A brand asks its China-based production partner to refresh a carousel campaign. The media report contains an ad summary and several product-card rows. One image has the lowest reported click-through rate, so the proposed scope is a new shoot. Before booking the studio, the team needs to establish what that row measures—and which delivered image it represents.
ONCE’s recommendation is to commission a specific information improvement. A card report can help locate a question; it does not, by itself, establish that photography caused the result or that the whole visual package needs replacing.
The documented capability has an important scope limit
Spotify’s Ads API v3 September 2026 release notes add carousel-card reporting dimensions and metrics. Its asynchronous report reference distinguishes card counts from creative totals: do not sum card impressions or clicks into those totals. Card CTR is clicks divided by impressions, with zero returned when impressions are zero. The position field describes configured position, not rendered position.
There is also a documentation discrepancy to preserve: the Carousel ads help page, checked on the same date, still says card-level reporting is unavailable. These are API definitions, not proof that every Ads Manager account or interface exposes the capability. Ask the media partner what it can actually retrieve before making card-level conclusions.
All three sources were checked November 3, 2026. This is an operating decision supported by existing documentation, not a claim of a universal launch today. None of these definitions proves that carousel beats video or that a particular image generated incremental sales.
Maintain ad results and component observations separately
Use the ad’s own metrics for the ad’s delivery and spend. Retain card rows as component observations. Do not append them to the summary as extra performance or force them to reconcile to a different counting unit.
Consider this illustrative report, entirely invented for explanation and unrelated to any client account or ONCE test:
| Image task | Card impressions | Card clicks | Card CTR | What the row supports |
|---|---|---|---|---|
| Whole product | 1,000 | 20 | 2% | 20 recorded card clicks |
| Construction detail | 100 | 4 | 4% | A smaller exposure base; percentage alone is not a winner |
| Packed-away use | 0 | 0 | 0% | No card exposure on which to judge the image |
Adding 20 and 4 yields a sum of those card rows, not permission to relabel it as the ad’s click total. Averaging 2% and 4% does not produce the ad’s CTR. A missing field should remain unavailable; it is different from a returned zero with no exposure.
Likewise, a card configured in position three does not prove a person saw it third. A report of configured positions is insufficient to tell a story about where viewers abandoned a sequence.
Build an ONCE card-to-production receipt
This is a proposed shared working record, not a Spotify feature. For each card, retain the ad identifier, product or card identifier, delivered image version, copy, destination, configured position, market and observation dates, raw card metrics, unresolved questions and proposed production action. Note when product, offer or destination changes; results for one combination should not silently become the reputation of another.
Re-establish the delivered asset before diagnosing it
Imagine a camping brand sourcing cookware and content from China. A detail image receives few clicks. First confirm which file ran, which product it represented and where its link led. An unclear image, mismatched product and unsuitable destination are different hypotheses. If the report cannot identify the card’s actual asset, record that limitation rather than assigning blame to the photographer.
Then specify the missing buyer information: how the lid locks, how small the kit packs, or what comes in the bundle. This turns a vague request for “better engagement” into a reviewable creative task.
Match the purchase order to the confirmed gap
| Confirmed gap | Work to scope first | Acceptance item |
|---|---|---|
| Results cannot be mapped to an image version | Asset mapping and reporting reconciliation | A traceable card-to-file record |
| Existing material contains the information, but crop, copy or link is wrong | Artwork, copy or configuration revision | Revised asset and its release configuration |
| No usable view or action shows the required product fact | A targeted photography or video pickup | Named new shots, with reusable work retained |
| Only a weak metric is known; cause remains uncertain | A focused modification hypothesis | A subsequent observation plan |
If an approved storage image merely lost its size reference in the crop, try a composition revision. If no existing image explains the locking mechanism, a pickup may be justified. These are production judgments, not automatic conclusions from CTR. To evaluate a change, agree on comparable conditions and a primary outcome; changing the product, offer and entire design at once makes a photography-only explanation difficult to support.
Price the work that changes
A useful quote structure is: agreed asset review and diagnosis + card artwork and adaptation + necessary pickups + checks on affected versions. This is a scoping model, not ONCE pricing or an industry benchmark; media spend sits separately.
Do not copy the full ad spend onto each card and add it again. Without a defensible allocation and outcome definition, avoid manufacturing a card-level acquisition cost. Ask the supplier to identify reusable source files, the exact card revisions, any required set work and the person responsible for matching the release to the approved assets.
For briefs covering both photography and footage, the ONCE production collection provides related planning guidance; the marketing collection covers how those assets are used and assessed.
In a recurring service, the useful asset is the history of a product-information task, its image version, observation conditions and subsequent change. That record can guide the next commission across a brand and its production partners. Any cost saving or performance improvement still needs project-level validation; this article reports neither.
Source and method note
The official sources linked above define the reporting scope. The card-to-production receipt, purchase-order method and numerical example are ONCE’s analysis. No advertising account was accessed or modified, and no delivery test was performed. The original AI cover is a production-themed illustration, not a client shoot.
