Home / Blog / Amazon PPC
Keyword Harvesting in Amazon PPC: A Step-by-Step Guide
Keyword harvesting is the routine of pulling search terms that already convert out of your auto, broad and phrase campaigns, and adding them as exact match keywords in a manual campaign. You then add them as negatives in the campaign they came from, so each search is bid on in one place. Done every week or two, it moves spend toward proven terms and away from waste.
Key takeaways
- Harvest from the search term report, not from the keyword list. The search term is what the shopper actually typed.
- Promote terms with orders at or below your target ACoS. Negate terms with many clicks and no sales.
- Always negate a harvested term in its source campaign, or your campaigns compete with each other.
- Harvest ASINs too. Converting product pages belong in a product targeting campaign.
- Set your targets from break-even ACoS first, so "good" and "bad" mean something.
What keyword harvesting is (and why it works)
Amazon shows your ad when a shopper's search matches one of your targets. In an auto campaign, Amazon picks the matches. In broad and phrase campaigns, one keyword can match hundreds of different searches. You pay for every click, whether the search was a perfect fit or a poor one.
The search term report lists the exact words shoppers typed before they clicked your ad. Some of those terms convert well. Some burn money. Harvesting is how you act on that:
- Promote the winners into exact match, where you can set a bid for that one search.
- Negate the winners in the source campaign so they don't get bid on twice.
- Negate the losers everywhere they waste spend.
Exact match gives you the most control. You can raise the bid on a term that converts at 15% and lower it on one that converts at 5%. In a broad campaign, both terms share one bid. Amazon's own guidance notes that exact match tends to have the highest conversion rates of the three match types.
Before you start: set your target ACoS
You can't call a term a winner until you know what you can afford to pay. ACoS (advertising cost of sale) is ad spend divided by ad sales. Your break-even ACoS is your profit margin before ad spend. Above it, every ad sale loses money.
Worked example. Say your product sells for $29.99. It is large standard, 12 to 16 oz, so the 2026 FBA (Fulfillment by Amazon) fee in the $10 to $50 price band is $4.60. Add the 3.5% fuel and logistics surcharge (about $0.16) for $4.76.
| Line | Amount |
|---|---|
| Sale price | $29.99 |
| Referral fee (15%) | −$4.50 |
| FBA fee + 3.5% surcharge | −$4.76 |
| Landed product cost | −$6.00 |
| Profit before ads | $14.73 |
| Break-even ACoS | 14.73 ÷ 29.99 = 49.1% |
Note that holiday peak fulfillment fees apply from 15 October 2026 to 14 January 2027, so your margin, and your break-even ACoS, is a little lower during that window.
Most sellers set a target ACoS below break-even so ads earn a profit. In this example, a 30% target leaves about $5.73 profit per ad sale. You can run your own numbers in the Break-even ACoS & Max Bid Calculator. For more on choosing the target, read break-even ACoS explained.
Where to find the data
You harvest from the Sponsored Products search term report. There are two ways to get it:
- Download it: In the Amazon Ads console, go to Measurement & Reporting → Sponsored ads reports → Create report. Choose Sponsored Products, report type Search term, and a date range. Download it as a spreadsheet.
- View it in Campaign Manager: Open a campaign, open an ad group, and click the Search terms tab. This is handy for quick checks on one ad group.
Two things to know about the data:
- Attribution lag. For sellers, Sponsored Products credits a sale to an ad if it happens within 7 days of the click. Orders from the last few days are still coming in, so don't judge yesterday's terms yet. Many sellers leave out the most recent 2 to 3 days.
- Limited history. The console only keeps roughly the last two months of search term data. Download and save it regularly if you want a longer record.
Our full walkthrough of the Amazon search term report covers every column.
Step-by-step: how to harvest keywords
Step 1: Pull 30 to 60 days of search terms
Use enough data to see a pattern. A term with 3 clicks tells you almost nothing. For low-traffic products, use 60 days. For busy ones, 14 to 30 days is often enough.
Step 2: Filter to terms with enough clicks
Sort by clicks and focus on terms with real volume. As a rough guide, a term needs enough clicks that you'd expect at least one or two sales at your normal conversion rate. If your CVR (conversion rate: orders ÷ clicks) is 10%, 10 clicks should produce about 1 order.
Step 3: Sort each term into a bucket
Use simple rules. These are common starting points, not laws. Adjust them for your price and margin.
| Bucket | Typical rule | Action |
|---|---|---|
| Winner | 2+ orders, ACoS at or below target | Add as exact in manual campaign, negative exact in source |
| Promising | 1 order, ACoS below target, few clicks | Leave running, check again next cycle |
| Too expensive | Orders, but ACoS well above break-even | Lower bid or move to exact with a lower bid |
| Waste | 10 to 15+ clicks, 0 orders | Add as negative exact (or phrase) |
| Irrelevant | Wrong product, any clicks | Negate right away |
Here is what that looks like on a sample report, using the 30% target from earlier:
| Search term | Clicks | Spend | Orders | Sales | ACoS | Bucket |
|---|---|---|---|---|---|---|
| bamboo sheets queen | 48 | $38.40 | 6 | $179.94 | 21.3% | Winner |
| cooling sheets queen | 31 | $27.90 | 3 | $89.97 | 31.0% | Winner (borderline) |
| silk pillowcase | 22 | $17.60 | 0 | $0 | – | Waste |
| bamboo sheets king | 9 | $7.20 | 1 | $29.99 | 24.0% | Promising |
| b0xxxxxxxx (an ASIN) | 14 | $9.80 | 2 | $59.98 | 16.3% | Winner (ASIN) |
The borderline term is a judgment call. At 31%, it is slightly above target but well below break-even. Most sellers would still harvest it and set a slightly lower exact bid.
You can speed this step up with the free Search Term Report Analyzer, which buckets your terms for you.
Step 4: Add winners as exact match keywords
Add each winner to a manual Sponsored Products campaign with exact match targeting. Many sellers keep a dedicated "exact" campaign per product so the budget for proven terms isn't shared with testing.
Choosing the starting bid. A simple formula ties bid to profit:
Max CPC = Price × CVR × Target ACoS
For "bamboo sheets queen": 6 orders from 48 clicks is a 12.5% CVR. So $29.99 × 0.125 × 0.30 = $1.12. The term actually paid $0.80 per click in the auto campaign ($38.40 ÷ 48). Starting at $0.90 to $1.00 gives it room to win more impressions while staying well inside your target.
Step 5: Negate the winner in the source campaign
Go back to the campaign where the term was found. Add it as a negative exact keyword. This stops the auto or broad campaign from bidding on the same search.
Skipping this step is the most common harvesting mistake. Without it, two of your campaigns enter the same auction. Amazon only shows one of your ads, but you no longer control which bid wins, and your reports get muddy.
Step 6: Negate the waste
Add irrelevant and non-converting terms as negatives. Use negative exact for a specific search. Use negative phrase when a word is always wrong for you (for example, "silk" if you sell cotton). Many sellers wait for 10 to 15 clicks with no sale (more if your conversion rate is low), so don't cut a term after 4 clicks. Our guide to negative keywords in Amazon PPC goes deeper.
Step 7: Harvest ASINs into product targeting
In auto campaigns, search terms that look like "b0…" are ASINs (Amazon Standard Identification Numbers). They mean your ad ran on that product's detail page and got a click. Converting ASINs go into a manual product targeting campaign, targeted individually. Then add them as negative product targets in the auto campaign.
A simple campaign layout for harvesting
You don't need a complex structure. This three-layer setup is widely used:
| Campaign | Role | Gets negatives for |
|---|---|---|
| Auto | Discovery: finds new searches and ASINs | Harvested exact terms and ASINs |
| Broad / phrase | Discovery on known themes | Harvested exact terms |
| Exact (manual) | Profit: proven terms at controlled bids | Nothing from the others |
| Product targeting (manual) | Proven competitor and complement ASINs | – |
Terms flow down from discovery into the exact and product targeting campaigns. Negatives flow up to stop overlap.
Common harvesting mistakes
- Judging too early. Terms with 3 to 5 clicks swing wildly. Wait for real volume.
- Not negating in the source. This creates self-competition, as covered above.
- Using last week's data before attribution settles. The newest days under-report orders.
- Negating with broad-ish negatives by accident. A negative phrase on "sheets" would block every search with "sheets" in it, including your winners.
- Harvesting on ACoS alone. A term with 30 orders at 35% ACoS may still be worth more than a term with 1 order at 10%. Volume matters.
- Never revisiting exact bids. Once a term moves to exact, it needs its own bid reviews. Harvesting is not the end of the job.
- Ignoring TACoS. TACoS (total advertising cost of sale) is ad spend divided by total sales, ads plus organic. Harvesting should help ad sales grow while TACoS stays flat or falls.
How often to harvest
A good default is weekly for high-traffic products and every two to four weeks for low-traffic ones. The goal is enough new clicks per cycle to make real decisions. Keep a simple log of what you promoted and negated each time, so you can check later whether the change helped.
Bottom line
Keyword harvesting turns your discovery campaigns into a pipeline: auto and broad find the searches, exact match pays the right price for the ones that work, and negatives stop the waste. Set your target ACoS from real margins, use clear rules for promote and negate, and always negate a harvested term in its source campaign. Repeat the cycle every week or two and your spend steadily shifts toward terms that make money.
Frequently asked questions
How often should I harvest keywords on Amazon?
Most sellers harvest every one to two weeks. Weekly works for products with plenty of clicks. Smaller products need two to four weeks of data before a search term has enough clicks to judge.
Should I negate a harvested keyword in the auto campaign?
Yes, in most cases. Add it as a negative exact keyword in the campaign it came from. Otherwise the auto or broad campaign keeps bidding on the same search, and you compete with yourself and lose control of the bid.
How many orders does a search term need before I harvest it?
A common rule of thumb is two or more orders at or below your target ACoS. One order can be luck. For high-priced products with few sales, some sellers promote a term after a single order if the click count is low and the ACoS is well under target.
Can I harvest ASINs as well as keywords?
Yes. Auto and some broad campaigns can show ASINs in the search term column when your ad ran on a product page and got a sale. Move converting ASINs into a manual product targeting campaign so you can bid on them directly.
What bid should I use for a newly harvested exact keyword?
Start near the cost per click (CPC) the term already paid in the source campaign, or a little above it. Then adjust after one to two weeks using your target ACoS and the term's real conversion rate.
Amazon fees, policies and features change often. Figures here were checked in October 2026; confirm current rates in Seller Central before making pricing or budget decisions. Sotrics is not affiliated with Amazon.