Retargeting Pools Rebuilt Around Suppression and Decay

Most retargeting accounts I open have one pool, a membership window near the platform maximum, and no suppression of people who already converted. Twenty to thirty-five percent of the impressions are being served to buyers, enrolled students or booked travellers. Fixing that costs nothing, returns immediately, and is usually the largest single change available in the account. Pools are cut by intent depth and recency: A person who abandoned a fee payment yesterday and a person who read a course page in March are not one audience and should not share a bid, a frequency cap or a creative. The architecture separates them into tiers with their own membership windows, typically three, seven and thirty days for high-intent behaviour, with a longer tier only where the buying cycle genuinely justifies it. Windows are argued from your own lag between first visit and conversion, measured, not copied from a template. Suppression is treated as infrastructure: Converters, refund cases, open support tickets, unsubscribed contacts, and in education anyone a counsellor has already reached by phone. Suppression files refresh daily and the refresh is monitored, because a suppression list that silently stopped updating three weeks ago is indistinguishable from no suppression at all and nobody looks until the quarter closes. Where the CRM cannot export daily, the fix is scoped in the same engagement. Event collection that survives browser policy: Pools built only on a browser pixel are shrinking on Safari and iOS and will keep shrinking. Membership moves to server-side collection through a server container and the DSP's server APIs, with event parameters and deduplication defined so the same conversion is not counted twice across the ad server and the platform. Consent state travels with the event rather than being applied afterwards. Frequency, honestly: There is no shared frequency cap between DV360 and The Trade Desk. Anyone promising cross-DSP capping is describing an approximation, and I will show you what the approximation costs in duplicated exposure before you decide whether to run both. Within a platform, caps are set per pool and per creative, and view-through conversions are discounted rather than reported at face value. Vertical shapes: Education: retarget only between funnel stages, suppress on counsellor contact, and hold the pool through the results window rather than across the year. Travel: the search-to-book window runs roughly three to twenty-one days depending on route and season, and both booking and travel-completion suppress. Medtech: pools are built from unbranded and corporate pages only, never from condition, symptom or treatment pages, and gated professional portals stay out of the pool entirely. Not included: Creative and dynamic feed production, though I will specify the feed's rules. Media buying. Reworking your product analytics. Third-party audience purchase to inflate a pool, which is the opposite of the work. Who this is not for: Advertisers who want the pool made larger. Larger pools with looser rules are where retargeting spend goes to be reported as efficient and to be worth nothing. Teams unwilling to run a holdout, because retargeting is the channel where last-click credit and incremental value diverge most sharply, and an account that has never held a control group does not know which of its results are real. If you would rather not know, this engagement will be uncomfortable and you should not buy it.