How do I find underperforming Ad Manager inventory?

Rank ad units by both revenue and eCPM over a recent window — the lists differ — then flag bottom-decile units that still carry meaningful impressions. Those are the ones occupying space without earning it.

Can an AI agent run Ad Manager reports?

Yes. With a read-only connection it can run the saved reports defined in your network; composing arbitrary ad-hoc dimension and metric combinations through the API requires a read-write connection.

Can it track MCM child publishers?

Yes — child-publisher readiness and MCM earnings can be read alongside the network's own numbers, so roster status and revenue trends come back in one answer rather than a per-publisher tour of the UI.

What about brand safety and ad review?

Review Center ads can be searched and blocked. The value is running it as a scheduled question rather than as incident response after someone complains.

Does Ad Manager show whether my inventory is profitable?

It shows revenue, not margin. If you buy traffic to monetised inventory, profitability needs Ad Manager revenue joined against acquisition spend — two systems that don't meet anywhere inside Google's own stack.

Playbooks

Ad Manager yield analysis with an AI agent: inventory, sites, and MCM publishers

TL;DR

Ad Manager reporting can answer almost anything, which is exactly why the useful questions go unasked — composing them is work. This is the set worth running regularly, from inventory ranking through site health to the MCM roster, plus the Review Center check most networks forget.

Rank inventory on evidence

1"Rank my ad units by revenue and by eCPM over the last 30 days." The two lists differ, and the gap is where inventory is high-volume but low-value.
2"Flag ad units in the bottom decile that still carry meaningful impressions." These are the candidates for consolidation — inventory occupying page space without paying for it.
3"Break revenue, impressions, eCPM, CTR and fill rate down per site." Site-level fill collapse is a distinct failure from site-level demand weakness.
4"Which sites are flagged with problems?" Approval and serving problems quietly cap revenue, and they're easy to miss in a large network.

Saved reports versus ad-hoc

Worth knowing before you plan: composing arbitrary dimension and metric combinations through the API needs a read-write connection. With read-only access, the agent can run the saved reports already defined in your network — which is often enough, since most networks have their standard views saved already. If a question needs a combination nobody saved, that's when the write scope matters.

The MCM roster

For networks managing child publishers, the recurring questions are readiness and trend: which publishers are onboarded and serving, which are earning less than last period, and where the roster's revenue actually concentrates. Reading child-publisher status and MCM earnings alongside the network's own numbers puts the whole picture in one answer instead of a per-publisher tour of the UI.

The Review Center check nobody schedules

Ad review is the part of Ad Manager that only gets attention after a complaint. Searching what's being served and blocking what shouldn't be is a genuine brand-safety task, and running it as a periodic question rather than an incident response is the difference between finding a problem and hearing about one.

Where the real margin question lives

None of the above tells you whether the inventory is worth the audience you paid for. For publishers buying traffic, the number that matters joins Ad Manager revenue against acquisition spend — two systems that don't meet anywhere in Google's stack. That's the same structure as optimising on profit instead of ROAS, applied to the sell side.

Put it on a schedule

Inventory decays quietly: a unit stops filling, a site develops a problem, a publisher drifts down. Monthly inventory ranking, weekly site-health checks and alerts on revenue movement turn that from something you notice late into something you're told. Keep the standing view on a self-refreshing dashboard so anyone who needs the numbers isn't asking you for them.

Frequently asked questions

How do I find underperforming Ad Manager inventory?: Rank ad units by both revenue and eCPM over a recent window — the lists differ — then flag bottom-decile units that still carry meaningful impressions. Those are the ones occupying space without earning it.
Can an AI agent run Ad Manager reports?: Yes. With a read-only connection it can run the saved reports defined in your network; composing arbitrary ad-hoc dimension and metric combinations through the API requires a read-write connection.
Can it track MCM child publishers?: Yes — child-publisher readiness and MCM earnings can be read alongside the network's own numbers, so roster status and revenue trends come back in one answer rather than a per-publisher tour of the UI.
What about brand safety and ad review?: Review Center ads can be searched and blocked. The value is running it as a scheduled question rather than as incident response after someone complains.
Does Ad Manager show whether my inventory is profitable?: It shows revenue, not margin. If you buy traffic to monetised inventory, profitability needs Ad Manager revenue joined against acquisition spend — two systems that don't meet anywhere inside Google's own stack.