Match-Type Harvesting with Claude Code for Amazon Sellers

What they did

A Seller Labs video describes a solo seller who had paid an agency $37,000/yr for 3 years to run 8 broad campaigns, fired them, and rebuilt his structure with Claude Code using match-type harvesting: broad → phrase → exact single-keyword ad groups, with negatives added back on the broad. The video gives the full Claude Code prompt and explains how to apply the output by hand or via a Bulk Operations upload.

How it went

Per Seller Labs, three months later the seller had 40 campaigns and his ROAS had doubled. The seller isn't named and no account data is shown.

We only report what the source says. Results are self-reported unless noted.

Worth knowing

Seller Labs sells an Amazon MCP server, and the video is narrated by an AI avatar. The prompt is designed so no promotion or negative goes live without approval.

Copy-paste prompt

Build me a match type harvester for my Amazon ad account. Ask me before any code: my target ACoS; my minimum spend floor (below this, ignore noise); my minimum conversion floor (below this, a term doesn't qualify for promotion); my conversion attribution window (7, 14, or 30 days); and the percentage above current CPC I want to bid when promoting a term to a tighter match type. Don't assume defaults.

Ask me to paste my Sponsored Products search term report CSV (Campaign Manager reports tab, last 30 days) or confirm I have live Amazon Ads data wired through an MCP. If I paste the CSV, treat the first row as headers and confirm which columns you're mapping to spend, clicks, orders inside the attribution window, ACoS, campaign, ad group, customer search term, match type of the targeting keyword, CPC, and advertised ASIN.

Classify every customer search term. A term qualifies for promotion to phrase match if its current targeting keyword is broad match, has spend above my floor, has orders above my floor inside my attribution window, and has ACoS at or below my target. A term qualifies for promotion to exact match if its current targeting keyword is phrase match and meets the same volume and ACoS criteria. Anything draining spend with zero orders gets flagged for negative keyword addition to the original broad keyword.

Output one decision table, one row per customer search term. Columns: current match type, recommended match type, advertised ASIN, spend, clicks, orders, ACoS, current CPC, recommended starting bid (current CPC plus my premium), recommended negative keyword adds for the original broad keyword, one-line reason. Then bundle promotions by advertised ASIN: for each ASIN with three or more promotions, output the suggested ad group structure (one exact match ad group per SKAG, one phrase match ad group containing the phrase-tier terms, plus the negative keyword list for the original broad keyword). If I ask for it, also produce the same structure formatted for the Amazon Sponsored Products bulk operations template (entity type, operation = create, campaign name, ad group name, match type, keyword text, bid) so I can upload it directly. No auto action. Every promotion and every negative keyword add waits for my approval.

This prompt is quoted or closely adapted from the original source. Replace anything in [brackets].