funnel slowed down and stopped refilling it. This is especially common when an account scales spend faster than it scales audience — the budget went up, the addressable pool didn't, and the auction charged you for the mismatch.

The fix lives upstream: the retargeting pool is only as healthy as the cold traffic feeding it. If saturation is your problem, the answer isn't more pressure on the warm audience — it's rebuilding the top of the funnel that fills it.

Cause 4 — Funnel Alignment Broke

This is the invisible cause, because everyone's attention is on the ads and the break is on the page. Something changed downstream — the landing page, the checkout, the offer — and it quietly severed the conversion path while the ads kept performing exactly as before.

How to identify it: CTR stable or even rising, but conversion rate declining. That divergence is the tell. The ad is still earning the click; something after the click stopped closing. A developer shipped a landing page redesign that buried the CTA. A checkout plugin updated and added friction — and checkout friction is one of the most reliable conversion killers there is. A discount code expired and the page still advertises it. The offer changed and nobody updated the funnel around it. Because the ads look healthy, nobody thinks to check the page — which is exactly why this cause survives so long. If your clicks are fine and your conversions aren't, stop looking at the ads and go walk the page as a stranger would.

The Andromeda Factor

A share of post-2025 performance drops have a dimension that simply didn't exist a few years ago, and it's worth understanding as a possibility rather than a default explanation.

Meta detailed its Andromeda delivery system publicly in late 2024, and it completed its global rollout through October 2025. Andromeda reads the content of your creative and uses it as a primary matching signal, while a companion system, GEM, learns from your conversion patterns and shapes what gets prioritized next. Meta published GEM's technical details in November 2025. What this means in practice: as these systems reweighted delivery across 2025, accounts that had been stable for years started behaving differently through no change of their own. Creative signals that used to perform got re-evaluated. Delivery patterns shifted.

I'll be careful here, because this is exactly the kind of claim that invites overreach: Meta has not published the weighting or scoring mechanics, and "it's Andromeda" is a tempting explanation precisely because it can't be checked. Not every performance drop is Andromeda-related, and blaming the algorithm is the most expensive reflex in performance marketing — it stops you looking right when the real, controllable cause is sitting in plain sight. Treat Andromeda as a genuine factor to understand, not a defendant to convict so you can stop investigating.

How to Diagnose Your Specific Cause in 30 Minutes

Run it as a decision tree, starting with the single most diagnostic question.

‍Did CTR drop, or hold?

If CTR dropped, the failure is at or above the click — the ad stopped earning attention. Now split further: is CPM holding while frequency climbs? That's creative fatigue. Did the drop line up with a site or tracking change? Check Events Manager and reconcile Meta's purchases against your backend — that's signal loss making delivery worse. Fatigue and tracking are your two candidates when the click itself weakened.

If CTR held or rose but conversions fell, the failure is below the click. Split again: are CPMs rising while reach stays flat and retargeting frequency is high? That's audience saturation. Is reach fine but on-page conversion down? Something broke on the page, in checkout, or in the offer — funnel alignment. When the ad is still working and the money still isn't, the problem is downstream, every time.

That one branch — did the click weaken or not — cuts the problem in half immediately and routes you to two candidates instead of twenty. From there, the confirming checks above take minutes, not days.

A Worked Example: Walking the Tree on One Account

Here's the tree running on a real-shaped case, because watching it eliminate suspects is more convincing than the abstract version.

A lead-gen account had been holding steady at roughly a $42 cost per lead for months. Over about ten days it climbed to $71 and stayed there. The owner's first move was the usual scramble — he'd already swapped in three new creatives, widened the attribution window, and bumped the budget by the time he stopped to think, which meant his data was now contaminated and he couldn't tell what any single change had done. First lesson: he'd turned a diagnosable problem into a fog machine before diagnosing anything.

Reset and pull the 90-day chart. The shape is a sudden step, not a slow slope — fine for months, then a cliff over about ten days. That classification alone kills the gradual-decline suspects: this isn't audience saturation slowly draining a pool or an offer eroding against competitors. Something happened, and it happened around a specific date.

Now the branch. Did CTR drop or hold? CTR actually held steady — cold traffic was still clicking at the same rate it always had. That single fact eliminates creative fatigue and, largely, the ad layer entirely. The ads are still earning attention. Whatever broke is below the click.

Below-the-click splits two ways: audience saturation or a funnel break. Check CPM and reach — both flat, and retargeting frequency is normal. Saturation is out. That leaves the funnel, so he walks his own landing page as a first-time visitor on his phone, which he hadn't done in weeks. The form loads, he fills it, he hits submit — and the confirmation hangs. A form plugin had auto-updated about ten days earlier and was silently failing to fire the lead event roughly a third of the time. Meta had been optimizing against a conversion signal that was now missing a third of its data, so it drifted toward cheaper, worse leads and reported fewer of them, driving the cost-per-lead up on both ends.

The break lined up exactly with the date on the chart. Total diagnosis time once he stopped thrashing: about twenty minutes. The three creatives, the widened window, and the budget bump — none of them touched the actual problem, and all of them made the numbers harder to read. The fix was a plugin rollback, not a single thing he'd reflexively changed.

The discipline that makes any of this work is refusing to change more than one thing at a time. The reason most accounts stay broken isn't that the owner doesn't care — it's that they care so much they change five levers at once, the number moves, and they can't tell which lever did it. Diagnose the layer, change one thing, read the result. Then move.

Two of these four causes — tracking and audience saturation — you have to check inside your own account. The other two live in your creative and your funnel, and that's exactly what FunnelBoost diagnoses: it scores your hook, copy, offer, message alignment, and landing page, and tells you where the leak is. $99, results in under an hour — start your audit at funnelboost.io.

Q: Why did my Meta ads suddenly stop working?

A: Sudden drops almost always trace to an external event with a date: a creative crossing into fatigue, tracking or CAPI breaking after a site change, a competitor's new offer, or an expired promo. Pull a 60–90 day chart, find the date the line turned, and look for what changed around it. Gradual declines point instead to structural causes like audience saturation or a weakening offer.

Q: How do I know if it's creative fatigue or a tracking problem?

A: Both can drop your CTR, so check frequency and Events Manager. Creative fatigue shows rising frequency (past 3–4 in 7 days), steady or rising CPM, and a collapsing hook rate. A tracking problem shows up as errors or low event match quality in Events Manager, and a gap between Meta-reported purchases and your actual orders in Shopify or Stripe.

Q: Can the Andromeda update cause my ads to stop working?

A: It can be a factor. Andromeda rolled out globally through October 2025 and changed how creative signals drive delivery, so some long-stable accounts started behaving differently. But Meta hasn't published its mechanics, and "it's the algorithm" is an easy way to stop investigating a controllable cause. Rule out fatigue, tracking, saturation, and funnel breaks first.

Q: My ads look fine but I'm getting fewer sales — what's wrong?

A: When CTR holds but conversions fall, the break is below the click. Either your warm audience is saturated (rising CPMs, flat reach, high retargeting frequency) or something changed on the landing page, in checkout, or in the offer. Walk your own funnel as a first-time visitor — the most common version of this is a page or checkout change nobody connected to the ad numbers.

Q: How long should I wait before deciding my Meta ads have stopped working?

A: Look at the trend, not a bad day or two. Daily numbers fluctuate, especially with attribution delay. A genuine problem shows up as a sustained multi-day or multi-week decline on a 30–90 day chart. If you're reacting to a single ugly day, you'll change things that were never broken and lose the ability to read what actually moved.

Related reading: The Complete Meta Ads Audit Guide · High CTR, Low Conversions · What Your Meta ROAS Is Actually Telling You · FunnelBoost home