Amazon Keyword Harvesting: Turn Search Terms Into Profit

Dana's best keyword was hiding in an auto campaign she hadn't opened in four months.

She sells silicone baking mats. Her auto campaign ran at 34% ACoS, which she treated as the price of "discovery." When she finally exported 60 days of search terms, one line stopped her: "macaron baking mat" had 14 orders at 16% ACoS. Nobody had ever bid on it on purpose. It was buried inside a campaign where Amazon shared one bid between that term and 300 others, most of them losing money.

She moved "macaron baking mat" into an exact match campaign, gave it its own bid, and blocked it from the auto campaign so the two would not compete. Six weeks later the term was producing 3x the orders at 18% ACoS, and her auto campaign's ACoS had dropped to 27% because its best term was no longer propping up the worst.

That is Amazon keyword harvesting. It is the single routine that turns discovery campaigns from a cost center into a pipeline of proven keywords. This guide covers exactly when a search term earns promotion, what bid to start it at, how to negate it at the source, and the mistakes that make harvesting backfire.

What Is Amazon Keyword Harvesting?

Amazon keyword harvesting is the process of finding customer search terms that already convert in your discovery campaigns (auto, broad, and phrase match), moving them into their own exact match campaigns with a dedicated bid, and adding them as negative exact keywords in the original campaign so the two targets don't compete for the same query.

It has three parts, and skipping any of them breaks the system:

1. Find: Pull the search term report and identify terms with enough orders at an acceptable ACoS.
2. Promote: Add each winner as an exact match keyword (or an ASIN as a product target) in a performance campaign.
3. Isolate: Add the same term as a negative exact in the campaign it came from.

The same logic applies to ASINs. When a competitor's product page shows up in your search term report with a string like "b0c7xk2m4q" and it converts, that's a harvestable product target.

Why Harvesting Beats Leaving Winners in Auto

If a search term already converts in an auto campaign, why bother moving it? Four reasons, all of them money.

Bid control. In an auto or broad campaign, one bid covers every query the ad group matches. Your 16% ACoS term and your 60% ACoS term pay the same price. In exact match, the winner gets a bid set for its own conversion rate. You can raise it without dragging 300 other terms up with it.

Placement control. Exact match campaigns let you add top-of-search placement adjustments to a term you know converts. You can't do that surgically in auto. Our placement guide covers the modifiers in detail.

Clean data. Once a term has its own keyword, you get keyword-level reporting for it every day, not just a line in a search term export you pull once a month.

Discovery stays discovery. When winners leave the auto campaign, its budget goes back to finding new terms instead of paying for ones you already know about. This is why Dana's auto ACoS dropped even though she did nothing to it except remove its best performer.

The Keyword Harvesting Funnel

Think of your account as a funnel with three layers, the structure described in our campaign structure framework:

Search terms flow down the funnel. Negatives flow up. Every term you promote to exact becomes a negative exact in the layers above it. That one rule is what keeps your campaigns from bidding against each other, the problem we unpack in the PPC cannibalization guide.

When a Search Term Earns Promotion: The Harvesting Thresholds

The most common harvesting mistake is promoting a term after one lucky order. One order is noise. The second mistake is waiting for 20 orders on a term that has converted cleanly for weeks, leaving money on the table the whole time.

Here are the thresholds we use as a starting point, based on a 30 to 60 day look-back:

The negate threshold depends on your conversion rate. A simple rule: negate after roughly 2x the clicks it normally takes you to get one order. If your ad group converts at 10%, you'd expect an order every 10 clicks, so a term with 20 clicks and zero orders is a strong negate candidate. At a 5% conversion rate, give it 40 clicks. Our negative keywords guide covers the full process for the negate side.

Relevance overrides numbers. If "silicone baking mat for dog food" got one order, it's still not a term you want to scale. If "half sheet baking mat" got one order at a good ACoS and describes your exact product, put it on the watch list and check again in two weeks.

How to Harvest Keywords on Amazon: Step by Step

Step 1: Pull the Search Term Report

In the Amazon Ads console, go to Measurement & Reporting, create a Sponsored Products report, and choose the Search term report type. Pick a 60-day range for low-traffic products and 30 days for high-traffic ones. Our search term report walkthrough covers every column.

Export to a spreadsheet and filter to your discovery campaigns only: auto, broad, and phrase.

Step 2: Sort and Classify Every Term

Add a column called Action. Sort by orders descending, then work through the list and tag each row:

• Promote: Meets the promotion thresholds above.
• Watch: One order, or ACoS slightly above target.
• Negate: Crossed the negate click threshold with no orders, or clearly irrelevant.
• Ignore: Too few clicks to judge.

Most rows will be Ignore. That's normal. A typical account with $3K/month in discovery spend turns up 5 to 15 promotable terms per month.

Step 3: Check Whether the Term Already Exists

Before adding anything, search your exact match campaigns for the term. If it already exists as an exact keyword and still shows up in auto, the fix isn't another keyword. It's a missing negative in the source campaign. Harvesting the same term twice creates duplicates that split data and bid against each other.

Step 4: Set the Starting Bid With Math, Not a Guess

A promoted term deserves a bid based on what it can afford. The formula:

Say Dana's baking mat sells for $32, her target ACoS is 25%, and "macaron baking mat" converted at 12% in the auto campaign.

$32 x 0.25 x 0.12 = $0.96

That's the break-even bid for her target. The auto campaign was paying $0.62 for the term, because that was the shared bid. At $0.96 in exact match, she had room to win more top-of-search impressions without breaking her target.

Step 5: Add the Term to an Exact Match Campaign

Put harvested keywords in a dedicated exact match campaign, grouped by product. Keep ad groups small enough that you can see each keyword's performance at a glance. Some sellers use one keyword per ad group for their top 10 terms, which gives each its own budget visibility.

For ASINs, create a product targeting campaign and add the converting ASIN as an individual product target. The ASIN targeting guide explains how to structure these.

Step 6: Negate at the Source

Now add the exact term as a negative exact in every discovery campaign where it appeared. Use negative exact, not negative phrase. Negative phrase on "macaron baking mat" would also block "large macaron baking mat" and "macaron baking mat set," longer terms that the auto campaign should still be discovering.

For harvested ASINs, add them as negative product targets in the auto campaign.

Skip this step and the auto campaign keeps entering the same auctions as your exact campaign. Amazon usually prefers the exact match when keywords match identically, but auto targeting is fuzzier and the two can still collide. You pay twice for the discovery and get messy data on both sides.

Step 7: Review Promoted Terms After 14 Days

Harvested terms don't always keep performing. Sometimes a term converted in auto because the auto campaign served it mostly on product pages at low CPCs, and at top-of-search it behaves differently. After two weeks, check each promoted keyword:

• ACoS at or under target: Raise the bid 10-15% and test for more volume.
• ACoS 1-1.5x target: Lower the bid toward the formula number.
• ACoS far above target with enough clicks: Lower the bid hard or pause, and remove the negative from the source campaign so discovery can run again.

A Second Example: Harvesting Competitor ASINs

Tomás sells a no-pull dog harness. His auto campaign's "complements" and "substitutes" targeting was quietly placing his ad on competitor product pages. When he sorted his search term report for rows starting with "b0," he found three competitor ASINs that together accounted for 22 orders at 19% ACoS. The competitors had fewer reviews and a higher price than his harness.

He created a product targeting campaign with those three ASINs, bid 25% above what auto had been paying, and negated them in the auto campaign. Within a month those three targets were his most profitable ad spend in the account. He then ran the same process every two weeks and found five more over the following quarter.

ASIN harvesting is overlooked because the search term report shows ASINs as unreadable codes. Paste each one into an Amazon search to see the product. It takes two minutes and often uncovers your best conquest targets.

How Often Should You Harvest Keywords?

Cadence depends on how much data your discovery campaigns produce:

Seasonality matters too. Harvest right after Prime Day and in early October before Q4 bids climb, so your best terms already have their own campaigns when traffic peaks. Our Q4 PPC strategy builds harvesting into the preparation timeline.

Seven Keyword Harvesting Mistakes That Cost Money

1. Promoting on one order. One conversion at 8 clicks might be a 12% converter or a 3% converter that got lucky. Wait for the second order.
2. Forgetting the negative. The most common cause of rising CPCs after harvesting. Your campaigns end up competing for the same query.
3. Using negative phrase instead of negative exact. It blocks the long-tail variations your discovery campaign should keep finding.
4. Pausing the auto campaign once it's "done." Search behavior changes constantly. Discovery campaigns should always run, just at lower bids.
5. Harvesting branded terms into generic campaigns. Your own brand terms belong in a separate brand defense campaign with their own ACoS target.
6. Copying the auto bid. The auto bid was a compromise across hundreds of terms. The harvested term deserves a bid set by its own economics.
7. Harvesting only keywords. Converting ASINs are often cheaper and more profitable than the keywords sitting next to them in the report.

Manual Harvesting vs Automation

Manual harvesting works. It's also the task most sellers drop first when they get busy. It means pulling reports, filtering hundreds of rows, making judgment calls, then keeping keyword additions and negatives in sync across several campaigns. Do it in bulk and you'll want bulk operations. Skip it for two months and your auto campaigns go back to paying a shared bid for your best terms.

Here's the time math for a seller with 20 ASINs: roughly 2 to 3 hours per harvesting session, every two weeks. That's 50 to 75 hours a year on one routine, and it's the routine that decides whether new keywords ever reach their full potential.

Daniks.AI runs this loop every day. It promotes converting search terms and ASINs into exact match and product targeting, sets their bids from your ACoS target, adds the matching negatives in the source campaigns, and blocks terms that burn clicks without orders. You set the ACoS target. The harvesting happens whether or not you have time that week. You can start a 14-day free trial and watch it run on your own account.

Frequently Asked Questions

What Is Keyword Harvesting in Amazon PPC?

It's the routine of moving converting search terms from auto, broad, and phrase campaigns into exact match campaigns with their own bids, then adding them as negative exact keywords in the source campaign. The goal is bid control over proven winners and cleaner discovery.

How Many Orders Does a Search Term Need Before I Harvest It?

Two orders at or below your target ACoS within 30 to 60 days is a reasonable minimum. For expensive, high-click terms, wait for three. One order is not enough to judge conversion rate.

Should I Use Negative Exact or Negative Phrase After Harvesting?

Negative exact. It blocks only the exact term you promoted, so the discovery campaign can keep finding longer variations. Use negative phrase only for terms that are irrelevant in every form.

Should I Pause My Auto Campaign After Harvesting?

No. Lower its bids if its ACoS is high, but keep it running. Auto campaigns keep surfacing new terms and ASINs as shopper behavior and competitors change.

What Bid Should a Harvested Keyword Start At?

Start from the formula Max CPC = Average Order Value x Target ACoS x Conversion Rate. In practice, begin at the higher of the source CPC plus 10-20% or about 80% of that formula number, then adjust after two weeks of data.

The Bottom Line

Your auto and broad campaigns are already finding your next best keywords. Without harvesting, those winners stay trapped behind a shared bid, competing with hundreds of terms that don't deserve their budget.

Pull the search term report. Promote terms with two or more orders at target ACoS. Set bids from the formula. Negate exact at the source. Review after 14 days. Repeat on a schedule that matches your spend, or let an autopilot do it every day.

For the ACoS side of the math, The FBA Girl's ACoS guide is a useful companion read, and Amazon's own Sponsored Products overview covers the match type mechanics.