Extending the Ad Context Protocol with Upfront and Reverse Auction Capabilities

[fblanco]

Summary

This proposal extends the Ad Context Protocol (adCP) to enable automated, recurring upfront marketplaces conducted through two coordinated auction types — Regular (Forward) and Reverse (Surplus) — between Seller Agents (Publisher Ad Servers or Seller Representatives) and Buyer Agents (DSPs or advertiser systems).

The goal is to create a protocol-level mechanism for efficient, transparent, and low-carbon trading of advertising inventory for future delivery, using adCP primitives for discovery, bidding, deal creation, and reporting.

---

Motivation

Today’s upfront and guaranteed advertising deals are fragmented and manual.
They rely on direct sales teams or intermediary platforms that:

• Require custom integrations and manual setup.
• Add unnecessary latency and cost through repeated bidstreams.
• Restrict publishers’ ability to combine different types of data and formats effectively.

As the industry moves toward privacy-compliant and interoperable data models, there is a growing need for a protocol-based mechanism that allows publishers and advertisers to transact directly in predictable, automated marketplaces.

The Ad Context Protocol already defines a standardized and verifiable way to represent inventory metadata, data provenance, and deal contracts.
This proposal extends adCP with auction primitives that support all data activation types — contextual, first-party, and matched datasets — improving market efficiency while maintaining transparency and interoperability.

---

Rationale for Two Auction Types

A single auction format cannot efficiently accommodate all supply and demand conditions.
This proposal defines two complementary auction mechanisms:

1. Regular (Forward) Auction — Upfront Allocation

• Targets high-demand, premium inventory.
• Allows advertisers to secure inventory packages in advance, defined by any combination of contextual, first-party, or matched data.
• Provides predictable revenue for sellers and guaranteed delivery for buyers.

2. Reverse (Surplus) Auction — Demand Reallocation

Even after the forward auction, many publishers retain unsold inventory, and many advertisers still have available demand.
The reverse auction addresses this by matching demand packages from Buyer Agents with available supply.

In this phase:

• Buyer Agents publish demand packages — each representing an advertiser or campaign with defined constraints, targeting rules, and available budget.
• Seller Agents compete to fulfill those demand packages by offering qualifying impressions at the lowest acceptable CPMs.

This mechanism:

• Maximizes utilization across both supply and demand.
• Automates demand reallocation, minimizing manual intervention.
• Enhances liquidity for long-tail publishers and smaller buyers.
• Improves sustainability by replacing per-impression RTB transactions with bulk impression or demand allocations handled in advance.

Together, the two auctions form a closed-loop system — the forward auction clears premium demand, and the reverse auction efficiently clears the residual market.

---

[bokelley]

Great thoughts! The protocol today handles placing guaranteed and non-guaranteed buys (effectively bids for price-prioritized inventory) that are placed at low priorities in the ad server.

I think the surplus use case you propose is effectively handled by the non-guaranteed use case in the spec: the buyer shares their criteria and asks for ad products; the seller attempts to match these; the buyer then sets and adjusts a price across a portfolio of sellers to minimize overall cpm. Not a reverse auction mechanically but I believe equivalent minus some game-theoretical nuances.

The forward auction is such a fun idea. My concerns are 1) that it conflicts with how agencies think about investment and 2) the forecasting of delivery is very difficult across complex constraints. There are many theoretical solutions to 2, and of course Yieldex (which we bought at AppNexus) was designed to solve the forecasting bit, but concerned that we're missing a key technology here unless Microsoft wants to sell it back to us!

[fblanco]

I understand the forecasting part. That's always been a challenge. But I believe publishers can be conservative for a large amount of supply. They don't need to promise things that they are not super confident about.