A retailer may now return product-level conversion data to a brand. That is useful. It is not yet a creative learning system.
Consider a global brand working with a production partner in China, a media agency in the US and several retail partners. The studio delivers three product videos. The agency uploads each file more than once, platforms assign new ad IDs, and the retailer reports performance by campaign and SKU. Everyone has data, yet no one can reliably tell whether the winning execution was the installation demonstration, the battery claim or simply a discount that happened to run at the same time.
The missing object is a durable join key between production and media. Brands should not merely buy neatly named files. They should require a stable creative identity that survives handoffs and connects each final asset to the product, claim, delivery context and valid rights window.
Three standards shifts make this practical now
On August 10, 2026, IAB finalized Campaign Data Standards 1.0. Its complementary Ad Type and Ad Inventory taxonomies provide a shared framework for describing advertising experiences and inventory across channels and campaign systems. The standard does not prove universal adoption, but it shows that cross-channel measurement is moving toward a common descriptive layer.
Video has a parallel identity layer. The IAB Tech Lab's Ad Creative ID Framework, last updated March 18, 2025, defines a persistent identifier for video creative throughout the supply chain. Its stated uses include cross-platform measurement, frequency management and campaign reconciliation. In other words, a video needs an identity that can survive more than one platform upload.
Commerce media is opening the outcome side of the equation. Google's current commerce media documentation says participating commerce partners can selectively share feed inventory, conversion and first-party audience data, while advertising partners can receive product-level performance metrics, including conversions. Availability still depends on participating merchants, account setup and permissions.
Taken together, the industry now has better ways to describe the media environment, identify video creative and return commerce outcomes. What most brand production contracts still lack is the table that joins those layers.
A report can identify an ad and still lose the idea
A platform ad ID is not the same thing as a production asset identity. The same master can be uploaded twice and receive two ad IDs. A small subtitle or packaging change may create a new file that looks similar but contains a materially different claim. A retailer may report against a SKU while the video actually shows an older regional package.
The IAB/MRC Retail Media Measurement Guidelines recommend transparent methods and granular reporting by dimensions such as device, ad format and creative type, placement, domain or app, time, geography, audience segment and online versus in-store sales. The guidelines also connect attributable sales to specific SKUs in the creative. This makes retail-media reporting more useful, but the brand still has to define which production asset those dimensions refer to.
Without that definition, a creative review becomes a guessing exercise. Teams may attribute a sales result to a message when the real difference was audience, inventory, price, promotion or availability. They may also merge two different versions or split the same version into multiple false “winners.”
The ONCE Creative Evidence Key
The ONCE Creative Evidence Key is a brand-owned mapping record, not an IAB requirement. It gives each deployable video version five linked identifiers:
- Product key — SKU, model, pack or regional product version.
- Creative version key — master, edit branch and approved version. A new shot, voiceover, subtitle or CTA needs an explicit decision: new creative or technical transcode?
- Claim module key — the message actually expressed, such as installation, battery life, material, compatibility or price.
- Delivery context key — market, language, duration, aspect ratio, ad type and placement. External taxonomies can normalize this layer without replacing the production vocabulary.
- Validity key — the rights and fact window for talent, music, source assets, product claims and promotions.
One additional field maps external IDs: platform ad ID, campaign ID,
retailer product ID and, where available, a registered creative ID. A
compact internal record might look like
SKU-A17 / CR-042-V03 / CLM-INSTALL / US-CTV-15 / RGT-2027Q2.
This is an illustration, not an industry standard.
The test is simple: starting with one line in a retail-media report, can the team reach one approved file, one product version and one claim definition without asking three vendors to search old email threads?
Put the mapping into the production deliverables
For work intended for retail media, CTV, programmatic video or several agencies, the procurement brief should request five operational files alongside the masters:
- Asset manifest: one internal creative ID per deployable file, plus checksum, duration, aspect ratio, language and approval status.
- Claim-module register: the products, claims, shot ranges and CTAs actually present in each asset.
- External-ID map: completed by media or ad operations after upload with platform ad IDs, campaign IDs and retailer product IDs.
- Measurement dictionary: conversion definition, attribution window, online and in-store sales scope, promotional period and known blind spots.
- Version and rights log: every change to footage, packaging, copy, claims, talent or music, with explicit retirement status for the previous version.
The production company does not need shopper-level data. It needs to deliver an accurate asset structure. The brand or its authorized data partner performs the outcome join inside the permitted environment. That separation limits unnecessary data exposure and keeps media-account access out of production acceptance.
Teams that still struggle to locate the approved master should first establish a basic searchable naming and asset-management convention. The Creative Evidence Key begins where file organization ends: it preserves identity after the file enters trafficking, reporting and retailer systems.
Measure cost per traceable decision, not cost per file
A practical comparison does not require an invented industry benchmark:
Cost per traceable creative decision = (production + tagging and QA + trafficking map + data access and analysis + governance) / decisions linked from creative ID to outcome and used in the next brief
The denominator is not the number of dashboard rows or generated variants. A decision counts only if it changes a shot, claim, version or delivery condition in the next production cycle.
If a team tags 80 variants and learns only that “video performed well,” the denominator is close to zero. If 12 controlled versions establish that an installation demonstration should be retained only for mobile product-detail placements, the system has produced a usable production decision. Those numbers are hypothetical examples, not ONCE results or market averages.
When the extra operating layer is worth buying
The system is most useful when all three conditions apply:
- one SKU will run across several retailers, markets or media environments;
- the next production cycle will change creative based on commerce outcomes;
- the brand can obtain sufficiently granular campaign, product or creative reporting and preserve its definitions.
For a one-off launch with only a few versions, no follow-on production budget or only aggregate retailer reporting, start with a simple asset ID and version log. A complex taxonomy that no one maintains is not an asset.
Finally, attribution visibility is not proof of incrementality. Commerce outcomes can identify a creative hypothesis worth investigating, but causal claims still require a credible counterfactual and bias control. ONCE's guide to bringing incrementality into creative production covers that next step. For teams adapting one product story across display, video and audio, the companion analysis on cross-sensory asset design is also relevant.
The useful procurement question is no longer “How many videos will we receive?” It is: “When a retail-media report returns a sale, can we identify the exact product, approved asset, claim and valid version that earned another production decision?”
