The OpenAI Ads Insights report can now break performance down by product, with ten new fields describing the item behind each row — feed, item id, title, brand, price, availability, description, image URL, target URL and seller.

OpenAI Ads exposes a product breakdown on its Insights API: instead of one row per entity per day, the report returns one row per product item per entity per day, with the feed metadata attached. Improvado now ingests it, so shopping performance can be read per item rather than only per campaign or ad.

The new fields

• Product Feed ID — the product feed the item belongs to.
• Product Item ID — the item's identifier inside that feed.
• Product Title, Product Brand, Product Seller Name — the item as it appears in the feed.
• Product Price — amount and currency in a single string, for example 119.99 EUR. Items priced in different currencies can appear side by side in one account, so parse the currency rather than assuming the account's.
• Product Availability — for example in_stock.
• Product Description, Product Image URL, Product Target URL — the feed's copy, image and landing page. Image URL is empty for items that carry no image in the feed.

The breakdown works at every aggregation level the Insights report offers — Ad Account, Campaign, Ad Group and Ad — so the product rows can be rolled up against whichever level you already report on.

Three rules the API enforces

• One breakdown at a time. Product cannot be combined with Country or Device Type in the same extraction. Selecting fields from two breakdowns is refused when you save the extraction, rather than failing the run later.
• Conversions are not available per product. OpenAI Ads does not report conversions at the product level. Keep Conversions in a separate, unsegmented extraction.
• Product Feed ID and Product Item ID are required together. The API refuses a product request that carries neither, and the feed id alone repeats across every item of the feed — so rows would not be distinguishable from one another. Selecting any product field therefore requires both.

What to expect in the data

Metrics on these rows are per product: impressions, clicks and spend are the item's own, not the campaign's repeated across its items. Row counts grow with the size of the feed, so start with a short date range and the entities you actually report on before widening.

Accounts without a product feed return no product rows. This is expected, not an error — the breakdown only applies to shopping delivery.