One row per automatic target per day, for Sponsored Products. This is the auto-targeting half of Amazon's spTargeting report: the same request that produces the keyword report, filtered to keywordType values TARGETING_EXPRESSION and TARGETING_EXPRESSION_PREDEFINED instead of BROAD, PHRASE and EXACT. Across 16 production files and 259,154 rows, (keyword_id, date) was unique on every one, so a target-day is genuinely the grain.

"Automatic" here names the filter, not the author of every row. TARGETING_EXPRESSION_PREDEFINED is the four groups Amazon creates for every auto ad group without being asked. TARGETING_EXPRESSION — 74,797 of the 259,154 measured rows — is the category and product targeting expressions, which Amazon documents as clauses an advertiser adds to a manual campaign. Both arrive in this file because both sit in Amazon's automatic keywordType vocabulary, so read keyword_type before assuming who chose a target.

The columns fall into four groups. Where the target lives — campaign_, ad_group_, portfolio_id, with the campaign's budget and its currency on every row. The target itself — keyword_id, keyword, targeting, keyword_type, match_type and keyword_bid. Yes, the columns are called keyword_ on a report about targets that are not keywords; keyword_id is the target's id, and keyword_type tells you which kind of automatic target the row is. Delivery — impressions, clicks, click_through_rate, top_of_search_impression_share, cost and cost_per_click. Outcomes, repeated at 1, 7, 14 and 30 days for purchases, units and sales, then again for the same-SKU subset of each — and, at 7 days only, the other-SKU remainder. acos_clicks_ and roas_clicks_* at 7 and 14 days are Amazon's own ratios of the money columns already in the row.

The one column pair worth knowing before you open the file: keyword and targeting are the same string on 96.4% of rows and different on the rest. Where they differ the row is a category target, keyword holds the category id (category="2892944031") and targeting holds its name (category="Intimo da uomo"). That is the opposite of the manual keyword report, where the two are equal on every row.

In the Amazon Ads console, this is Measurement & Reporting → Sponsored ads reports → Create report, ad product Sponsored Products, report type Targeting, daily time unit — the console offers the targeting report at two time units, summary or daily.

Through the Amazon Ads API v3, you POST to /reporting/reports with reportTypeId: spTargeting, adProduct: SPONSORED_PRODUCTS, groupBy: ["targeting"], timeUnit: DAILY and format: GZIP_JSON, listing the 52 columns above, and a keywordType filter of ["TARGETING_EXPRESSION", "TARGETING_EXPRESSION_PREDEFINED"] — which is Amazon's own sample request for this report. Every v3 report request uses that one endpoint; it returns a reportId with status PENDING, and you poll GET /reporting/reports/{reportId} until the status is COMPLETED and a url appears, then download the gzipped JSON array from it. Generation can take as long as three hours.

The filter is the report. The keyword report is this exact request with keywordType set to BROAD, PHRASE and EXACT; leave the filter off and you get both halves in one file with the same 52 columns and nothing but the values in keyword_type to tell them apart. Keep the two pulls separate and keep them separately named, because the files behave differently where it matters — see keyword versus targeting above.

Two properties of v3 to rely on. You get exactly the columns you asked for: across 16 files all 52 arrived every time, in one identical set, so a missing column is your request, not Amazon's. And the order of keys in the JSON is whatever Amazon felt like that day — it differed in all 16 files — so read by name, never by position. Reports are per-profile: a seller in five marketplaces runs five requests and gets five files, each in its own currency.

Deciding which automatic targets to keep bidding on. Group by keyword_type and targeting, sum cost and sales_7d, and compute ACOS yourself. The predefined groups behave very differently for most catalogues, and this is the only report that separates them by day. Amazon creates four of these groups for every auto ad group — close match, loose match, substitutes and complements — and lets you bid on each separately, which is what makes splitting them worth the trouble.

Setting per-target bids. keyword_bid is the bid you set and cost_per_click is what the auction actually charged. A target whose CPC sits well below its bid has headroom you are not using; one whose CPC equals its bid is being capped by you, not by competition.

Working with category targets. TARGETING_EXPRESSION rows include category targets. Join on keyword when you need the category id — it is stable and language-independent — and display targeting, which carries the name in the marketplace's own language. Never join the two files of different marketplaces on targeting.

Buying top of search deliberately. top_of_search_impression_share sits on the same row as cost and purchases_7d, so you can see whether the targets winning the top slot are the ones that convert, or just the ones that cost most.

Separating halo from direct sales. sales_other_sku_7d and units_sold_other_sku_7d are the products people bought other than the one advertised. An auto target with weak same-SKU ACOS and strong total ACOS is introducing shoppers to the range; judged on attributed_sales_same_sku_7d alone it looks like waste.

Reconciling against the campaign report. Sum cost by campaign_id and date, and the auto-target rows plus the keyword report's manual rows should meet the campaign report's total. When they do not, the usual reason is a missing filter value or a marketplace left out of one pull.

Attribution windows nest — never add them. purchases_30d contains purchases_14d contains purchases_7d contains purchases_1d, and the same holds for units and sales. Summing the four inflates the number roughly fourfold.

keyword is not targeting. They differ on 3.6% of rows — every category target — with the id on one side and the name on the other. A pipeline that assumes they are equal because they were equal on the keyword report will silently mismatch category targets.

The ACOS and ROAS columns are mostly null, and that is correct. acos_clicks_7d is null on 98.5% of rows (no attributed sales) and roas_clicks_7d on 91.5% (no clicks). They are present-and-null, not missing, and a null is not a zero: a target with spend and no sales has infinite ACOS, not 0%. Recompute both from cost and sales_7d over any aggregate.

ACOS is a percentage; ROAS is a multiple. In the measured files ACOS ran as high as 364.9 and ROAS as high as 15,463. click_through_rate and top_of_search_impression_share are percentages bounded at 100. None of the four is a 0–1 fraction.

Other-SKU exists only at 7 days. There is no sales_other_sku_14d or _30d, so a halo analysis has to be a 7-day analysis or has to approximate the remainder as sales_Nd minus attributed_sales_same_sku_Nd. Amazon defines all three columns independently — total, same SKU as advertised, different SKU — and asserts no identity between them, so the subtraction is a reasonable reading of the definitions and not a measured fact. Check it on your own file before you publish a number that depends on it.

keyword_type and match_type are the same column twice. Equal on every one of 259,154 rows, with the automatic vocabulary — never BROAD, PHRASE or EXACT. If you see those values you are holding the keyword report.

The predefined groups' literal names are not documented. Amazon publishes the four auto targeting groups and the API names for their expression types — CLOSE_MATCH, LOOSE_MATCH, SUBSTITUTES, COMPLEMENTS — but says nothing about what string the report's targeting column carries for each. The example rows on this page use the console's wording as an illustration; only the category form, category="…", comes from a measured file. Read the distinct values off your own first pull rather than hard-coding any of them.

Budget columns repeat per target. campaign_budget_amount is the campaign's budget, copied onto every target row in that campaign. Summing it across an ad group multiplies the budget by the number of targets.

There is no spend column. cost is the only spend figure, which is a difference from the campaign report where both names are carried and equal.

Recent days keep changing. Impressions and clicks land within about 12 hours but can still move for three days while Amazon filters invalid traffic; conversions land within 24 hours and are restated 1, 7 and 28 days after the conversion event. Because a click today can attribute a purchase up to 30 days out, the last month restates on every re-pull.