One row per campaign per day, for Sponsored Products. The left-hand side is what you spent — impressions, clicks, cost — and the right-hand side is what Amazon attributed back to it.

What makes this report bigger than it looks is that the outcome side is repeated four times, once per attribution window: 1, 7, 14 and 30 days. Each window is repeated again for units and again for same-SKU, which is how a report about campaigns ends up 43 columns wide. They are not alternative measures of the same thing and they do not sum: purchases_30d includes purchases_7d. Picking a window is picking how much credit a click gets for a slow decision.

The row also carries the campaign's configuration on the day — budget, budget type, bidding strategy, and any budget rule that overrode the configured amount. That is what makes the report self-contained: a campaign that flatlines at its campaign_budget_amount is capped, and you can see it without joining anything.

In the Amazon Ads console, this is Measurement and Reporting → Sponsored ads reports, with report category Sponsored Products and report type Campaign. Choose daily time unit, pick your window, and the console emails a link when it is ready.

Through the Amazon Ads API v3, you POST to /reporting/reports with reportTypeId: spCampaigns, adProduct: SPONSORED_PRODUCTS, groupBy: ["campaign"] and timeUnit: DAILY, naming every column you want. Amazon returns a report ID; poll it until the status is COMPLETED and download the gzipped JSON from the URL it hands back. Generation can take as long as three hours, and startDate to endDate may not exceed 31 days.

Two things about v3 that the v2 API did not do. First, you get exactly the columns you asked for — nothing arrives that you did not request, so a missing column is your request, not Amazon's schema. Second, some columns are only available at certain groupBy values: the campaign_* configuration columns in this report exist because it groups by campaign.

Reports are per-profile, so a seller in five marketplaces runs five requests and gets five files, each in its own currency.

Calculating ACOS and ROAS you can defend. ACOS is cost ÷ sales_Nd and ROAS is its inverse. Fix the window across every report you publish — most teams settle on 7 or 14 days — and say which one you used, because the same campaign can show 22% ACOS at 7 days and 17% at 30.

Finding budget-capped campaigns. Compare daily cost against campaign_budget_amount. A campaign that hits its cap before the day ends stops bidding, and the demand it missed appears nowhere in this report — it is invisible loss, and the only signal is the cap itself.

Separating halo from direct sales. The gap between sales_7d and attributed_sales_same_sku_7d is revenue from other products bought after the click. A campaign with poor same-SKU ACOS but strong total ACOS is working; judging it on the advertised SKU alone would have you switch it off.

Auditing a CPC step change. campaign_bidding_strategy and campaign_budget_amount on the row mean a jump in cost_per_click can be checked against a same-day settings change before anyone blames the auction.

Measuring how long your category takes to convert. The ratio of purchases_1d to purchases_30d is a direct read on deliberation time. High-consideration products show a wide spread; impulse buys converge almost immediately, and that tells you which window to standardise on.

Attribution windows nest — never add them. sales_30d contains sales_14d contains sales_7d contains sales_1d. Summing them inflates revenue roughly fourfold, and it is by some distance the most common error made with this report.

Recent days keep changing. Amazon restates conversion data 1, 7 and 28 days after the conversion event, and because conversions are reported against the date of the ad interaction a 30-day window can still be revised nearly two months after the advertising day. Amazon says conversion data is subject to change for up to 60 days back, and recommends re-pulling monthly. Click and impression data moves too, for up to three days, as invalid traffic is filtered out. Any snapshot you take of yesterday will disagree with the same day re-pulled in a fortnight — that is attribution settling, not a bug.

cost and spend are the same number. Both are requested and carried because both are in Amazon's vocabulary, but they are not independent measures and adding them doubles your spend.

Campaign names are mutable and reported as-of-now. A campaign renamed today has its new name on rows from three weeks ago. Join on campaign_id.

Don't average the rate columns. click_through_rate and cost_per_click are computed per row. Averaging them across days or campaigns weights a 12-impression day the same as a 120,000-impression one. Recompute from clicks, impressions and cost.

Sponsored Brands and Sponsored Display are separate reports. This one covers Sponsored Products only. A total ad spend figure needs all three, and their columns do not line up.

KENP is invisible to ACOS. For KDP advertisers, kindle_edition_normalized_pages_royalties_14d is real revenue that no sales_* column includes, so ACOS computed the usual way overstates the true cost of advertising a Kindle Unlimited title.