Amazon Marketing Cloud has 7 tables and 761 columns. Most sellers never touch it
By Anthony Nguyen 🟢 · June 8, 2026 · Curated by George's Blog
Amazon Marketing Cloud has 7 tables and 761 columns. Most sellers never touch it. Most agencies query maybe 5% of it. That's a problem.
The tables that matter:
- amazon_attributed_events_by_traffic_time, your bread and butter for ad performance
- amazon_attributed_events_by_conversion_time, same data, different time alignment
- sponsored_ads_traffic, impression and click-level data
- dsp_impressions, DSP campaign delivery
Each table has _for_audiences variants that return user_id instead of aggregate metrics. That's how you build custom audiences from actual purchase behavior, not just targeting signals.
Here's what trips people up: ad_product_type now includes sponsored_television alongside the classic sponsored_products, sponsored_brands, and sponsored_display. If your queries filter on ad type and you wrote them before STV launched, you're missing data. No error. No warning. Just incomplete results.
Metrics in AMC are binary, 0 or 1 per record. You SUM() them to get totals. There's no pre-aggregated "total clicks" column sitting there waiting for you.
761 columns sounds overwhelming. In practice, 30-40 columns drive 90% of useful analysis. The trick is knowing which 30. Most people never figure that out.
AdsCrafted runs these audits in one command, P&L, reimbursements, catalog health, AMC queries, all conversational → adscrafted.com