Ad Manager rebuilt from the inventory taxonomy up: under forty ad units, one key-value schema, floors in pricing rules, Data Transfer into your own BigQuery.
About this service
Ad Manager networks that can still be operated in year three carry fewer than forty ad units. The ones we inherit average several hundred, with six spellings of the same key-value and nobody left who knows which reports depend on them. The taxonomy is the first thing we rewrite, because pricing, reporting and every future header bidding change sit on top of it.
Inventory:
The ad unit tree maps to page templates and slot positions, never to who sold what in 2019. Sizes belong to the slot, not to the unit. Placements exist only where a salesperson genuinely sells across the tree, which in practice is two or three of them. Labels carry the competitive exclusions. For a banking title, no competing lender in the same page view is a clause in the insertion order, and it has to survive as a label, a line item setting and a QA step rather than as tribal knowledge held by one trafficker.
Key-values:
One schema, written down before a line item exists: page template, slot position, section, content signal, consent state, and the hb_ keys reserved to Prebid. It is published as a document your developers and your sellers both read, and we delete the twenty keys nobody has targeted in a year. Ad Manager charges nothing for unused keys. The cost is paid in reporting people quietly stop trusting.
Pricing:
Floors go into Unified Pricing Rules by geography, device and format. They do not go into individual line items, and optimised pricing stays off, because a floor you cannot read is not a floor and the first time yield drops you will want to know which number was in force. Where your Prebid version and your Ad Manager entitlements allow header-bid demand to be priced through pricing rules rather than a generated ladder of price-priority line items, we will argue for it. Where they do not, you get the ladder generated by script, with granularity, currency and the rounding rule documented.
Data you keep:
Data Transfer files to Cloud Storage and then BigQuery: NetworkImpressions, NetworkBackfillImpressions, NetworkRequests. That is the only place an argument with an SSP dashboard can be settled, which is why it belongs in setup rather than in a later project. You also get the daily pull built on the Ad Manager API, ReportService with a PQL query set, so finance stops exporting spreadsheets by hand.
What we do not do:
We take a fee, never a share of revenue, and no commission from any SSP, so nothing here contains a recommendation we are paid to make. We do not switch on Open Bidding by default: it settles inside Google's auction and your log-level view of those bids ends there. We do not write creative, we do not sell your inventory, and we do not run your network under our own MCM parent unless you ask and we agree it is the cheaper structure for you rather than for us.
Who should not book this:
Sites running AdSense only, or under roughly two million monthly ad impressions. Ad Manager will cost more in operational overhead than it returns, and we say so on the first call rather than after the invoice. Publishers who want the current network tidied without touching the taxonomy: that is a fortnight of somebody's time and the next redesign will undo it. Anyone who needs this finished inside three weeks, since the parallel reporting period alone runs longer than that.