The maths of creative fatigue

Most creative refresh schedules are built on a calendar, a new batch every two weeks regardless of how the current slate is performing. That approach either wastes production budget refreshing concepts that still have life left, or leaves a decaying concept running past the point where it is quietly inflating CPI. The fix is tying refresh cadence to spend per concept and a measured decay curve, not to the clock.

What the decay curve actually looks like

Creative fatigue is not a cliff. It is a gradual decline in click-through rate and install rate as the same audience sees the same concept repeatedly, and the shape of that decline differs meaningfully by channel because frequency accumulates at different speeds. Meta and TikTok's broad, algorithm-driven distribution tends to concentrate impressions on a narrower slice of the highest-propensity audience early, which means fatigue often shows up faster in CTR than in raw frequency reports would suggest, since the same users are being served the same ad more often than the average frequency number implies. Rewarded video and playable inventory inside other apps tends to decay more slowly, because supply is more fragmented across publishers and the same user is less likely to see the same creative twice in a short window.

Genre also changes the curve's slope. Hyper-casual and casual concepts, built around a single visual hook, decay faster because the entire value of the creative is delivered in the first exposure. Strategy and RPG creative, which often relies on a narrative or progression tease, tends to hold performance longer because repeat viewers pick up additional information on a second or third exposure rather than simply re-seeing the same hook.

Building the formula

A workable refresh trigger combines three inputs rather than relying on any one signal alone.

• Spend per concept since launch, tracked daily.
• The rolling seven-day change in CTR relative to the concept's own first-week baseline, not relative to the account average.
• Frequency, where available, as a secondary confirming signal rather than the primary trigger, since reported frequency understates true exposure on algorithmic placements.

A practical trigger point: flag a concept for review once its seven-day CTR has fallen more than roughly 20 to 25 percent below its own first-week baseline, and confirm the read once cumulative spend on that concept passes a threshold set relative to typical daily spend per concept in that account, so a fast-scaling account flags earlier in absolute days than a slow-spending one. This avoids the two most common mistakes: retiring a concept early because a single bad day dragged the average down, and holding a concept too long because frequency alone looked acceptable while CTR had already turned.

Setting the threshold per channel

Because decay speed differs by channel, a single global threshold either fires constantly on fast channels or never fires on slow ones. The cleaner approach is a channel-specific baseline built from a studio's own historical data: track how many days of spend at typical daily volume it took previous concepts on that channel to hit the 20 to 25 percent CTR decline, and use the median of that history as the expected life of a new concept on the same channel. A concept that starts declining meaningfully earlier than that median is a genuine early-fatigue signal worth investigating for a weak hook rather than normal decay, and one worth pulling before CPI visibly climbs, since CTR decline reliably precedes a CPI increase by several days on most channels.

Tying refresh volume to the curve, not the calendar

Once a channel's expected concept lifespan is known, refresh planning becomes a capacity question rather than a scheduling one. A channel with a fourteen-day median lifespan and five concepts running concurrently needs a new concept roughly every two to three days to maintain a stable rotation without gaps, which sets a production target the creative team can plan against directly, rather than committing to an arbitrary "new batch every fortnight" cadence that has no relationship to how fast the account actually burns through creative life.

The output worth building from this is a single dashboard column: days remaining at current decay rate before a concept crosses its channel's fatigue threshold. That number, refreshed daily, replaces the guesswork in a weekly creative review and gives production a lead time long enough to have a replacement ready before the gap opens, rather than reacting to a CPI spike that has already cost several days of inefficient spend.

Related archive reading

These articles provide related context and remain subject to their stated review status.

• Creative Decay Curve by Genre Looks Different
• Creative Fatigue Leading Indicators to Watch Before CPI