Back to Blog
GuidesNovember 7, 2026· 13 min read

KDP Ads Keyword Harvesting: How to Turn Auto Campaign Search Terms into Manual Keywords

I

InteliAds Team

KDP Ads

KDP Ads Keyword Harvesting: How to Turn Auto Campaign Search Terms into Manual Keywords

You launched an auto campaign, a few weeks later the report lists dozens of customer search terms, and you have no idea which ones deserve a manual campaign. Some terms have a click or two, a few have orders, and many look like noise. KDP ads keyword harvesting is the routine of reading those auto campaign search terms, keeping the ones with real evidence, and moving them into manual Sponsored Products campaigns where you control bids and match types. Done carelessly, it just copies wasted spend into a new campaign.

Short answer: Harvest search terms from your auto campaign only after they have enough impressions and clicks to judge, and prioritize terms that produced orders. Move winners into a manual campaign as exact match first, with a bid tied to your royalty per sale rather than to a guess. Send clear losers to negatives instead of keywords. Review the report on a fixed schedule so you act on patterns, not on single clicks. This guide is part of our wider KDP ads guide for authors running Amazon Ads on books.

What an auto campaign's search term report actually gives you

An auto campaign lets Amazon decide where your book appears. It matches your ad to shopper searches and to product pages it considers related, using targeting groups such as close match, loose match, substitutes, and complements. The search term report shows what shoppers actually typed or which products triggered an impression, along with the clicks and spend that followed.

This is the raw material for harvesting. You are not guessing keywords from a brainstorm. You are reading real shopper language that already produced impressions and clicks for your specific book.

The report has limits that matter for decisions:

  • It is a sample of behavior, not a verdict. A term with 3 clicks and 1 order is a lead, not proof.
  • It mixes keyword phrases with ASINs. Product page targets appear alongside typed searches and need separate handling.
  • It reflects your book's current state. Price, reviews, and category position all change what a click is worth.
  • Attribution can lag. Orders can post after the click, so very recent terms often look worse than they are.

If you want the report mechanics in isolation, our post on the KDP search term report covers where to find it. Here the focus is what to do after you have it.

Pulling and cleaning the report: the columns worth keeping

Export the search term report for a date range long enough to collect meaningful clicks, typically 30 to 60 days for a book with modest daily budgets. Then strip it down. Most exported columns do not help you decide anything, and a cluttered sheet hides the few numbers that do.

Keep a working sheet with only the columns that answer a harvesting question:

Column to keep Harvest question it answers How to use it in the working sheet
Customer search term What exact words did the shopper use? Sort alphabetically first to spot near-duplicates, then re-sort by orders.
Impressions Does this term have enough visibility to judge? Very low impressions mean wait. Do not harvest or negate yet.
Clicks Is there enough traffic to read conversion? Group terms into 0, 1 to 4, and 5 or more clicks to set priority.
Orders Did this term actually sell the book? Any term with orders goes to the top of the review list.
Spend What did the evidence cost? Compare to the royalty from one sale before judging.
ACOS Is the term inside your break-even ceiling? Read it only when orders exist. ACOS on zero sales is meaningless.
Royalties What did the sales actually pay you? Use your royalty per sale, or KDP Sync data if linked, to check profit.
Targeting type (keyword or ASIN) Is this a phrase or a product page? Split into two tabs so you do not harvest both the same way.

Before you analyze anything, clean the data. Merge duplicates caused by plurals or minor spelling differences, remove your own book title and author name into a separate branded list, and flag any term that is clearly off-topic for the book. Branded terms behave differently from discovery terms, so mixing them skews your read.

Bid-desk tip: add one calculated column for break-even ACOS, which is your royalty divided by your list price. Every ACOS you read then has a fixed yardstick next to it.

A step-by-step KDP ads keyword harvesting routine for one book

Run this routine for one book at a time. Blending several titles into one sheet makes it impossible to tell which book's page converted the click. If you manage several books or accounts, repeat the routine per title, ideally from one KDP advertising dashboard so you are not juggling exports.

  1. Set the date range. Pull at least 30 days of data, and avoid including the first few days of a brand new campaign when Amazon is still testing placements.
  2. Export and clean. Keep only the columns in the table above, merge near-duplicates, and split keyword phrases from ASIN targets.
  3. Sort by orders, then by clicks. Terms with orders are harvest candidates. Terms with several clicks and no orders are negative candidates. Everything else waits.
  4. Check relevance by eye. Search the term on Amazon. If the results show books unlike yours, an order may have been luck, and the term is risky to scale.
  5. Add winners to a manual campaign. Create or open a manual Sponsored Products campaign for that book and add each term as a keyword, starting with exact match.
  6. Set the starting bid from economics. Base it on what the term already cost per click, capped by what your royalty can support.
  7. Negate losers deliberately. Add clear non-converters as negative keywords in the auto campaign so it stops spending on them.
  8. Log the date and the terms moved. A simple note lets you judge the harvested terms against their old auto performance at the next review.

Manual work like this is where an InteliAds pricing comparison can be worth a look. InteliAds, a Sponsored Products manager, lets you filter targeting by spend, clicks, and orders, preview IF/THEN rules with a dry run before anything changes, and see royalty-aware profit when KDP Sync is linked. It supports the routine rather than replacing your judgment.

Deciding which search terms graduate to manual campaigns

Graduation should depend on evidence plus economics, not on a single lucky order. A term earns a manual keyword when it shows a sale, stays relevant to the book, and can be bid on profitably at your royalty level.

Consider this labeled hypothetical for a cozy mystery paperback priced at $14.99 with a $4.20 royalty per sale. The search term "small town cozy mystery series" collected 14 clicks, 2 orders, and $11.90 in spend over 45 days. Sales were $29.98, so ACOS was about 39.7%. Royalties were $8.40 against $11.90 of spend, a small loss of $3.50 on that term so far.

Should it graduate? Probably yes, but carefully. Two orders from 14 clicks is a conversion rate of roughly 14%, which is healthy for a book. The loss comes from the average click cost, so the answer is a lower manual bid in exact match, not abandoning the term. A break-even ACOS of 28% (royalty divided by price) tells you the bid must come down before the term can pay for itself.

Use these checks when sorting candidates:

  • Graduate: at least one order, a relevant phrase, and a bid that fits your break-even ACOS.
  • Watch: a few clicks, no orders yet, and a relevant phrase. Leave it running in the auto campaign.
  • Negate: many clicks, no orders, or a clearly mismatched topic.
  • Treat separately: any ASIN or product target, which belongs in a product targeting campaign with its own review.

Profit matters more than ACOS alone, because a term can look acceptable by ACOS and still lose money after printing costs. Our explanation of KDP royalties vs Amazon Ads sales shows why reported ad sales and actual royalties differ. When you are ready to run this routine across several books, you can create an InteliAds account and connect your ad data in one place.

Choosing match types and starting bids for harvested terms

A harvested term already has a history in your auto campaign, so use that history. Start with exact match for terms that produced orders, because it keeps the manual campaign close to the shopper language that sold the book. Phrase and broad match can come later, once the exact keyword has proven itself.

For the starting bid, work backward from your royalty instead of copying the auto campaign bid. Take the cozy mystery paperback from the previous section: a $4.20 royalty and a conversion rate of about 14% on "small town cozy mystery series". The break-even cost per click is roughly $4.20 x 0.14 = $0.59. The term averaged about $0.85 per click in the auto campaign, so a manual bid of $0.45 to $0.50 leaves room for a conversion rate that turns out lower than the small sample suggested.

Bid-desk tip: a low bid on a good keyword will often under-deliver, and that is cheaper than overbidding on two orders of evidence. Raise it in small steps after you have more clicks.

Harvested term evidence Starting match type Starting bid logic Where it goes
Relevant phrase, 1 or more orders, ACOS near break-even Exact Royalty x conversion rate, then shave 15 to 25% for a thin sample Manual keyword campaign for that book
Relevant long phrase, 2 or more orders, ACOS under break-even Exact, plus phrase after a few weeks Near the old auto CPC, capped by break-even CPC Same campaign, separate ad group if bids differ
Short generic phrase with orders but mixed relevance Exact only Below the auto CPC, tighten after 10 clicks Test ad group with a small budget
Single order, off-topic search results Do not harvest yet No bid until a second order or clearer relevance Stays in the auto campaign
ASIN or product page target Not applicable Judge the competing book and price first Product targeting campaign

In the Amazon Ads console, the setup looks like this:

  1. Open the manual Sponsored Products campaign for the book and choose Add keywords.
  2. Paste the harvested phrases and select Exact only.
  3. Enter a custom bid per keyword instead of accepting the suggested bid.
  4. Record the date, the old auto CPC, and the new bid in your working sheet.

Negating the terms that waste spend without deleting your auto campaign

The auto campaign is your discovery engine, so keep it running and prune it. Deleting it stops the flow of new search terms, and you lose the data that feeds the next harvest. Negative keywords let you cut specific waste while the campaign keeps exploring.

Here is a fresh labeled hypothetical. A $12.99 puzzle book earns a $3.10 royalty, so break-even ACOS is about 24%. The search term "free printable puzzles" collected 19 clicks, 0 orders, and $9.88 in spend over 40 days. That spend equals more than three full royalties with no sale, and the phrase signals shoppers who want free content. This is a clear negative.

Use a rough rule to separate noise from waste. Once a relevant term has spent about two to three times your royalty without an order, it deserves a negative. Here that threshold is $6.20 to $9.30, and the term is already past it. Irrelevant terms, such as a different genre or a "free" modifier, can be negated sooner.

  • Negative exact: use it for a specific losing phrase that has close relatives which might still sell.
  • Negative phrase: use it to block a whole pattern, such as "free" or "coloring" on a puzzle book.
  • Campaign level: apply the negative in the auto campaign so it stops matching those searches.
  • Review later: note why each term was negated, so a seasonal or changed-price case can be revisited.

To do this in the console, open the auto campaign, go to Negative targeting, add the terms as negative exact or negative phrase, and save. For bigger catalogs, InteliAds works as a Sponsored Products manager: you can filter targeting by spend, clicks, and orders, preview IF/THEN rules with a dry run before anything goes live, and see royalty-aware profit when KDP Sync is linked. It narrows the list you review, and you still decide what gets negated.

FAQ

Q: How many clicks or impressions should a search term have before I move it to a manual campaign? A: Any relevant term with at least one order is worth considering, ideally with 5 or more clicks behind it. Terms with plenty of impressions and no clicks are not harvest candidates, since nothing shows shopper interest yet.

Q: Should I harvest ASINs and product targets the same way I harvest keyword phrases? A: No. An ASIN reflects a competing book's page, price, and reviews, so judge it separately and add it to a product targeting campaign. Keyword phrases go into keyword campaigns with a match type.

Q: Do I keep the same term running in the auto campaign after I add it as a manual keyword? A: You can, but the two campaigns may compete for the same impression and blur your data. Many advisors add the harvested term as a negative exact in the auto campaign so the manual keyword gets the traffic and you can read its results cleanly.

Q: Which match type should a newly harvested search term start with, and why? A: Start with exact match. It mirrors the wording that already produced the order and limits spend on loosely related searches while the bid is still unproven.

Q: How do I handle search terms that get clicks but no sales when the book is new and has few reviews? A: Give relevant terms more room, perhaps up to three times your royalty in spend, because a thin page converts less. Negate only terms that are clearly off-topic, and recheck after reviews and your price settle.

Q: How often should I check the search term report without reacting to noise? A: Every two to four weeks is enough for most books, using a window of at least 30 days. Weekly checks tend to reward single clicks and cause overreaction.

Q: What should I do if my auto campaign produces almost no search terms to harvest? A: Check that the campaign is actually spending, that the bid is competitive, and that targeting groups are not all paused. Then give it more time or budget, and use related keyword ideas as a manual starting list in the meantime.

Q: When is it right to add a harvested term as a negative instead of a keyword? A: When the term has spent two to three times your royalty with no order, or when it clearly signals the wrong intent, such as "free" or another genre. If the term is relevant but only lightly tested, keep watching it.

Good kdp ads keyword harvesting comes down to evidence, exact match first, bids tied to your royalty, and negatives that trim waste while the auto campaign keeps discovering. If you want to run this across several books with royalty-aware views, See InteliAds plans or Create an account. Essential has no trial, while Pro and Publisher include a 7-day trial with a card required.

Browse more KDP ads guides on InteliAds.

See InteliAds plans: compare Essential, Pro, and Publisher (Essential has no trial; Pro and Publisher include a 7-day trial). InteliAds is a Sponsored Products manager for KDP authors: royalty-aware profit when KDP Sync is linked, user-written IF/THEN rules with dry run, and targeting filters for keywords, product targets, and search terms. Create an account.

Tools and guides: KDP ads optimizer · break-even ACOS calculator

Put these rules to work

Set IF/THEN conditions, run them every hour, and revert any batch when you need to.