Programmatic Guaranteed vs PMP vs Preferred Deal vs Open Auction: Which Deal Type to Use

Three questions separate the four types. Is the volume reserved? Is the price fixed? Who is allowed to bid? The table answers those and the practical questions that follow. Ad server priorities use Google Ad Manager as the example, because Google publishes its values (Google Ad Manager Help, line item types).

Every deal type below uses real-time bidding underneath. The difference lies in the terms the seller attaches to each bid request. The auction mechanics are explained in how real-time bidding works.

The open auction is the default. A seller offers an impression to every buyer the exchange allows, sets a floor price, and the highest bid above it wins. Most open auctions are first price, so the winner pays its own bid. A shoe brand chasing cheap reach across thousands of sites uses the open auction and lets its demand-side platform (DSP) decide impression by impression.

A private marketplace (PMP), also called a private auction, is an auction with a guest list. The seller creates a deal ID, invites some buyer seats and sets a deal floor, often higher than the open-market floor. Invited buyers compete on price above that floor. A sports publisher might invite five sportswear and drinks brands to bid on its marathon coverage at a $6 floor; that example is illustrative.

A preferred deal gives one buyer a fixed price and a first look before the auction, with no obligation to buy. If the buyer passes or bids below the agreed price, the impression drops to private and open auctions (Google Ad Manager Help, Preferred Deals). A bank might agree $12 per thousand for a news site's personal-finance section and take only the impressions that match its own audiences.

Programmatic guaranteed is a direct deal run through programmatic pipes. Buyer and seller agree a fixed price, a volume and a flight, and the publisher reserves that inventory. The buyer then bids on each impression it receives, at the agreed price, through its DSP. A carmaker that must own every first slot in a streaming service's broadcast of a cup final would book it this way.

Both fix the price; only one fixes the volume. That single difference changes who carries the risk.

With programmatic guaranteed, the publisher takes the risk of delivering the volume and the buyer takes the risk of paying for it. The OpenRTB standard says a bidder "must bid" on a guaranteed deal (IAB Tech Lab, OpenRTB 2.6). If the DSP filters out impressions with its own brand-safety or frequency rules, the deal under-delivers and both sides have a reporting problem.

With a preferred deal, the buyer keeps its freedom. It looks first and bids only when the impression fits, and the publisher sells the rest elsewhere. That makes preferred deals a good test before a guaranteed commitment. A travel brand might run a preferred deal for a month, check viewability and audience fit, then convert the best placements into a guaranteed buy for the summer season.

A PMP and the open auction share the same auction rule. The differences are the guest list, the floor and usually the quality of the supply.

Higher prices do not always mean worse value. The ANA's Q2 2026 benchmark found that its higher-performing advertisers paid more per thousand ($7.90 against $6.55) but less per quality impression ($11.94 against $13.80) (ANA via Fiducia, 6 August 2026). They also bought from about 19,000 fewer domains and apps. Those figures are not split by deal type, so treat them as a pointer, not proof.

The open auction still has a job. It finds new supply, sets a market price and gives small buyers access without negotiation. Many buyers run a PMP for proven publishers and a tightly filtered open auction for discovery. The cost side of both is covered in the ad tech tax explained.

Programmatic guaranteed keeps the commercial shape of a direct deal and changes the plumbing. IAB Tech Lab calls this "Automated Guaranteed". The inventory and price are guaranteed, and the campaign "runs at the same priority as other direct deals in the ad server" (IAB Tech Lab, OpenDirect, updated 8 July 2024).

The trade is control against cost. A brand that buys from a dozen publishers can cap frequency across all of them and see one report. In return it pays platform fees that a plain insertion order avoids. IAB Tech Lab released OpenDirect 1.0, its API for automating guaranteed buys, in January 2015 (IAB Tech Lab, OpenDirect). IAB Tech Lab's own Deals API announcement describes deal entry as still largely manual and prone to mismatches (IAB Tech Lab, 4 December 2025).

Deals travel inside the bid request, in a pmp object attached to each impression (IAB Tech Lab, OpenRTB 2.6). The snippet below is a trimmed, valid request from an illustrative running magazine. Slot 1 is a homepage billboard sold as programmatic guaranteed. Slot 2 is an article rectangle that carries a preferred deal and a private auction, and stays open to the open auction.

Read it field by field with the table. The definitions come from the OpenRTB 2.6 specification; exchanges can add their own rules on top.

Two notes from the specification help buyers. A DSP should return a deal bid and an open-auction bid when both apply, which the guidance calls "M+N" bidding (IAB Tech Lab, OpenRTB implementation notes). Returning one bid only can lose the impression. The notes also warn against sorting deal bids by price and cutting the list short, because that can starve a deal of delivery.

Deal IDs fail more often than people expect. At the Deals API launch, Magnite's Nick Allen said that "two-thirds of deals are configured with the desired supply, yet deliver no or very little revenue" (IAB Tech Lab, 4 December 2025). Typos, mismatched floors and targeting set on the wrong object are the usual causes. Test each deal in the planned OpenRTB bid request builder, and see OpenRTB 2.6 explained for the rest of the request.

Take the article rectangle from the snippet above, served through Google Ad Manager. The prices are illustrative; the order follows Google's published priorities (Google Ad Manager Help, priorities).

1. Guaranteed line items go first. A sponsorship or standard line item booked for this slot would serve before anything else. Programmatic guaranteed deals sit at this level.
2. The preferred deal gets its first look. With no guaranteed line due, the bank's DSP sees the impression at the agreed $12. The reader matches its audience, so it bids and wins.
3. The auction follows. Had the bank passed, the invited sportswear seats would bid above the $6 deal floor. Open-auction bids above the $2.50 floor would compete in the same unified auction.
4. Leftovers fall through. If nobody clears a floor, the slot goes to lower-priority demand such as house ads, which sit at priority 16 (Google Ad Manager Help, priorities).

The order explains how sellers price deals. A first look is worth a premium because it comes before the crowd. A guarantee is worth more because it comes before everyone, and because the publisher gives up the chance to sell those impressions elsewhere.

The OpenRTB implementation notes list anti-patterns that platforms have seen in deal bidding (IAB Tech Lab, OpenRTB implementation notes). Most stalled deals trace back to one of these:

• The deal ID sits on the wrong object. The guidance treats a deal ID as a "targeting token" for an order, line item or campaign. Tied to a whole buyer seat, it can switch on too many campaigns; tied to an advertiser, it can limit that advertiser to one deal.
• The private auction flag is ignored. When private_auction is 1, bids are restricted to the listed deals (IAB Tech Lab, OpenRTB 2.6), and open-market bids may be rejected.
• The prices do not meet. The DSP line item is priced below the deal's bidfloor, so every bid fails the floor check.
• Only one bid comes back. The DSP returns its open-auction bid and drops the deal bid, or the reverse.
• Filters fight the guarantee. A buyer's brand-safety list or frequency cap removes impressions the publisher reserved, and the guaranteed deal under-delivers.

A weekly check of bid requests, bids, wins and spend per deal ID catches most of these early. Plenty of requests with no bids usually points to targeting or price. Bids without wins usually point to the floor or to a higher-priority line.

Curation took the PMP idea and moved it to the middle of the supply chain. IAB Tech Lab defines it as "the selection and organization of inventory using technology and data, to create effective packages for advertisers through prepackaged or real-time operations" (IAB Tech Lab, 23 September 2025). In practice, a curator gathers impressions from many publishers, adds audience or quality data and sells the result as one deal ID that any DSP can bid on.

To the DSP, a curated deal looks like any other PMP deal in the pmp object (PPC Land, 15 September 2026). That is why curation pushed deal usage up so fast. The ANA's own release linked the shift to "curated, higher-quality environments" when PMPs reached nearly 88% of tracked spend (ANA, 14 August 2025).

Sellers and data owners built curation businesses around this demand. Magnite combined discovery, packaging and activation in ClearLine on 1 October 2025. Roku launched Roku Curate and Walmart Connect launched Connect Select on 27 April 2026. Magnite added 27 named partners to ClearLine Activate on 9 September 2026 (PPC Land, 15 September 2026).

Fees are the main concern. In Jounce Media's April 2025 benchmarking, about half of curated deal IDs carried no fee beyond the exchange's standard take (PPC Land, citing Jounce Media, 15 September 2026). About 34% used a percentage markup with a median of 14%, and about 10% charged a fixed CPM with a median of $0.48. Because the fee is usually built into the deal price, the bid request does not show it.

IAB Tech Lab's response is the Deals API. It syncs deal terms from supply-side platforms (SSPs) to DSPs and names the seller, packager and curator of each deal (IAB Tech Lab, 4 December 2025). Version 1.0 was finalised on 6 February 2026 (IAB Tech Lab, Deals API). Before buying a curated marketplace deal, ask who curated it, what fee type applies and whether the same supply is available more cheaply direct. The fee chain is tracked in the working media benchmark.

Match the deal type to what the campaign cannot do without. The examples are illustrative.

1. Start with delivery. If the campaign fails without specific impressions on specific dates, such as a product launch on a streaming final, book programmatic guaranteed. Connected TV (CTV) ad breaks add their own mechanics, covered in SSAI and CTV ad serving.
2. Then price certainty. If you want a known price on premium supply but cannot promise volume, ask for a preferred deal. Use it to test a publisher before any guarantee.
3. Then access. If you want quality supply at a market price, join or request a PMP. Ask the seller what the floor is and what the package contains.
4. Check curated options against direct ones. A curated deal can add data and reach across many sites. Compare its effective price with a direct PMP on the same publishers.
5. Keep an open-auction line. Run a filtered open-auction line for discovery and reach. Move publishers that perform into deals.
6. Review quarterly. Compare cost per quality impression by deal type, not just CPM. A seller planning its own deal desk can start from the private ad exchange setup guide.

Publishers face the mirror choice. Guaranteed deals give revenue certainty but lock inventory and need forecasting. Preferred deals and PMPs keep flexibility. The publisher independence hub and the real-time bidding (RTB) architecture hub cover the seller's tooling.

1. IAB Tech Lab, OpenRTB 2.6 specification (Pmp and Deal objects), GitHub (release 2.6-202606, 11 June 2026). https://github.com/InteractiveAdvertisingBureau/openrtb2.x/blob/main/2.6.md
2. IAB Tech Lab, OpenRTB release 2.6-202309 notes, GitHub (28 September 2023). https://github.com/InteractiveAdvertisingBureau/openrtb2.x/releases/tag/2.6-202309
3. IAB Tech Lab, OpenRTB 2.x implementation notes, section 7.3 "PMP & Direct Deals", GitHub (accessed 9 November 2026). https://github.com/InteractiveAdvertisingBureau/openrtb2.x/blob/main/implementation.md
4. ANA, "The ANA Q2 2025 Programmatic Transparency Benchmark Finds $26.8B in Wasted Programmatic Spend" (14 August 2025). https://www.ana.net/content/show/id/pr-2025-08-programmatictrans
5. Fiducia, ANA Q3 2025 Programmatic Transparency Benchmark release (5 November 2025). https://www.fiducia.eco/post/the-ana-q3-2025-programmatic-transparency-benchmark-reveals-13-6b-in-efficiency-gains-as-programmat
6. Fiducia, ANA Q2 2026 Programmatic Transparency Benchmark release (6 August 2026). https://www.fiducia.eco/post/ana-programmatic-transparency-benchmark-q2-2026-working-media-efficiency-reaches-a-record-45
7. Marketing Dive, "Programmatic ad spend grows alongside efficiency gains: ANA" (16 December 2024). https://www.marketingdive.com/news/programmatic-efficient-transparent-ctv-marketing-ana/735645/
8. eMarketer, "US Programmatic Advertising Forecast and Ad Tech Trends H2 2026" (17 July 2026; figures verified via search snippet). https://www.emarketer.com/content/us-programmatic-advertising-forecast-ad-tech-trends-h2-2026
9. Google Ad Manager Help, "Line item types and priorities" (accessed 9 November 2026). https://support.google.com/admanager/answer/177279?hl=en
10. Google Ad Manager Help, "Add proposal line items" (accessed 9 November 2026). https://support.google.com/admanager/answer/9789673?hl=en