A quick primer
Before calling this a leak, it helps to understand the logic behind channel selection in the first place. Marginal cost optimization means the system is constantly hunting for the most cost-efficient conversion opportunity at any given moment, not aiming for one uniform average CPA across every impression. A channel getting budget at a lower cost per impression often means the system found a cheaper, but still qualifying, conversion there, not that the spend is aimless.
Google's own FAQ addresses this exact scenario directly: if PMax is serving heavily on a channel that fit your brand, keep in mind you can't directly control budget allocation by channel, but you can influence it indirectly, through asset optimization for the channels you want, refined audience signals, correctly configured conversion values, search themes, and content suitability plus account-level exclusions for brand safety. That's Google acknowledging the channel isn't necessarily broken; it's a signal that the inputs could be sharper.
Separately, there's an honest, less intuitive scenario worth knowing: if a channel isn't spending at all and the Status column reads "Eligible" with no warnings, that most likely just means PMax is bidding on other channels with a higher predicted ROI in the moment, not a technical problem. Even zero spend on a channel with a clean status is part of normal marginal cost optimization behavior.
What to check before you touch anything
- Is the channel really underperforming on conversions or CPA, or is it just getting a smaller spend share at a lower cost per conversion. Cheap isn't the same as inefficient.
- What does the Status column show for that channel in the channel distribution table. "Eligible" with no warnings points to normal budget allocation, not a problem.
- Are conversions set up correctly with the right value assigned. An undervalued or misconfigured conversion action can make the system undervalue a channel's real contribution.
- How specific are your audience signals. Vague, broad signals give the model less to work with for precise targeting within any given channel.
Possible approaches
- If a channel is technically cheap but the conversions there are real and count toward the right goal, leave it alone. That's expected marginal cost optimization behavior, not a problem.
- If conversions exist but their assigned value undersells their real importance to the business, fix the conversion value setup before changing anything else. An undervalued conversion can artificially depress the model's view of a channel.
- If the concern is brand fit rather than efficiency, use the content suitability center and account-level exclusions rather than trying to solve a brand question through budget.
- If audience signals are too broad, sharpen them (more specific customer lists, search-term-based custom segments) to steer the model more precisely, instead of trying to cap a channel directly.
- Before writing off a channel as a leak, weigh its contribution in the context of the campaign's overall goal, not in isolation; a channel can look weak on direct conversions and still be adding value another way.