New Tab and Extension Inventory Buying, Consent-First

An extension audience decays from the day it installs, and new-tab inventory is almost always priced as though it does not. We price a buy on the age and retention of the install cohort sitting behind the placement, not on the network's flat CPM. Below 35 percent thirty-day retention we treat the supply as a burst and cap the flight at two weeks; above it we will negotiate a direct rate with the publisher. That is the whole argument, and it is why our rate card here looks nothing like a display plan. What we buy, precisely: The extension's own surfaces. The new tab page it legitimately overrides through chrome_url_overrides, its side panel, its onboarding and options pages. Not units injected into a third-party page the user opened for an unrelated reason. Under Manifest V3 the difference is visible in the manifest and in the host permissions, so before a test we ask for the manifest, the Web Store listing ID and the Edge or AMO equivalents. Supply that arrives without a listing ID attached is supply nobody will stand behind. Diligence before the first dollar: Listing age and review velocity, and the shape of the review text rather than the star average. Install channel, named: organic store search, paid installs, cross-promotion from another extension, or bundling. Requested permissions read against the stated single purpose. Whether a default search change is disclosed on a consent screen the user actually sees. Publisher-reported uninstall rate compared with the decay we observe in our own delivery, which is the number that catches an inflated deck. Publishers who will not name the install channel do not get bought, and that rule has cost us inventory we wanted. How the flight runs: Cohort-priced, one new-tab unit per session, nothing on the first tab a user opens after launching the browser. Creative built to the surface rather than resized from a display kit, because a 300x250 dropped into a new tab reads as malware to the user and performs like it. Measurement runs on a first-party click ID with a server-side postback; extension contexts break third-party cookies often enough that a pixel-only setup will under-report by a margin nobody can reconcile at quarter end. What we refuse: Bundled installs and sideloaded distribution. Pay-per-install partners. Extensions that change the default search engine or homepage without a consent screen. Injected advertising over publisher content, which takes revenue from a site that did not agree to the trade. Any supply whose chain we cannot follow back to a named store listing. We also decline briefs that want this reported inside the display line so nobody upstairs has to be told which channel it is. If the channel cannot be named in the board pack, do not run it. Not included: Building or acquiring an extension. Store submission and policy review for a client's own extension is separate work we do not take. We do not write the consent screens for publishers we buy from, though we will refuse the buy if theirs is inadequate. Who this is not for: Performance teams who need spend live this week. Diligence on a new publisher takes about ten working days and we will not shorten it. Advertisers who need guaranteed monthly reach, since cohort decay makes that guarantee a fiction that someone eventually has to explain. Brands in regulated categories where an unexpected placement screenshot becomes a compliance incident: the surface is legitimate, but the supply chain has fewer people checking it than you are used to, and we would rather you spent the money elsewhere than assume we can make that risk disappear.