Building a Creative Fatigue Detection Cadence That Works

By the time CPI has visibly drifted upward on a concept, that concept has usually been decaying for a week or two already. That gap is the whole problem. Most teams review creative fatigue reactively, when a weekly performance report flags a cost increase, which means the review always arrives after the damage has already worked its way into spend efficiency. A cadence built on leading indicators, rather than the lagging CPI number everyone already watches, catches the decay while there is still time to act on it instead of explaining it afterwards.

Why CPI is a lagging signal

CPI reflects an auction outcome that has already priced in whatever the algorithm learned about a concept's declining relevance to its audience. By the time that repricing shows up clearly in a weekly report, the engagement decline underneath it has typically been building for several days, and every one of those days is spend running against a hook that is already past its useful life. Waiting for CPI to move leaves a team permanently a few days behind the point where a concept actually started losing its audience.

Leading indicators worth watching instead

• Hook-window completion rate, the percentage of viewers still watching past the first two to three seconds, tracked daily rather than weekly. It tends to soften before overall completion rate or CPI moves, since it isolates the exact moment a hook stops earning attention.
• Frequency-adjusted engagement, comparing engagement rate among users seeing a concept for the first time against the fourth or fifth time. A widening gap between those two groups is one of the clearest early fatigue signals available, and it shows up well before blended metrics shift.
• Video ad drop-off timestamps, as covered in Video Ad Drop-Off Timestamp Analysis Beats a Guess, read at the individual second rather than the aggregate completion rate. A drop-off point that creeps earlier week over week on the same concept is decay in its most granular, diagnosable form.
• Click-through rate stability among fresh impressions only, excluding repeat viewers, since a concept can look stable in blended CTR purely because frequency capping keeps refreshing the pool of first-time viewers even as true engagement quality declines.

A worked creative fatigue detection cadence

A cadence that catches decay early doesn't need to be a daily deep dive on every concept; what it needs is a fixed rhythm with escalation built in.

• Daily, five-minute check. Hook-window completion rate for the top three to five spending concepts, flagged automatically if it drops more than roughly 10% from its seven-day average.
• Weekly, thirty-minute review. Frequency-adjusted engagement and fresh-impression CTR across the full active concept set, comparing this week against the prior two, with any concept trending down on both metrics marked for refresh planning.

Fortnightly, anything flagged twice in the weekly review has to get a decision. Refresh it properly, which means genuine new creative rather than a re-edit; retire it from the active rotation; or write down a decision to hold it and the reason behind that, so the same concept doesn't sit in limbo indefinitely.

Genre changes the clock, not the method

Genre changes how quickly this cadence needs to catch a decaying concept. In practice the pace at which a hook decays is not the same in a puzzle title as it is in a hybrid-casual or strategy one, and faster-paced genres burn through concepts noticeably quicker than the slower systems-heavy ones like the survival and kingdom-builder titles currently dominating UA spend charts. The cadence above doesn't need separate metrics per genre. What it does need is escalation thresholds tuned to a title's own historical decay pattern rather than a single studio-wide default, because a strategy title's normal week-to-week variance can look alarming by a puzzle title's faster-moving standard.

Avoiding false alarms

Leading indicators only earn their place if they don't flood a team with false positives. Normal day-to-day variance in a small-spend concept can look like fatigue on a five-minute daily check when the escalation threshold sits too tight, and a team that chases every minor dip ends up refreshing concepts that were never actually decaying, burning production capacity on replacements nobody needed. Set the daily threshold against a rolling seven-day average rather than the previous single day and most of that noise disappears. Reserving an actual refresh decision for the fortnightly review, not the daily check, keeps a team from overreacting to a one-day blip that self-corrects overnight.

What this buys a team

Catching fatigue on leading indicators doesn't produce a dramatic single save. It produces a steady reduction in the number of days any concept spends decaying unnoticed, compounded across every concept in rotation over a quarter. A team running this cadence alongside the concept-output framework covered in Creative Volume Needs After an Ad Network CPI Reset gets both halves of the problem covered: enough new concepts entering rotation, and a reliable trigger for retiring the ones that have stopped earning their spend.

There is a cultural benefit too, and it tends to get overlooked in a purely metrics-driven pitch for this cadence. Creative teams asked to justify a refresh only after a CPI spike have to explain a failure that has already cost the business money, which puts every refresh conversation on the back foot. A cadence built on leading indicators reframes the same conversation as a routine decision made on schedule, before the concept has cost anything. That shift changes how a creative team experiences the review process, from defensive postmortem into forward-looking planning input, and it tends to produce more honest, faster reporting of early warning signs than a system where flagging fatigue early feels like admitting a concept is already failing.

Adoption depends on making the daily check genuinely fast rather than aspirationally fast. A five-minute daily review that in practice requires pulling data from three separate dashboards will not survive a busy week. Worse, a team that lets the check lapse during exactly the weeks when a launch or a spend surge makes fatigue most likely has built a cadence that fails precisely when it matters most. Automate the daily flag, even with a simple spreadsheet formula against an exported report. The setup time is worth it to keep the habit alive past its first few weeks.

Related archive reading

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

• Creative Fatigue Leading Indicators to Watch Before CPI
• Detecting Creative Fatigue Before It Shows Up in CPI