If a measurement service cannot reproduce the decision it was meant to support, the dashboard is not the deliverable. It is only the interface.

On September 22, 2026, the IAB released the Measurement Services Addendum v1.0 for public comment. It is designed for relationships involving advertisers, agencies, publishers, platforms and providers of measurement, verification, attribution and analytics. The draft works with the IAB’s modular General Terms v1.1, and comments remain open through October 22, 2026.

This needs a careful boundary. The document is a draft industry framework, not law, and its publication does not prove universal adoption. Its structural importance is narrower and more useful: measurement is moving from a feature discussion into procurement language. The buyer increasingly needs to specify what is measured, which inputs are valid, how methods change and who is accountable when two systems disagree.

A dashboard can display an answer without making it decision-ready

The same reported ROAS can be built from different attribution windows, customer definitions, device matching coverage and refund rules. Two identical numbers may not justify the same budget move.

Platform documentation makes the problem tangible. Google notes that some asset-level metrics are non-summable and that ratios such as CTR, CPA and ROAS should be treated as directional because assets are served in combinations. Campaign Manager 360 separately publishes a measurement methodology, data-retention rules and a notification process for methodology changes.

The procurement question is therefore not, “Can the vendor show creative performance?” It is, “Can this evidence support a named action—pause a version, fund another shoot, move budget across markets or reconcile an invoice?” If that action is not defined in the scope, the reporting layer may be informative without being operational.

The ONCE Measurement Decision Contract

This seven-part structure is an ONCE procurement framework, not a summary of clauses in the IAB draft and not legal advice. It can inform an RFP, statement of work or kickoff, with local legal, privacy, finance and media review.

1. Decision

Name the action the work must support: increase spend, retain a creative route, stop a channel, commission new production or conduct a quarterly review. Do not buy continuous measurement for a question with no owner or action.

2. Outcome definition

Specify the event, value logic, attribution window, customer lookback period, cancellation and refund treatment, and the hierarchy between engagement, performance KPIs and business outcomes.

3. Data ledger

List exposure, site, order, CRM, offline and third-party inputs with market coverage, refresh cadence, exclusions and known gaps. Missing or non-exportable data belongs beside the conclusion, not buried in a methods appendix.

4. Method card

State whether the service uses platform attribution, third-party attribution, incrementality testing, MMM or a combination. Record assumptions, sample requirements, control design and failure conditions. Project Eidos is explicitly about shared structure and comparability; it does not mandate one model or vendor.

5. Version chain

Keep data version, model version, rule changes and rerun dates. For video and dynamic ads, map the master, edit, language, market, rights state and platform asset ID. The IAB Tech Lab’s Ad Creative ID Framework exists so a creative can retain a persistent identifier across the video supply chain. Even teams that do not register a UniversalAdId should maintain stable internal IDs from production through trafficking.

6. Discrepancy rule

Define which source controls which decision, what variance triggers review, and who must rerun, explain or suspend a conclusion when platform, ad-server, measurement-provider and commerce data diverge.

7. Handoff and exit

Require usable exports, field definitions, query or configuration snapshots, methodology versions, change logs and retention periods. At contract end, the brand should be able to reconstruct important conclusions without an active vendor login.

Price the service by usable decisions, not report volume

A practical comparison is:

Cost per usable decision = (provider fees + data onboarding and cleaning + creative/media ID mapping + reruns + human review and dispute handling) / reproducible decisions with an assigned action owner

A decision counts only when inputs are traceable, definitions are agreed, uncertainty is visible and a result triggers a named action. More pages, charts or API endpoints do not increase the denominator.

This also changes production handoff. If one product shoot becomes multiple hooks, durations, languages and market versions, the production partner should not deliver a folder called “finals.” It should deliver a version ledger connecting every file to its claim, shots, language, rights status, media ID and launch date. Otherwise the media team may identify a “winner” that nobody can accurately reproduce.

When to keep measurement simple—and when to pause

Native platform reporting may be enough when one platform, one market and a short optimization cycle are involved, and the result will only drive in-platform creative rotation.

A limited external measurement pilot is appropriate when the same creative system runs across platforms or markets, stable business outcomes exist, data can be exported and creative versions have unique identifiers.

More tooling should wait when marketing and finance disagree about “new customer,” revenue or valid order; versions cannot be matched to delivery; critical data cannot be exported; or the provider cannot explain methodology changes and rerun conditions. Adding another model in that situation usually produces conflicting answers faster.

The IAB draft may change before finalization. The useful move today is not copying model language into every contract. It is asking whether the brand is buying a data display or a reproducible decision that can change media, creative and production plans.

This article also draws on the IAB’s State of Data 2026, Campaign Data Standards 1.0 and the IAB Tech Lab ACIF materials. It is commercial and production analysis, not legal advice.