I think marketing teams should keep one written copy of how they define a conversion, because standards bodies and analytics vendors are each writing their own.
Two places the definition is being written
On November 3, IAB Tech Lab finalized guidelines for sending ECAPI events to data clean rooms. The same post describes a core value of the ECAPI specification as “the standardization of the data attributes and definitions used to measure conversions.” Event names, user fields and event types now have a shared form, and that form is not yours.
The same day, Chord announced its acquisition of Zenlytic and described a problem it says the combination solves. Its release says that without shared definitions and controls, “teams can receive conflicting answers from the same data.” It also says Chord keeps definitions, metrics and financial goals current as they change. That is a second place where your definition of a customer or a conversion gets written, this time by an analytics agent.
Both efforts are sensible. A marketing leader who relies on either one without a written copy of their own is trusting that the definition each one holds matches the one finance, sales and the CEO use.
What I would write down
The document I am asking for fits on a page and has three parts. First, the list of events you count: the name, the funnel stage, and the system of record. Second, the rules around them: the attribution window, how duplicates are handled, and which events are excluded. Third, an owner and a review date.
The IAB Tech Lab launch post for ECAPI makes the first part easier. It says advertisers “don’t need new events, just alignment with the standard.” Your list is the thing you align. Without it, the mapping exercise the post describes starts from whatever each platform happens to be sending today.
The strongest objection
The best argument against this is that vendors maintain definitions better than a team document will. A page in a shared drive goes stale in a quarter, while Chord says its platform keeps context current, and a standard goes through public comment. On that view a team’s own copy duplicates work that someone else is already doing.
My answer is that the vendor’s own release says definitions change, and a change somebody else makes is a change you did not review. The Chord CEO, Bryan Mahoney, framed the risk of handing context to others in the release: “We saw what happened in commerce when brands became dependent on Facebook and Google to reach their own customers.” He was describing enterprise data and the AI vendors that use it, and the same logic applies one level down, to the definitions inside the analytics tool you rent. A written copy with an owner and a review date costs an hour a quarter. It also gives you something to compare against when an agent’s answer looks different from last month’s.
The Vibe example shows what that comparison can look like. When the self-serve CTV platform added Walmart Audiences, it described an agent called Smart select that reads four inputs: the advertiser website, the category and any category rules, the campaign creatives, and the budget and schedule. Vibe says relevance scores are “an estimate of how closely a segment aligns with the inputs the agent considered.” Its post states the inputs and the limit of the score, which is the level of disclosure to ask for from every agent that touches your numbers. Vibe also says nothing goes live from Smart select on its own: the picks land in the audience list, where the advertiser can remove or add segments before launch. That review step is the same job a written definition does for analytics. It gives a person something to check the agent against before money moves.
I made a related request for retail media in Ask Retail Media Networks for Dated Measurement Promises Before 2027 Budgets, and the argument for approval steps in Own the Approval Step Before Agents Own Your Stack rests on the same point. A control you wrote and own is the one you can enforce.
What I would do on Monday
Pull the conversion events your top three platforms receive today. Compare them with the one-page list. Where they differ, decide which side is wrong and write the decision down. Then ask each analytics or data vendor where its copy of the definition lives and how you would see a change to it.
Source: Chord, PR Newswire
