Why your Meta ROAS dropped: the usual causes in order, and how to tell which one it is

When Meta ROAS drops, it is almost always one of four things, and you should check them in this order: the measurement changed, your best ads have tired, the offer or the page stopped converting, or your campaigns are bidding against each other for the same people. The first one costs nothing to check and explains more drops than people expect. The second is the most common real cause. The last two are rarer, and they are the ones people jump to first.

Most of the articles ranking for this search pick one cause and tell you it is the answer. Some say it is always tracking. Some say it is always creative. In the accounts we run, it is usually one cause hiding behind another: a measurement change makes a tired ad look worse than it is, or a tired ad makes a weak offer look like a creative problem. So the order you check things in matters more than any single fix.

First, check the measurement changed and not the business

Before you touch a campaign, compare what Meta says with what your store says. Pull orders from Shopify, or whatever you sell on, for the same days, and set them next to the purchases Meta reports. If your real orders held up and only the Meta number fell, nothing is broken in the account. The scoreboard moved.

This happened to a lot of advertisers this year. On 18 March 2026 Meta changed what counts as a click for attribution, so only outbound clicks to your website now earn click-through credit. Before that, a like, a save or a comment could. An analysis of 50 accounts by Common Thread compared the two weeks either side of the change and found reported ROAS fell on average, with no change in the real business behind it.

The other half of measurement is signal. If the pixel misses purchases, because of browser blocking or a weak server side setup, Meta is learning from part of the picture. Journify suggests that a gap of more than 10 to 15 percent between your backend and Ads Manager points to a signal problem, and that event match quality on a default setup usually sits between 40 and 55 percent against 70 to 90 for a strong one. Both are worth a look in Events Manager.

What we would add from our own accounts: decide your target on the numbers you trust and keep it there. If you judge ads against a target set before a measurement change, every ad looks like a loser and you cut the ones that were working. We set each account's target return once, write it down, and judge every creative against that same bar, so a drop in the reported number shows up as a drop across everything at once rather than a run of individual "losers".

Second, check whether your best ads have tired

If the store numbers fell too, the most common real cause is creative fatigue. The people most likely to buy have seen your best ad several times, they have either bought or decided not to, and Meta is now showing it to people further from the sale.

You can see it in three numbers, ad by ad rather than campaign by campaign:

Signal	What it looks like	What it means
Frequency rising	The same people are seeing the ad more often each week	Meta has run out of fresh people it thinks will buy
Click-through rate falling	Fewer clicks per thousand views than the ad's own first fortnight	The ad has stopped stopping the scroll
Cost per result rising while CPM holds	You pay the same to be seen but get less for it	The ad, not the auction, is the problem

Fatigue comes faster than most brands plan for. On a UK luxury laundry brand, four videos launched on 9 September and two statics launched on 17 September had all tripped our fatigue check, frequency up or click-through falling, by early October. On a kids' supplement brand, three UGC videos launched in mid September did the same within about three weeks. Neither account had a broken offer. The audience had simply finished watching those ads.

The fix for fatigue is new creative, not a new offer and not a new audience. Rewriting the offer because one ad tired is the most expensive mistake on this list, because it throws away the part that was working. Keep the offer, keep the targeting broad, and give Meta new ads that say the same thing to different people.

That last part matters. A new ad that is the old ad with a different first frame often tires in days, because Meta treats it as the same idea. What lasts is a different person and a different reason to buy. In our first August Weekly we wrote about a UK luxury home brand where the ad written for the buyer who hesitates on price beat the ads written for people who already loved the category, at nearly double the target return. That is a new idea, not a new edit.

Third, check the offer and the page

If your ads are fresh and the measurement is sound, look at what happens after the click. A ROAS drop with steady click-through rates and steady cost per click usually means the traffic is fine and the conversion is not.

Check these before you blame the ads:

• Price or stock changed. A best seller went out of stock, a price went up, a bundle came off the site. The ads are still sending people to a page that no longer sells what the ad promised.
• The page changed. A new theme, a new checkout app, a slower page on mobile. Compare conversion rate by device for the weeks either side.
• The offer and the ad disagree. The ad promises one thing and the page leads with another, so the click arrives and leaves.
• Average order value fell. Same number of orders, smaller baskets, lower ROAS. That is a merchandising question, not an ads one.

One thing we would warn against: reaching for a bigger guarantee or a sharper discount to rescue a falling number. On the same UK luxury laundry brand, the ads built around a risk reversal, the promise that you lose nothing if you do not like it, came in at a little over half the target return. On a London supercar hire company, the ads that led with the founder's offer barely got delivered at all. Meta tried them and moved the money elsewhere. An offer cannot fix an ad that is not talking to the right person, and it costs you margin while it fails to.

The opposite case shows what a fixed page and a clear offer can do. One of the brands on our case studies page, Le Luxe, went "from a 0.7x ROAS to a consistent 3.2x ROAS in 3 months". That kind of move is rarely one change. It is the measurement trusted, the creative refreshed, and the page selling what the ads promise, all at the same time.

Fourth, check whether your campaigns are fighting each other

Audience overlap is real but overdiagnosed. It happens when two or more ad sets target the same people, so you bid against yourself in the auction and each ad set learns from fewer results.

Signs it is your problem:

• You have many small ad sets with similar or broad targeting, each spending a little.
• Ad sets sit in the learning phase for weeks because none of them gets enough results on its own.
• Cost per result rose across the account at the same moment you launched new campaigns, while the ads themselves did not change.

The fix is usually fewer, larger ad sets with broad targeting, and letting the creative do the work of finding the buyer. Meta's delivery now leans heavily on the ad itself to decide who sees it, so a varied set of ads inside one broad campaign usually beats a dozen narrow audiences.

Be careful with one related trap. When you put several new ads in one ad set, Meta picks favourites quickly. For the London supercar hire company in our second August Weekly, the ads that sold the occasion got almost no spend, while the ads aimed at someone who wants to make a statement in London brought leads in at about two thirds of the target cost. An ad that cannot get delivered is telling you something, but it is not the same as an ad that was tested and lost.

A ten minute check, in order

When the number drops, work down this list and stop at the first one that explains it:

1. Store orders against Meta purchases for the last 14 days and the 14 before. Real orders steady? It is measurement. Adjust the target, change nothing else.
2. Event match quality and the backend gap. Big gap or low match quality? Fix the signal before judging any ad.
3. Frequency and click-through rate on your top five ads by spend. Rising frequency and falling click-through on the ads carrying the account? It is fatigue. Launch new ideas, keep the offer.
4. Conversion rate, stock and average order value on the site. Clicks steady but sales down? It is the page or the offer.
5. Number of ad sets and time in learning. Many small ad sets, stuck in learning? It is structure. Consolidate.
6. Only then, the auction. Seasonal CPM rises, like the run into Black Friday, are real, but they hit everyone, and they are the one cause you cannot fix from inside the account.

A drop is also not always a reason to pull back. Revice, another brand on our case studies page, "scaled Meta spend nearly 4x in a month while ROAS rose from 4.2x to 5.7x". Spend and return can rise together when the creative keeps up.

When it is time to get help

If you have worked through the list and the answer is fatigue, the question becomes whether you can make enough new ideas, each a different person and a different reason to buy, fast enough to stay ahead of it. That is the problem most brands spending $10k or more a month on Meta run into, and it is the one we spend our week on. If you want to see what that looks like on your account, start at /free. If you are weighing up agencies, our guide to what a good Meta ads agency does covers what to ask.

FAQ

Why did my Meta ROAS drop overnight? A drop that arrives overnight, across every campaign at once, is usually measurement rather than performance. Compare store orders with Meta's reported purchases for the same days before changing anything.

How do I know if it is creative fatigue? Look at your top ads one by one. If frequency is rising and click-through rate is falling while CPM holds steady, the ad has tired. The fix is new ads with new ideas, not a new offer.

Should I change my offer when ROAS drops? Only if the ads are fresh, the measurement is sound and the page is where people are leaving. In our accounts, a bigger guarantee or discount has not rescued ads that were talking to the wrong person.

Does audience overlap really hurt ROAS? It can, when many small ad sets chase the same people and none leaves the learning phase. It is less common than people think. Fewer, broader ad sets with varied creative usually solve it.