How to diagnose ad fatigue before it tanks your ROAS

Ad fatigue is the slow kind of failure. It doesn't crash a campaign overnight. It bleeds out the unit economics over 10 to 14 days, until the day you finally look at the dashboard and the ROAS is down 32% from the previous week. By then you're already two weeks behind the fix.

The good news: the metrics warn you before the ROAS does. The early signals show up 3 to 5 days before the bottom line moves. If you know what to look for, you can intervene at the diagnostic stage and not at the autopsy.

Here's the diagnostic flow we use, the early signals to monitor, and the framework decision that comes after: refresh, rotate, or replace.

Fatigue is not one thing

The biggest mistake teams make is treating "fatigue" as a single failure mode. It isn't. There are at least three distinct things people call fatigue, and the fix for each is different.

Creative wear-out. The creative itself has been seen too many times by the same people. The audience has stopped reacting to the message. Frequency is high, click rates are dropping, and the ad is still relevant but no longer fresh.

Audience saturation. You've reached the bulk of the audience that was ever going to convert at this price point. Frequency might be moderate, but you're now serving the long tail that wasn't going to convert at any frequency. The creative is fine; the audience is exhausted.

Offer fatigue. The market has internalized your offer, or worse, your competitor has shipped a better one. The creative is fine, the audience is fresh, but the trade you're proposing isn't winning anymore.

Each of these reads similarly in the top-level metrics, which is why teams confuse them. They read very differently in the second-level diagnostics.

The early signals (in order of usefulness)

These are the metrics we watch first, before the conversion volume notices. Order matters because the upstream signals fire first.

1. Hold rate / video watch through

If you run video, hold rate at the 3-second and 6-second marks is the most sensitive early indicator we know of. When viewers who used to hold through second 6 start dropping by second 3, the hook has decayed even though the rest of the creative might still be fine.

Watch the change in hold rate, not the absolute value. A 7-day moving average of "% of viewers reaching 3s" is the cleanest read. A 4-point drop over 5 days is the early warning.

2. Click-through rate (link CTR, not all-CTR)

All-CTR moves on engagement signals that include reactions and comments. Link CTR is the one that tracks intent. When link CTR drops on a fresh audience, the creative is the issue. When link CTR drops on the same audience but stays steady on cold audiences, you're looking at saturation, not creative wear.

The rule of thumb: if link CTR drops 15% over a 5-day moving average, start diagnosing.

3. Cost per landing-page view

If your CPC is holding but cost per landing-page view is rising, the post-click experience is the leak, not the creative. This sounds like fatigue but isn't. Don't fix the ad. Fix the landing page.

4. Frequency

Frequency is overrated as an early signal. By the time frequency is high enough to alarm you, the wear-out is already 5 days deep. Use frequency to confirm a hypothesis, not to form one.

Useful rule: frequency above 3.5 on a prospecting audience is the threshold where wear-out becomes a meaningful contributor. Below 2.5 you're almost never in wear-out territory.

5. Comment quality

Often overlooked. If the comments on a paid creative shift from "where can I buy this" to "how is this still being shown to me," you're past the wear-out point. Read the comments before you trust the metrics. Real audiences tell you they're tired before the numbers do.

The 8-dimension diagnostic

Once a creative shows fatigue signals, the question isn't "is it fatigued." It's "which dimension stopped working?"

This is where the Adverdly Method becomes a diagnostic tool, not just a review tool. Re-score the creative against the eight dimensions and look for which scores have functionally dropped relative to the current audience:

• Hook. Did the audience stop reacting to the opening signal? Common in week 2-3 of a flight.
• Hierarchy. Are viewers no longer reading the creative in the order you designed? Less common, but possible if you're targeting a wider audience than originally.
• Copy. Does the language feel less relevant as the audience has shifted? Frequent in scaling phases.
• Persuasion. Has the persuasion mechanism (social proof, scarcity, contrast) lost weight? Social proof claims often decay first.
• CTA. Has the action become less specific to where the audience is now in the funnel?
• Platform fit. Has the platform changed (new feed UX, new placement rules) in a way that disadvantages this creative?
• Offer alignment. Has a competitor moved the floor on what this audience expects?
• Trust. Have your trust signals (reviews, badges, testimonials) become stale or outdated?

The dimension that has functionally moved is the one that needs the fix. If Hook has decayed and the other seven are intact, you don't replace the creative. You replace the first 3 seconds and keep the rest.

Refresh, rotate, or replace

The three diagnostic patterns map cleanly to three different decisions.

Refresh when the creative has 6+ dimensions still working and 1-2 specific dimensions that have decayed. Most common case: hook needs a swap, or trust signals need an update.

The mistake teams make here: refreshing more than the decayed dimension. If only Hook has dropped, only change Hook. Changing the body of the ad introduces noise and you won't be able to attribute the recovery cleanly.

Rotate when 3-5 dimensions have decayed and the audience is showing saturation signals. The creative isn't broken; it's done. Pull it from active, file it for a return in 60 days when the audience has cycled.

Rotation is underused. Teams default to "kill it" when they should default to "rest it." A creative that worked once will often work again on the same audience after a 60-day gap, because audience memory of paid creative is shorter than people think.

Replace when the underlying offer or angle has stopped working, or when 6+ dimensions have decayed at once. This is uncommon outside of seasonal pivots or competitive responses. When it happens, don't iterate on the dead creative. Build the next one fresh from the audience and angle, not from the corpse of the last one.

The signal for replacement: when you look at the scored analysis and the priority fix is "rewrite the angle," not "rewrite a dimension."

What to do this week

Pick one underperforming ad set. Pull the last 14 days of hold rate, link CTR, frequency, and comment sentiment. Plot the 7-day moving averages.

If hold rate is dropping but link CTR is stable, you have a hook problem. Refresh the first 3 seconds.

If link CTR is dropping but hold rate is stable, you have a copy or offer problem downstream of the hook. Refresh the body, not the open.

If both are dropping in parallel, you have audience saturation. Rotate, don't refresh.

If everything is dropping faster than your moving averages can keep up, you have a competitive or seasonal shift. Replace the angle.

The discipline that saves the ROAS

Fatigue diagnostics work because they front-load the question. Most teams diagnose after the ROAS has moved, when the cost of being wrong is already paid. Teams that monitor the 5 early signals catch wear-out 3 to 5 days earlier and ship the refresh before the bottom line moves.

The hard part isn't the diagnostic. The hard part is checking the early metrics on a cadence that's faster than your dashboard's weekly summary. Once a campaign is in flight, hold rate and link CTR deserve a daily glance, not a weekly review.

Score a creative that's been running for 14+ days and you'll often see the fatigue you missed. Run the analysis on Adverdly and the priority fix will point to which dimension has decayed. Refresh that dimension, hold the rest, and watch the recovery.

The ROAS will follow the hold rate. It always does. Lead the dashboard, don't trail it.