Search term harvesting: promote the queries that pay, block the ones that proved they will not.

Search term harvesting is the one habit that turns a broad campaign into an exact-match campaign you can actually price — and most sellers either never do it or do it on a gut feeling. Every automatic or broad campaign is a free research tool: shoppers type real queries, Amazon matches them to your product, and the report tells you which of those queries paid. Harvesting is the step where you act on that report. You promote the terms that earned their place, you block the ones that proved they will not, and you leave the rest alone until the data is good enough to judge.

What harvesting actually is

A search term is what the shopper typed. A keyword is what you bid on. The two are not the same, and the gap between them is the whole opportunity. When you bid on "running socks" in broad match, Amazon is free to show your ad for "merino running socks for women", and your report shows the second phrase as a search term with its own clicks, spend and orders.

Harvesting means taking a search term that has proven it converts at a profitable cost, and giving it its own exact-match keyword with its own bid. The discovery campaign keeps finding terms. The exact campaign spends on the ones you already know.

The reason to move a winner out is control. Inside a broad campaign, a winning term shares a bid with every other query that campaign matches, so you cannot price it for what it is worth. In an exact keyword it has a bid of its own. That is the same trade-off described in the guide to broad, phrase and exact match: broad finds, exact controls.

Start from break-even, not from a round number

Most harvesting rules you will find online say something like "promote any term with two or more orders". That is a fine starting filter, but it ignores the only question that matters: does the term make money after ad spend? The answer depends on your margin, and the arithmetic is short.

Take an example product that sells for €25 with 35% margin before advertising. Profit per unit before ads is €8.75. Your break-even ACoS is therefore 35%: spend more than that share of revenue on a term and each sale loses money. If the product converts at 10%, your break-even cost per click is €25 × 35% × 10%, which is €0.875. Those three numbers are the yardstick for every decision below. They are illustrative inputs; swap in your own.

The three buckets for every search term

Open the search term report for the last 30 to 60 days, sort by spend, and put each term into one of three buckets. We covered the negative side in more detail in how negative keywords cut wasted spend; here is the full picture.

1. Harvest. The term has at least two orders and its ACoS sits under your break-even. In the example, a term called "merino running socks" with 58 clicks at €0.80 each spent €46.40 and produced 7 orders, or €175 in revenue. That is a 26.5% ACoS against a 35% break-even and a 12% conversion rate against the 10% assumption. It earned an exact keyword.
2. Negate. The term has spent more than one unit's margin with zero orders. In the example that line is €8.75. A term that has burned €14.20 without a sale has had enough chances.
3. Wait. Everything else. A term with €3.40 of spend and no orders is not a loser, it is a small sample. Negating it is how accounts quietly starve their own discovery.

The line is per product, not universal. A €60 product with a 40% margin has a €24 threshold, and the same €10.60 that justifies a negative above would be a perfectly reasonable cost of finding out. That is why a fixed money rule fails: it treats a cheap product and an expensive one the same.

How to harvest in practice

Start the new exact keyword at roughly what the term cost you in the discovery campaign, in the example €0.80, and keep it under the €0.875 break-even CPC. You are not guessing. You have a real observed cost and a real conversion rate. Then add the same term as a negative exact in the campaign you found it in. If you skip that, your discovery campaign and your exact campaign compete for the same click, and you pay for the privilege of running both.

Where the term came from also matters. A term harvested from an automatic campaign tells you something the keyword list could not: shoppers use that phrase. It can be worth checking the Search Query Performance report to see how many shoppers search it at all before you commit a bid.

Why the weekly rhythm beats the monthly clean-up

Harvesting once a quarter works in the sense that it is better than never. But a converting term found in week one earns for twelve more weeks if you act on it, and a bleeding term found in week one costs you for twelve more weeks if you do not. The report does not get better with age. The work is small, repeatable and decided by arithmetic, which is exactly the kind of work that gets skipped when the account is busy.

Mirox is built for this kind of repeatable decision. It sets bids automatically, records the numbers behind every bid change, and lets you export them, so you can check any decision against your own break-even instead of taking it on trust. The 14-day free trial bids live on up to 5 of your products, on campaigns Mirox builds for them, so your existing campaigns stay untouched. The maths in this post is something you can check yourself either way.

Common mistakes

1. Harvesting on orders alone. One order at a 90% ACoS is not a winner. Check the cost against break-even first.
2. Forgetting the negative. A harvested term left active in the source campaign keeps spending there.
3. Starting the exact bid at the category average. Start from what the term actually cost you.
4. Negating on a few clicks. If the term has not spent a unit of margin, it has not had a fair chance.

The one-line version

Promote search terms that convert under your break-even ACoS, block the ones that have spent a unit of margin for nothing, wait on the rest, and always negate the source after you harvest.

Want to see how Mirox shows its reasoning for each bid? Read how transparency works, then pair this guide with the match types explainer.