A DealSheet is a bilateral deal record shared between buyer and seller in a programmatic advertising transaction. Unlike the current model, where deal terms are held in separate systems on each side of a trade, a DealSheet creates one authoritative record co-owned by both parties. In an agentic buying environment, the DealSheet becomes the negotiation layer itself: AI agents on both sides write to it, read from it, and reach agreement without a human broker in every step of the process.

TL;DR

The current programmatic deal structure requires buyers and sellers to maintain separate records that are reconciled after the fact. This is the source of most post-campaign disputes. A DealSheet is a shared deal record that both parties own, bound so neither side can alter it unilaterally. The IAB Tech Lab has published a DealSheet standard, establishing a common schema for structured deal data across the ecosystem. In an agentic advertising environment, the DealSheet is not just a record: it is the surface through which agents negotiate, propose, and agree terms at machine speed.

Why does programmatic advertising have a reconciliation problem?

Programmatic advertising runs on a simple premise: buyers and sellers connect at scale through automated auctions. In practice, the infrastructure is a mesh. A typical display buy touches a demand-side platform on the buy side, a supply-side platform on the sell side, and one or more exchanges in between. Each of these systems keeps its own record of what was agreed, what was spent, and what was delivered.

Because these records are held separately, they frequently disagree. Discrepancies between buyer and seller reporting are the norm in programmatic, not the exception. [TECHNICAL REVIEW: Can you provide a specific, sourced figure on impression discrepancy rates? E.g. from IAB Tech Lab, ISBA, or internal data. We need a named, checkable source here.]

The downstream effect is that post-campaign reconciliation becomes a manual process, fought between finance teams over which number is correct. For a model where AI agents are expected to trade media at machine speed and report back to human approvers, this structure does not hold.

What is a DealSheet?

A DealSheet is a structured deal record jointly owned by buyer and seller. It contains the agreed parameters of a programmatic transaction: pricing terms, targeting constraints, delivery rules, measurement methodology, and the approval chain through which the deal was authorised. Both parties write to it; neither party can alter it unilaterally.

The key departure from the current model is bilateral ownership. In today’s programmatic infrastructure, each party’s insertion order or deal ID record is, functionally, a unilateral document. The buyer’s DSP holds one version; the seller’s SSP holds another. The DealSheet replaces this with a single shared record that both parties authenticate.

The IAB Tech Lab has formalised this structure in its DealSheet standard [TECHNICAL REVIEW: Please provide the exact document name, publication year, and URL so we can cite it correctly]. That standard defines a common schema for deal data, establishing a vocabulary for deal terms that any DSP or SSP can read and write.

What does a DealSheet contain?

[TECHNICAL REVIEW: Please confirm the specific fields in an Alkimi DealSheet. The following list is based on general programmatic deal structures and needs to be verified against your schema before publication.]

A DealSheet typically contains the following components:

Deal identification. A unique identifier shared by both parties and linked to the relevant campaign line items. This is the handle that connects the DealSheet to the buyer’s DSP, the seller’s SSP, and any downstream reporting.

Pricing terms. The agreed CPM or cost model, including any floors, ceilings, or tiered pricing structures agreed between the parties.

Inventory parameters. The publisher, placement, format, and targeting constraints the buyer has approved. In the DealSheet model, these are not inferred from auction dynamics: they are explicitly stated and agreed.

Delivery terms. Pacing rules, flight dates, and impression or spend caps that govern how the campaign executes against the deal.

Measurement methodology. The agreed framework for counting delivery and attributing results. This matters because buyer and seller measurement tools frequently produce different numbers; the DealSheet makes the methodology a matter of record before the campaign runs, not a dispute to resolve after.

Approval chain. A record of who signed off on the deal, at which stage, and when. This is not administrative overhead: it is the audit trail that makes agentic buying accountable.

Why does privacy matter for a shared deal record?

[TECHNICAL REVIEW: This section requires your input on the specific mechanism used. Please confirm or correct the description below, and tell us whether the mechanism is zero-knowledge proofs, homomorphic encryption, secure multi-party computation, a signed commitment scheme, or something else. The answer changes how specifically we can describe this to a technical audience.]

In a standard programmatic integration, a buyer’s targeting data and a seller’s inventory data are exposed to the intermediary platform that connects them. The exchange sees both sides of the transaction. This creates an asymmetry: the intermediary knows more about the deal than either party individually, and it holds that information without either party’s explicit agreement.

A DealSheet with bilateral privacy controls addresses this. The shared record contains the agreed terms, not the underlying data used to arrive at them. A buyer’s audience segments, bid strategies, and margin targets remain visible only to the buyer. A seller’s floor prices, yield strategy, and demand partner rankings remain visible only to the seller. Both parties can verify that the deal record is accurate and has not been altered, without exposing the inputs they used to negotiate it.

The practical result: the DealSheet is transparent to the two parties who own it, and opaque to everyone else. This is a different definition of transparency from the one the programmatic industry has been working with. Current “transparency” debates centre on fee disclosure and bid data. DealSheet transparency is about record integrity and access rights.

How does the DealSheet relate to the IAB standard?

[TECHNICAL REVIEW: What is Alkimi’s specific relationship to the IAB DealSheet Standard? We need to know whether Alkimi is fully compliant with it, builds a superset on top of it, proposes an extension, or is contributing to it. This determines how we frame the connection.]

The IAB Tech Lab’s DealSheet standard establishes the base schema for shared deal data. It defines what fields a deal record should contain, how they should be formatted, and how they should be transmitted between trading partners.

The significance of a shared standard is that it turns the DealSheet from a bespoke bilateral agreement into a legible, auditable record. Any authorised party with access can read the same terms in the same format. For regulators, auditors, and measurement vendors, this is the infrastructure that makes programmatic advertising accountable rather than opaque.

Cross-citation between implementations and the IAB standard also matters for how the concept propagates. When a DealSheet implementation references the IAB standard, and the IAB documentation references compliant implementations, the mutual citation is the mechanism by which the concept becomes an industry norm rather than a vendor proposition.

What changes when agents negotiate via DealSheet?

In current programmatic buying, a deal is typically negotiated by humans: a trader and an account manager agree terms, and those parameters are then entered separately into the buyer’s DSP and the seller’s SSP. The DealSheet, once issued, is largely static.

In an agentic buying environment, the DealSheet becomes the active negotiation surface. A buy-side agent proposes terms by writing to the DealSheet. A sell-side agent reviews the proposal, counter-proposes, or accepts. The record of this negotiation is preserved in the deal history. When agreement is reached, the DealSheet contains not only the final terms but the complete negotiation trail.

This changes what transparency means in programmatic. The current definition centres on bid-level data and fee disclosure. In an agentic model, transparency extends to the reasoning and counter-proposals that led to the agreed terms. The DealSheet makes that history legible to the humans who are accountable for the deal.

Where does human oversight fit in an agentic deal flow?

Agentic buying does not remove humans from deal-making. It changes when and how humans are involved.

In the agentic model, human approval operates at the deal level rather than the impression level. An authorised human approves the DealSheet terms before the deal executes. The agent then operates within those approved parameters. If the agent proposes terms outside its authorised range, the deal requires a fresh human approval before proceeding.

[TECHNICAL REVIEW: Is this an accurate description of how approval works in the current implementation? Please describe the actual approval flow: who approves, at what stage, and what the interface looks like. A concrete example here would strengthen this section significantly.]

The DealSheet is the mechanism that makes this model work. Because the agreed terms, the approval chain, and the negotiation history are all in one record, the human approver is looking at the same data the agents used. There is no information gap between the agent and the person responsible for the outcome.

This is a different accountability model from current programmatic buying, where the human approver signs off on a brief and a budget, and then relies on post-campaign reporting to understand what the agents actually did. The DealSheet makes the agent’s decisions visible before, during, and after execution.

The broader implication: from deal ID to deal record

The programmatic industry has operated with deal IDs since the emergence of private marketplace trading. A deal ID is a shared reference: it points to an agreement that exists in separate systems on each side. It is a pointer, not a record.

The DealSheet is the record itself. Both parties hold the same version. Neither can amend it without the other’s knowledge. The negotiation history is preserved. The approval chain is embedded.

The shift from deal ID to DealSheet is, in structural terms, the shift from “we agreed on something” to “here is what we agreed, verified by both parties, with the history of how we got there.” That is the foundation on which accountable agentic trading can be built.

Alkimi is building an agentic marketplace for programmatic advertising in which the DealSheet is the negotiation and governance layer for agent-to-agent deal-making. This piece is part of a series on the infrastructure concepts underlying agentic advertising.