How to Run a Weekly Amazon Search Term Harvest That Actually Lowers ACOS
Most Amazon accounts we take over have the same problem: a Sponsored Products auto campaign that has been running untouched for eight months, quietly spending 30% of the ad budget on search terms nobody has ever read. The account owner knows the harvest is supposed to happen. It just never gets on the calendar. Here is the exact weekly routine we run on client accounts, start to finish, in about 40 minutes. Pull the right report, not the convenient one Use the Search Term Report from Campaign Manager, not the Targeting Report. The Targeting Report tells you how your keywords performed. The Search Term Report tells you what shoppers actually typed, which is the only place discovery lives. Pull a rolling 30-day window, not 7 days. Seven days gives you almost no statistical weight at typical mid-size volume, and you will end up negating terms that were simply unlucky. Export it, drop it in a sheet, and add one column: conversion rate, calculated as orders divided by clicks. That single column does most of the work in the next three steps. Promote winners into exact-match, one term at a time Filter for search terms with at least 2 orders and an ACOS at or below your target. Those are your graduates. Move each into a dedicated exact-match campaign or ad group so you control the bid directly instead of letting the auto campaign guess. Set the starting bid at roughly the term's current cost-per-click, not the suggested bid Amazon shows you. Amazon's suggestion is built off marketplace-wide competition; your own historical CPC on that exact term is the better signal, because you already know it converts at that price. Then negate the harvested term inside the source campaign as a negative exact. If you skip this, your auto campaign and your new exact campaign bid against each other and you pay a premium for the privilege. Kill the bleeders with a click threshold, not a hunch The rule we use: any search term with 12 or more clicks and zero orders in the 30-day window becomes a negative exact. Twelve is not magic, but it should scale to your product. If your typical listing converts at 10%, twelve clicks with no order is already a meaningful signal. If you convert at 3%, raise the threshold to 30 or you will negate terms that would have converted eventually. Be careful with terms that show clicks and no orders but high impressions with a very low click-through rate. Those are usually relevance problems in your title or main image, not ad problems. Negating them hides the symptom; fixing the listing solves it. Read the near-misses for listing and catalog signal The most valuable rows in the report are often the ones with clicks, no orders, and a term that clearly describes your product. That combination almost always means the customer arrived, looked, and found a mismatch. Common culprits: a size or count the listing does not carry, a use case your bullets never mention, a price gap versus the competitor set on that query. We log these in a running sheet each week. After a month, patterns show up that justify a real change: a new variation, a bullet rewrite, a bundle, sometimes a whole new ASIN. Search term data is the cheapest customer research on the internet and most sellers throw it away. Make it repeatable or it will not happen Pick one day. Same day every week. Ours is Tuesday, after the prior week's data has settled. Build a template sheet with the filters pre-set so the only manual step is pasting the export. Keep a change log with the date, the term, and the action taken, because in six weeks you will want to know whether a bid change caused a swing, and Campaign Manager history will not tell you cleanly. Budget 40 minutes. If it takes two hours, your campaign structure is too fragmented and that is the real problem to fix first. Why this matters Search term harvesting is not a growth hack, it is hygiene, and that is exactly why it compounds. Every negated bleeder permanently redirects spend toward terms that convert. Every promoted winner moves a proven query into a campaign where you control the bid and the budget. Accounts that run this weekly typically see ACOS drift down 15 to 25% over a quarter with no increase in effort and no new creative. The sellers who fall behind on Amazon rarely lose on strategy. They lose because the routine work never got scheduled. Put it on the calendar this week.