A quick primer
The key idea many people miss: PMax isn't chasing uniform average CPA across every impression. Instead, the system constantly hunts for the "next cheapest" conversion, adjusting bids to the predicted likelihood and cost of each specific auction. This is officially called marginal cost optimization, and a direct result is that conversion cost naturally varies by segment of traffic, including by channel (Search, YouTube, and Display within one PMax campaign, for example). Different CPA by channel isn't an anomaly. It's built into how the system works.
That explains why average campaign CPA can rise even with stable conversion quality inside every individual channel. If the system shifts more budget toward a channel that already had a higher baseline cost per conversion, even while that channel's own CPA stays flat, the overall weighted average rises purely from the mix shift, with no decline anywhere underneath it. Google's own guidance flags this directly: PMax optimizes for marginal, not average, return, so average ROI or CPA by channel can be misleading and doesn't tell the whole story; the system is constantly hunting across channels for the most valuable conversions to maximize total campaign return in real time.
A separate official source confirms why campaign-level evaluation is the right lens here: PMax campaigns should be evaluated campaign-wide, since the system optimizes toward your conversion goals across all channels at once, and billing itself differs structurally by channel (CPM for YouTube in-stream versus CPC for Search, Shopping, and Discover, for example). Even the unit you're billed in changes by channel, which is one more reason average CPA moves when the mix shifts.
What to check before you touch anything
- How CPA has moved inside each individual channel over time, through the channel distribution table. If CPA is stable inside Search and Shopping but the overall campaign CPA rose, that's a strong sign of a mix effect, not a decline.
- How the spend share by channel has moved over the same period. A growing share going to YouTube or Display, which typically carry a higher baseline CPA, will mechanically raise the campaign average.
- Whether the Display or Video share started rising right when you added new video assets or changed your goals or bid strategy. The time series chart shows this clearly.
- Whether there's enough data volume in each channel for a meaningful comparison. On low-volume channels, CPA swings can just be noise, not a trend.
- Whether you're looking at "Conversions" (what the campaign optimizes for) or "Results" (every goal in the account). These give different pictures.
Possible approaches
- Break the change in overall CPA into two parts: the change in CPA within each channel, and the change in spend share between channels, comparing the channel distribution table across two comparable periods. If CPA within each channel is stable and only the overall figure rose, that's a mix effect, and no intervention is needed.
- If CPA genuinely rose inside a specific channel (a within effect), look at traffic quality or the landing page for that channel specifically, rather than the channel mix as a whole.
- Don't try to artificially "pin" channel mix in place to keep average CPA flat. Budget distribution across channels is itself a result of the optimization, not a separate problem to fix.
- If a growing Display or Video share coincided with new assets, judge whether that growth is justified by added value (Results or Results value), rather than judging it purely by the average CPA, which will almost always tick up temporarily in that case.
- Evaluate campaign success at the campaign level against your CPA or ROAS target, rather than judging individual channels in isolation.