Creative fatigue and comparing audiences: the Demand Gen monitoring method

In short: a Demand Gen campaign doesn't break suddenly, it wears out: CPA climbs week after week with nothing changing. The signature of fatigue is CTR falling + frequency rising + impressions flat, measured per creative and audience. Reference thresholds against the average: warn < 0.8, refresh < 0.7, pause < 0.5 — calibrated per account.

A Demand Gen campaign doesn't break all at once: it wears out. The creatives burn out, an audience saturates, and the CPA climbs week after week without anything "changing". The monitoring that prevents that decay is a simple calculation repeated with discipline. This lesson formalises it as a method — the one Demand Gen Optimizer implements — so you can either do it by hand or understand what the machine is doing.

How do you measure creative fatigue?

Per creative (ad or asset) and per group/audience, every week:

The decisive combination: CTR falling + frequency rising + stable impressions. A CTR drop with impressions cut in half is something else (budget, competition, a bid change).

At what thresholds should you refresh a creative?

• Warning (amber): CTR/peak < 0.8 two weeks running → get the next creative out of the bank.
• Refresh (orange): CTR/peak < 0.7 or CPA/average > 1.3 with volume → add a new one, overlap for a week, withdraw the burnt-out one.
• Pause (red): CTR/peak < 0.5 or zero conversions with meaningful spend for two weeks → pause the creative (via checkbox).

The thresholds are calibrated per account: with small remarketing lists fatigue comes faster and the thresholds are looser; with broad prospecting, the other way round.

How do you compare performance across audiences?

With the "one group per audience" structure (lesson 2), a weekly table:

Sorted by CPA, with three decisions:

1. Groups with a CPA on target and limited volume → widen the reach (broader lookalikes, expansion) before adding budget.
2. Groups with a CPA off target consistently (4 weeks) and enough volume → pause the audience or change the creative (if the CTR was good, the problem is the landing page or the audience; if it was bad, the creative).
3. Hot remarketing is left out of the prospecting comparison.

How do you automate pauses safely?

Two operating rules that avoid the typical damage:

• Pause only with approval: the table proposes, you tick. A creative paused at the wrong moment (in the middle of your season) costs more than one more week of fatigue.
• Alert anti-spam: one warning per creative and state, not one a day; and no repeats while the state doesn't change. Alert fatigue kills the monitoring just as creative fatigue kills the campaign.

Do it by hand or automate it?

By hand: a sheet with the table by creative and by group, filled in every Monday from the asset and audience reports (30-45 minutes per campaign). It is workable with one or two campaigns; with five, it stops happening in the third month.

💡 Ninja trick: Demand Gen Optimizer is this lesson turned into a script: a fatigue engine by creative and audience (CTR vs peak, frequency, CPA vs average, with minimum volumes), a sorted comparison between audiences, pause proposals via checkbox (executed through the Google Ads API once you approve them), alert anti-spam and a weekly email with whatever has changed. The thresholds are yours; the discipline is the machine's.

What you should remember

• Fatigue = CTR falling + frequency rising + stable impressions; measure by creative and by audience, with a minimum volume.
• Thresholds: warning < 0.8 · refresh < 0.7 (or CPA/average > 1.3) · pause < 0.5; calibrated per account.
• Audience comparison: widen what performs on little volume; pause what fails consistently; remarketing kept separate.
• Pauses via checkbox and alert anti-spam.