Every Amazon account generates a search term report. Very few accounts do anything with it on a schedule. Harvesting is the schedule: a weekly pass that moves converting terms into exact-match campaigns and stops the account from buying the losers again.
Done every week, it compounds. Research spend gets cheaper. Profit campaigns get bigger. The search term report gets shorter. Done once a quarter, it is a cleanup that never catches up.
What harvesting actually decides
Each search term with meaningful spend gets one of four outcomes:
- Harvest. It converted at or below target ACoS with enough orders to trust. Add it as exact match to the profit campaign for that product, and negate it as exact in the campaign that discovered it.
- Watch. It converted, but the sample is small or the ACoS is above target. Leave it, note it, and decide next week with more data.
- Negate. It spent without converting past a threshold you set from the product's price and conversion rate. Add it as a negative at the right level.
- Ignore. Spend too small to matter yet. Do nothing. Not every term needs a decision.
The value is in doing this consistently, not in the sophistication of the rules.
Thresholds that make sense
Fixed thresholds like "negate after 10 clicks" ignore the product. A term on a $12 product with a 15% conversion rate should convert within a handful of clicks. The same term on a $180 product with a 4% conversion rate might reasonably need 25 or more clicks before you judge it.
We set the negation threshold per product from expected clicks per order, then apply a multiplier. If a product typically needs 12 clicks per order and a term has taken 30 without one, that term is not going to work. Below that line, it is still an open question.
Harvest thresholds are the mirror image. Two orders on a term is a signal. It is not proof. We harvest to exact match with a conservative bid and let the profit campaign confirm it over the next few weeks.
Where to negate
Negatives go at the level that stops the waste without blocking useful traffic.
| Situation | Negative level | Match type |
|---|---|---|
| Term is irrelevant to the product entirely | Campaign | Phrase |
| Term is relevant but converts poorly for this ASIN | Ad group | Exact |
| Term was harvested to exact match elsewhere | Source campaign | Exact |
| Competitor brand names you do not want to conquest | Campaign or portfolio | Phrase |
The most common mistake is negating with phrase match too aggressively. A phrase negative for "cheap" blocks "cheap" and every query containing it, including some that convert. Use exact negatives unless you are certain the whole family of queries is worthless.
The weekly workflow
This is what a harvesting session looks like on a managed account:
- Pull the search term report for the last 30 days with a 7-day comparison. Longer windows smooth out noise; shorter ones catch changes.
- Filter to terms with spend above the product's negation threshold or with at least one order.
- Sort by spend descending. Work the top of the list first, because that is where the money is.
- Apply the four outcomes. Log every harvested term in a keyword tracker with its source campaign and date.
- Review last week's harvested terms in their new profit campaigns. Adjust bids that overshot.
The whole pass takes less time than most sellers expect, because after a few weeks the list of undecided terms is short.
Harvesting across ad types
Sponsored Products search terms are the main source, but not the only one. Sponsored Brands search term reports surface category and comparison queries that often belong in Sponsored Products exact campaigns. Auto campaign "close match" and "substitutes" targeting groups behave differently and should be harvested separately, because substitutes reveal competitor ASINs worth targeting directly with product targeting.
What breaks the loop
Three things stop harvesting from compounding:
- No structure. If there is no exact-match profit campaign to harvest into, terms get added to the same mixed campaign and nothing changes. Read campaign structure first.
- Skipping the negation step. Harvesting a term without negating it in the source campaign means the auto campaign keeps buying it, now at a worse price.
- Judging by ACoS alone. A term with a high ACoS on a launch product might be building rank. Harvest decisions need the campaign role, not just the number.
Harvesting is not glamorous. It is the mechanism by which an advertising account learns. Accounts that run it every week get cheaper to operate every month.
Want to know what your search term report is hiding?
Get Your Free Amazon Audit