How to Know When to Kill Your Facebook Ad Before Your ROAS Drops

Your own dashboard tells you an angle died after you paid to find out. A competitor's account tells you earlier, and it puts a number on it.

The problem with killing on your own metrics

Most advice says kill a Facebook ad on your own numbers. Watch CTR fall, watch CPC climb, watch ROAS flatten, then cut.

The logic is sound and the timing is bad. Those are all lagging indicators of the same underlying event: an angle that the market has stopped responding to. By the time your account has enough data to show it clearly, you have already bought that data.

Creative angles rarely fatigue for one advertiser alone. They fatigue across a category, and the brands with the biggest accounts hit the wall first because they have the most impressions running through it. Which means the evidence exists outside your account before it exists inside it.

What the product actually labels, and what it does not

One correction first, because it matters if you go looking.

Brandsearch does not have a "Phase" filter in Discovery with Testing, Scaling, Winning and Inactive values. What it has is a lifecycle classification on a brand's Ad Rank tab, and the four labels are Scaling, Evergreen, Fading and Not measured. They are not applied by guesswork; each one states its own rule.

Fading is the kill signal, and it is defined as a loss of rank.

This is Meller, a Barcelona sunglasses brand. The header reads Fading 46, lost 6+ rank positions, and each row shows the slide rather than a position: #27 to #34, #21 to #36, #26 to #38, #26 to #44.

Now read the right-hand column, because that is where the warning is. The ad that fell from #27 to #34 has been running 207 days and carries 14 duplicates. The one that fell to #44 ran only 35 days but carries 13 duplicates.

Those are not abandoned tests. A brand does not make fourteen copies of a creative it does not believe in. These are ads the brand committed to, duplicated, and is now watching slide. That is what a fatiguing angle looks like from outside, and it is visible without access to anyone's ad account.

Their decline is an opening as well as a warning, which is the subject of what to do while a competitor's best ad is dying and the replacement is not ready.

Check whether the angle you are running is fading in your category

Read the whole account before you read one ad

A single fading ad is noise. A brand can retire a creative for a dozen reasons that have nothing to do with the market. The signal is the balance.

The legend does the work: Scaling 38, Evergreen 1, Fading 46, Not measured 35. More of Meller's measured ads are losing rank than gaining it. This is an account rolling over, not one bad creative.

The scatter underneath says something the counts alone cannot. Red and blue points sit on top of each other at nearly every value on the Days axis, from under 30 days out past 500. Fading is not something that happens only to old ads: the fading list above includes an ad at 35 days, while blue points are still climbing near the right-hand end of the same axis. Age is not the variable here, and any rule of the form "kill at N days" is fitting a number to the wrong axis.

A healthy account looks visibly different in this same view, with the red almost absent. That contrast, and how to read the gaining side of it, is covered in how to tell if competitor ads are actually scaling.

What to do with the signal

The useful question is not "is this ad dying" but "is this angle dying across the brands that spend the most in my category".

1. Pick five or six brands that genuinely compete with you and open each one's Ad Rank tab.
2. Read the legend on each. You are looking for the ratio of Fading to Scaling, not the raw count, because a big account has more of everything.
3. When the brands with the heaviest accounts tip toward Fading on an angle you are also running, treat that as your notice. They reached the frequency wall first.
4. Open their Scaling bucket in the same visit. The angle they are moving budget toward is the more useful half of the signal, and it is in the same view.

The move is not to copy the creative that is climbing. It is to understand which underlying desire it addresses, and to have your own version of that ready before your own ROAS tells you what the competitor's account already did.

Pull the scaling and fading split for the brands in your niche

One number that is easy to misread

Rank in Brandsearch is worldwide. It is computed across everything a brand runs in every market, so when Meller's ad slides from #27 to #44 it slid within the brand's whole account.

The euro spend figures elsewhere in the product are not worldwide. They are the slice EU and UK disclosure rules force advertisers to publish. So do not read a small euro figure as a weak ad, and do not read a rank as a regional result. They are measuring different things, and only one of them is complete.

Method

Everything quoted here is transcribed from two Brandsearch captures taken on 13 September 2026, both from the Ad Rank tab of one brand: mellerbrand.com, which the app's own brand card describes as a Barcelona sunglasses brand. No figure from any other account is quoted on this page, which is why the healthy-account comparison above links out rather than citing numbers it does not show.

The Fading definition, the counts, the rank moves, the days-run values and the duplicate counts are all printed in the screenshots above. The threshold is set per brand: Meller's panel reads "lost 6+ rank positions", so read the definition on the account in front of you rather than assuming six.

Brandsearch indexes 223M ads across Meta and TikTok and 13.7M stores (9M of them Shopify). The Fading label depends on daily rank history, which is why it can describe a direction rather than a position.

No claim is made here about what any of these ads cost to run, or about anyone's ROAS, because neither is observable from outside an ad account. What is observable is rank, movement, runtime and duplication, and that is what this page uses.

FAQ

How many competitors do I need to watch before the signal is real?
Enough that one brand's internal decision cannot swing it. Five or six accounts that genuinely compete with you is a workable floor, and you are reading the Fading-to-Scaling ratio on each rather than counting individual ads.

Does a fading ad mean the brand is in trouble?
No. Meller has 38 ads gaining rank at the same time as 46 losing it. Accounts rotate creative constantly. What matters is the balance and whether it is moving.

Is "Fading" the same as an ad being switched off?
No. Fading means the ad lost rank positions while still being observed. An ad that has stopped is a different state, and ads without enough history to have a direction sit in Not measured instead.

Can I see this on TikTok too?
No. Ad rank and disclosed spend in Brandsearch are Meta only. The TikTok data is organic engagement, which answers a different question.