Ad Server Build and Trafficking: GAM, Kevel, Adzerk

Of the last twenty-three Google Ad Manager networks we took over, seventeen had guaranteed sponsorship line items running at Standard priority instead of Sponsorship. The deals the sales team had signed were losing auctions to remnant demand, and the publisher was paying make-goods on inventory it had effectively sold twice. Ad serving faults are rarely exotic. They are priority, taxonomy and ads.txt, and they compound quietly. Google Ad Manager: We rebuild the ad unit hierarchy so it maps to how the site is organised now rather than how it was organised when the network was created, then set a key-value taxonomy meant to survive the next reorganisation: one naming rule, written down, with the values that must never be free text. Line item priority is set deliberately, sponsorship guaranteed at four, standard guaranteed at eight, price priority and network at twelve, house at sixteen, and nothing is promoted because somebody escalated. Forecasting is only as good as the targeting keys, so those come first. Prebid is wired with price granularity that matches your floors and a timeout taken from real device latency on Etisalat and STC rather than a default copied out of a US deployment. Lazy load fetch margins are set against actual scroll behaviour. Then ads.txt, app-ads.txt and sellers.json are reconciled against every SSP seat you genuinely hold, and the reseller entries nobody remembers authorising come out. Kevel and Adzerk: Server-side decisioning for sponsored placements that live inside a product, promoted listings in a marketplace, native units in an app, a sponsored row in a catalogue, where a client-side tag is the wrong shape and gets blocked, mis-measured or renders after the row it belongs to. We define the decision request contract, the flight and pacing rules, priority and frequency inside the decision engine, and the reporting fields your warehouse needs. Then we hand the endpoint to your engineers with the failure modes written down. We build against the API. We do not become its owner. Consent and regional signals: TCF 2.2 and GPP strings passed intact down every demand path and verified at the SSP rather than assumed from the consent platform's own dashboard. UAE and Saudi data protection considerations for first-party identifiers and cross-border transfer, decided with your counsel and then implemented, in that order. What we refuse: We will not run a network for a site whose inventory is also sold through a path its own ads.txt does not declare. We will not build a Kevel integration for a team with no engineer to own the endpoint after we leave, because it will rot within two quarters and the failure will look like ours. We do not take header bidding yield chasing as a standalone engagement, since moving an RPM a few percent by shuffling partners and timeouts only pays when the plumbing beneath it is already right. And we do not inherit a live network without a read-only period first, usually a fortnight, because a change made before we understand the traffic is a guess with your revenue behind it. Not included: Creative production, buy-side campaign management and DSP trading are separate. Reporting lands in your warehouse through the Ad Manager BigQuery transfer. We do not build dashboards, and we will point you at whoever should. Who this is not for: Anyone looking for a party to hold ad server credentials permanently and answer tickets. We build it, document it and train the person who will own it. If nobody on your side is going to own it, the network is back where it started inside a year, and neither of us wants that engagement on the record.