Programmatic Audience Buying: The Data Cost Nobody Puts on the Media Plan

Targeting an audience and buying one are different transactions

It's easy to conflate audience targeting with audience buying, but they're separate steps with a separate cost attached. Targeting is deciding which audience a campaign should reach. Buying is acquiring the actual data that defines that audience — a segment of device or advertising IDs a data provider has classified as matching a set of criteria — and it isn't free. Third-party data providers license access to segments through a data marketplace, typically integrated into a DSP or a connected data management platform (DMP), and the cost of that license is added directly onto the media cost as a separate line, most often a flat data fee per thousand impressions.

The distinction matters most at the invoicing stage, where the two costs are easy to lose track of. A media plan showing "$5.00 CPM" as its line-item cost might actually be $5.00 in media plus $2.00 in stacked data fees, reported as a single blended number by a platform or agency that isn't required to itemize the two separately. Asking for the data-fee line broken out on its own, before approving a plan, is the only reliable way to see whether a targeting layer is worth what it's actually costing.

A worked example: what a segment actually adds to the bill

Say a DSP campaign is buying media at a $5.00 base CPM. Layering a third-party "in-market for home renovation" audience segment onto that targeting typically adds a separate data fee — commonly in the range of $1 to $3 CPM depending on the provider and how narrow the segment is — bringing the effective CPM to $6.00–$8.00 before any DSP platform fee is even applied. That data fee doesn't buy media; it buys the right to target people the provider has already classified into that segment. A campaign layering three separate third-party segments at $1.50 CPM each is adding $4.50 to the media cost before a single impression is served, which is worth knowing before assuming a targeting checkbox is free just because it's easy to click.

The pipeline: how a segment actually reaches a DSP

A DMP collects and organizes audience data from multiple sources — a publisher's own first-party data, a data provider's third-party panels, or an advertiser's own uploaded lists — and makes it available as a segment. When a campaign requests that segment, the DSP queries the DMP, the DMP identifies which device or advertising IDs match the criteria, and those IDs get passed into the ad exchange so the DSP knows which bid requests to respond to. The advertiser never sees the raw underlying data; it sees a segment size estimate and a price.

The segment size estimate itself deserves a second look before a campaign launches. A DMP typically reports an audience size based on its own matched device pool, which rarely aligns exactly with what a specific DSP or exchange can actually reach — the usable, addressable portion is often smaller than the headline estimate once match rates and inventory overlap are accounted for. Treating the DMP's stated segment size as the guaranteed reach of a campaign is a common planning error that shows up as under-delivery once the campaign is live.

Why the third-party half of this is shrinking

Third-party data — audience segments built by a data broker from sources the advertiser doesn't own — has been declining in reliability as browsers phase out third-party cookies and privacy regulation tightens the rules on cross-site tracking. That's pushed audience buying toward first-party data (an advertiser's own customer list) and second-party data (a direct data-sharing partnership with one other company, rather than a broad marketplace purchase), both of which tend to be more stable and often cheaper per impression than a broadly-licensed third-party segment, because there's no broker margin sitting between the source and the buyer.

The mistake: buying a segment before asking whether it's incremental

The recurring objection in nearly every conversation we have about audience-driven display spend — on Amazon DSP or anywhere else — is a version of "is this incremental, or would they have bought anyway." It's not really a measurement question; it's a trust question, and the honest answer starts with naming who designs and computes the test. A holdout run by the party being paid on the campaign's result isn't a real holdout. The genuinely useful version of this, applicable to any vendor including us, is to require that the test design, the holdout definition, and the reporting cadence get written into the scope of work before spend starts — not offered after the fact as a favor.

What to check before buying an audience segment

Check whether the segment is first-, second-, or third-party before pricing it — the reliability and often the cost differ meaningfully by source. Check the data fee as its own line, separate from media CPM, before approving a targeting layer — it's easy to approve a checkbox without seeing the dollar figure attached to it. Check whether the segment's performance has ever been tested against a holdout, not just reported on a last-click basis — a segment that looks efficient purely on delivered ROAS may simply be reaching people who were already going to buy.

Where Dr. DSP fits

Dr. DSP is Amazon DSP run as a managed product by Full Circle, which has managed more than $500M in revenue across 100+ brands. Amazon's own targeting draws heavily on first-party retail signal rather than a licensed third-party segment, which sidesteps some of the data-cost stacking described above — but the incrementality question still applies regardless of the data source. A reader who never buys anything from us should still leave this page knowing to ask for the data fee as its own number before approving any audience layer, on any platform.