Curation and Supply-Side Targeting

For fifteen years the division of labour in programmatic was settled. The sell side decided what inventory to offer. The buy side decided whom to target. SSPs moved impressions; DSPs applied intelligence.

That division is coming apart, and the move is the most consequential structural change in programmatic since header bidding. If you only follow one current trend in this curriculum, follow this one.

The thing that is actually happening

Start with facts rather than vendor narrative.

• Supply-side targeting spend doubled year over year in 2025 (OpenX, Supply-Side Intelligence Trend Report, October 2026).
• Microsoft wound down the Xandr Invest DSP while retaining its sell-side Monetize and Curate products.
• Direct DSP-to-publisher paths and curated marketplaces are projected past 35% of programmatic display spend by the end of 2026.
• The average RTB-enabled publisher now integrates with 25–30 SSPs, up from about 24 the year before, authorising 16–17 to initiate resold auctions (Jounce Media, 2026).

A company with Microsoft's resources concluding that the sell side is where value accrues, and shutting a DSP to prove it, is the single loudest signal in that list.

Curation is not supply path optimisation

These get conflated constantly. They are related and distinct, and the distinction is the article.

Supply path optimisation is a buy-side discipline: of the many routes to this impression, which do I allow? Score each path by win rate, fee transparency and fraud, then block the bad ones. SPO removes paths.

Curation is a sell-side product: a pre-assembled package of inventory, delivered as a Deal ID, that a buyer can activate in one line instead of constructing targeting themselves. Curation creates a path — a better one, in principle.

OpenX draws a further line inside curation that is worth adopting, because vendors blur it on purpose:

• Curation proper is inventory selection. Which sites, which formats, what viewability threshold, which brand-safety floor. Choosing the shelf.
• Supply-side targeting goes further: audience data, quality signals and predictive models applied at the SSP to decide which impressions within that inventory to surface, and at what price.

The second is the one that takes work away from the DSP. Ask which of the two a vendor is selling you, because the fee is attached to the second and the reassurance is attached to the first.

Why it is happening now

Four forces, and they compound.

1. The buy side ran out of signal

Covered in The Cookie That Did Not Die: a DSP's traditional advantage was a third-party cookie graph it had built. On 30–50% of traffic that graph is empty.

The publisher's advantage did not erode the same way. A logged-in reader is a first-party relationship, and it survives every browser change. Signal moved to where the authentication is, and authentication lives with the publisher.

2. Header bidding made the sell side the real decision point

Header bidding was supposed to give publishers leverage by making SSPs compete. It did — and it also meant the publisher's wrapper became the place where a single impression is offered to 25–30 parties at once. If one company sits at the point where every bid for an impression converges, that company is positioned to decide things, not merely to relay them.

3. Fee pressure pushed SSPs to find new revenue

A pure auction-running SSP charging 15% has a problem: the service is a commodity, SPO has made buyers ruthless about comparing paths, and the fee only goes down. Selling an audience layer on top of the auction is a higher-margin product that SPO does not commoditise, because it is differentiated by data rather than by price.

This is why SSPs have been signing data partnerships with the likes of Experian and TransUnion — building, in effect, their own data marketplaces where buyers pay for an extra audience layer on top of the media.

4. Clean rooms made the match legal and auditable

A publisher and an advertiser can now compare audiences without either handing over a file, which is what makes curated audience deals a thing a compliance team will approve. Clean rooms moved from infrastructure to the workspace where these deals are actually assembled.

What a curated deal looks like in practice

Mechanically, nothing exotic — which is part of why adoption is fast.

1. A curator (SSP, data company, or agency operating in the SSP's curation tool) defines a package: inventory list, audience layer, quality thresholds, price floor.
2. The platform mints a Deal ID.
3. The buyer adds that Deal ID in the DSP. One line.
4. Bid requests matching the package arrive carrying the deal, usually in the pmp.deals array of the OpenRTB request.
5. The buyer bids against the deal rather than assembling site lists and segments.

The buyer's targeting complexity collapses from a hundred line items to one. That convenience is real, and it is most of why this is winning.

What you give up

• Transparency inverts. You see the package's performance, not the composition decisions behind it. The curator knows why an impression was included; you do not.
• Verification is harder. You are trusting the curator's quality claims about inventory they also profit from selling. Self-marking homework is a structural conflict, not an accusation about any one firm.
• Attribution of credit gets murky. When a curated deal outperforms, is that the audience layer, or just better inventory that would have outperformed under any targeting? Only a holdout answers that, and curators rarely offer one.
• Portability is gone. A Deal ID is specific to a curator. Performance you build inside one is not a thing you can take elsewhere.

How to evaluate a curation partner

Six questions. The useful ones are uncomfortable, which is the point.

1. Which fee disappears? As above. Get it in basis points, not adjectives.
2. Curation or supply-side targeting? Inventory selection, or audience decisioning? Price the second separately.
3. Where does the audience data come from, and what is the match rate on my inventory mix? A platform-wide match rate is marketing. Ask for yours.
4. Will you supply a no-targeting control deal? A no here tells you how confident they are.
5. What is the reseller depth inside the package? Curation that bundles resold inventory has reintroduced the hop SPO was built to remove.
6. Can I see the inventory list? If the composition is a secret, the quality claim is unfalsifiable.

The honest uncertainty

Two readings of this trend are live, and nobody can yet settle it.

The optimistic reading: the supply chain is consolidating around fewer, better paths. Signal sits with the publisher, decisioning moves next to the signal, hops get removed, and the buyer gets a simpler and cheaper product.

The pessimistic reading: the DSP's fee is being duplicated rather than replaced, inventory transparency is being traded for convenience, and the industry is reinventing the blind network it spent fifteen years dismantling — this time with a clean room attached.

The evidence currently supports both. What settles it per campaign is the control deal. Insist on one.

Key takeaways

• Curation is a sell-side product that packages inventory as a Deal ID; SPO is a buy-side discipline that removes paths. They are not the same thing.
• Supply-side targeting goes beyond curation by applying audience data and models at the SSP — this is the part that displaces DSP work, and the part that carries the fee.
• It is happening because buy-side signal eroded, header bidding concentrated decisioning on the sell side, SSP fees were commoditised, and clean rooms made the audience match auditable.
• Always ask which fee the curated path removes. If none does, it is a cost increase.
• Always ask for a no-targeting control deal on the same inventory. If the curated deal cannot beat it, you are paying for a site list.