Skip to content
AppDues
Log in
Contact salesSign up

Creative fatigue and ROAS: why a few winning ads stop scaling

When delivery systems concentrate spend on two or three assets, those ads set the quality baseline for everything else. If nothing new beats that baseline, frequency rises, click-through falls and costs climb — even though the campaign once looked efficient. The durable answer is continuous creative supply, with short-term levers used only to buy time.

Why concentration happens

Audience precision is no longer the main lever. Delivery systems now weigh clicks, likes, comments, shares, saves, views, watch time, click-through and conversion behaviour, partly because privacy limits other signals. That richness lets them identify a leading creative quickly and move most budget toward it.

The risk is concentration. Spend follows engagement at the upper funnel, so install cost can drift away from the quality of the audience actually arriving. Two or three assets may absorb 90% or more of budget. Until a newcomer beats them, the same assets keep receiving impressions, frequency builds, and efficiency erodes.

The core decision: supply beats pausing winners

Pausing every winner to force delivery elsewhere rarely helps. The system then spends on assets below the established baseline, often raising cost per action beyond what fatigue itself would have caused.

A practical decision rule:

  • If you can produce assets, rotate ad groups with fresh work at least weekly and keep searching for new outperformers.
  • If production is temporarily blocked, use a bridge lever below, but set an expiry date and a plan to restore supply.
  • Judge a bridge by whether it preserves signal volume, not only by whether cost dips for a few days.

Pause and reactivate in rotation

Pause one of the top two or three assets for roughly a week so the remaining one or two carry delivery, then pause those in turn. The gradual reset can work for up to two weeks. After one reset, wait at least a couple of weeks of strong performance before expecting another.

Bid or cost caps

Where the network supports them, bid or cost caps limit spend per action for the chosen goal. They may soften fatigue, yet delivery will taper without new assets. Fewer signals then degrade learning, so caps suit short-term use only.

Refresh the copy

New text paired with the same video or banner forces a retest and can postpone fatigue. Expect limited effect: copy is a smaller performance driver, and delivery often settles back to the earlier level quickly.

Change the campaign goal

As a last resort, optimise toward an event just above or below the main action and start learning again. The same assets will probably dominate, so the benefit is delay rather than a cure. Moving to a later event demands enough daily budget to exit learning; otherwise a learning-limited state can worsen results. Audiences may also shift, hurting conversion and ROAS.

Best-case setup: BAU plus testing

With production capacity, use one campaign per country and two ad groups. The BAU group takes 70–80% of budget on proven assets. The testing group takes 20–30%, adding two or three new creatives every two to three days depending on impressions, installs and event signals. Winners graduate to BAU, spend spreads more evenly, and fatigue hits later.

Always optimise toward the most important action. Creative choice is effectively targeting, so optimising elsewhere can make creative comparisons misleading.

Compare each test asset with BAU across CPA, CPI, CTR, install rate, hook rate, hold rate, engagement rate, IPM, install-to-action conversion, frequency and average watch time. That comparison reveals the baseline to beat and shows which elements — hooks, music, headlines, transitions, features, messaging or creators — deserve more production.

Takeaways

Algorithms favour a few assets because engagement drives spend. Without fresh creatives, costs rise and click-through falls. Weekly rotation is the strongest defence; copy refreshes, caps, staged pauses and goal changes only mitigate temporarily. Sustained growth comes from continuous testing judged against clear KPIs.

Frequently asked questions

What signals creative fatigue?

It happens when delivery concentrates on a few assets, frequency climbs, and efficiency falls — click-through drops while costs rise at both upper and lower funnel stages. The fix is a steady supply of new assets, not simply pausing winners, which usually forces spend onto below-baseline work.

How does the 5–7 day pause work?

Pause one top asset for about a week while the remaining one or two absorb spend, then pause those in turn. This can help for up to two weeks; a repeat reset needs a couple of weeks of good performance first.

Are bid caps a long-term solution?

They can delay fatigue by capping spend per action, but without fresh assets delivery shrinks, signals thin out, and performance worsens. Treat caps as a short-term bridge.

Does refreshing ad copy help?

Changing text forces retesting of the same visual, which may delay fatigue briefly. Because copy carries less weight than the creative itself, the algorithm often returns to the prior performance level quickly.

When should you change the campaign goal?

Only as a last resort. Moving to an adjacent event starts a new learning phase, but the same few assets will likely dominate again. If the new event sits after your main action, budget must be enough to leave learning, or a learning-limited state can make results worse.

What setup finds new winners fastest?

Run one campaign per country with a BAU ad group holding 70–80% of budget on proven assets and a testing ad group holding 20–30% with 2–3 new creatives every 2–3 days. Compare testing assets against BAU on cost, click, install, hook, hold, engagement, IPM, install-to-action conversion, frequency and watch time to learn what beats the baseline.