Amazon Keyword Harvesting: The Complete Guide to the AUTO to PHRASE to EXACT Funnel

You run auto campaigns on Amazon. Clicks come in. Some convert, most do not. But do you actually know which search terms are driving your sales? And more importantly, are you putting those winning terms to work in tightly controlled manual campaigns? For most sellers the answer is no, and that gap is where the ACoS hides.

Amazon keyword harvesting is the process that turns raw search-term data into an optimized keyword portfolio. It is the backbone of a profitable Amazon PPC strategy and the difference between an account that drifts and one that compounds. This guide walks through it end to end, from the theory of the AUTO to PHRASE to EXACT funnel to the day-to-day practice.

What is Amazon keyword harvesting?

Keyword harvesting is the practice of extracting the high-performing search terms from your auto campaigns and moving them into manual campaigns. The goal is to graduate from loose, algorithm-driven targeting to precise targeting where you control the bid on every single term. It is keyword research powered by your own conversion data instead of a guess.

In an auto campaign, Amazon decides which search terms trigger your ads. Some are relevant and convert; others quietly drain your budget. Harvesting identifies the winners and isolates them in a controlled environment where you own the bid, the match type, and the negation. That control is what lets you scale spend without scaling waste.

Without harvesting, your AUTO campaigns stay an unfiltered blend of strong and weak terms. Your ACoS stalls because the good is buried under the bad. Harvesting is the mechanism that separates the signal from the noise, one search term at a time.

The funnel: AUTO, PHRASE, EXACT

The optimal harvesting funnel has three stages. Each stage filters terms against stricter and stricter criteria, so only the genuine winners reach the end of the process. Most tools skip the middle stage and jump straight from AUTO to EXACT, which lets weak terms slip through and misses valuable long-tail variations.

Stage 1: AUTO campaign (discovery)

The auto campaign is your net for catching search terms. Amazon picks the terms based on your listing (title, bullets, backend keywords). The bid is deliberately moderate to keep the cost of discovery low. The goal here is not immediate profitability but data collection.

Review your search-term reports over a 7 to 14 day window. Terms with at least one or two orders and an ACoS below your target are candidates for promotion to PHRASE. Terms with plenty of clicks and no conversions get added as negative keywords so the campaign stops paying for them.

Stage 2: PHRASE campaign (validation)

A winning AUTO term is harvested into a PHRASE campaign. Phrase match triggers the ad on the exact term and close variations that keep the word order, so it validates the term in a more controlled context while still capturing useful variants. This is the validation layer that EXACT-only workflows never get.

The moment a term lands in PHRASE, it is added as a negative in the source AUTO campaign. That avoids cannibalization: the same term should never run in two campaigns at once. After 14 to 30 days in PHRASE, terms that clear a higher sales threshold (say three orders) are promoted to EXACT.

Stage 3: EXACT campaign (performance)

Exact match triggers the ad only on the precise term (and very close variants like plurals). It is the maximum level of control. EXACT bids are usually higher because the term has already proven its profitability across two filtering stages, so it has earned a more aggressive bid.

Once a term reaches EXACT, it is added as a negative in the PHRASE campaign. The chain is complete: every term lives in exactly one campaign at a time. Budget concentrates on the best-performing terms with the tightest match type, which is precisely where you want your money to land.

Winners vs bleeders: how to tell them apart

A winner is a search term that converts with an ACoS at or below your target. A bleeder is a term that spends your budget without producing proportional sales. The line between the two rests on quantitative thresholds, not on gut feel.

Typical winner criteria: at least one order over the last 7 or 30 days, with an ACoS below your strategy's target. Those thresholds shift with your bidding strategy. In an aggressive mode you accept a higher ACoS to win share; in an economical mode the bar is stricter and only the cleanest terms qualify.

Typical bleeder criteria: more than 15 to 25 clicks with zero sales, or an ACoS far above your target over 30 days. The worst bleeders are terms with many clicks and no conversions at all; a second tier runs at two to three times your target ACoS. Both deserve a negative, but the order in which you cut them matters.

The judgment has to rest on enough data. Calling a term on three clicks is premature. Wait for at least 15 clicks (in economical mode) or 25 clicks (in aggressive mode) before declaring a term a bleeder. Patience here keeps you from cutting terms that were about to turn profitable.

Auto-negation: the anti-cannibalization safeguard

Cannibalization is the number-one problem in multi-campaign structures. If the same term is active in both AUTO and PHRASE, the two campaigns compete for the same ad slot and you end up bidding against yourself. Your CPC climbs, and your ACoS climbs with it.

Auto-negation solves this. Every time a term is promoted up a level (AUTO to PHRASE, PHRASE to EXACT), it is automatically added as a negative in the source campaign. The negation sits at the campaign level so the term is blocked across every ad group in that campaign, not just one.

Without auto-negation, harvesting creates more problems than it solves. Promoted terms keep burning budget in their original campaign, diluting the whole point of a tiered structure. It is a small technical detail that makes an outsized difference in the results.

Manual vs automated harvesting: the comparison

Done manually, harvesting means downloading the search-term report from Seller Central, filtering in a spreadsheet, identifying winners and bleeders, building the manual campaigns, adding the negatives, and adjusting the bids. For an account with 10 campaigns across 5 marketplaces, budget 3 to 5 hours every week.

Manual errors are common: forgetting a negative, promoting a term without checking its 30-day performance, mixing up negative exact and negative phrase. Each mistake quietly costs money for days before anyone notices it.

Automation runs the same process every day with no omissions and no slip-ups. Terms are evaluated over rolling windows (7 and 30 days), thresholds are applied consistently, and negatives are added immediately after each promotion. Nothing falls through the cracks because it never gets tired.

The cost of automation typically pays for itself within 2 to 4 weeks through reduced waste and better coverage of high-performing terms. The larger your catalog and the more marketplaces you run, the bigger the payoff.

The SellerPPC approach: strategy-aware harvesting

SellerPPC implements the full AUTO to PHRASE to EXACT funnel with thresholds tuned to each bidding strategy. In aggressive mode the promotion thresholds loosen to maximize coverage; in economical mode only the strongest terms get promoted. The funnel adapts to the goal instead of forcing one rule on every account.

Bid optimization uses a proportional formula: the new bid is computed from the ratio between the target ACoS and the term's actual ACoS. No arbitrary plus 10 percent or minus 15 percent moves. Every adjustment is mathematically calibrated and held to a 5 to 7 day cool-down so the data has time to stabilize.

Negatives are added at the campaign level (not ad group) for a complete block, and every action lands in an audit log. You keep full visibility over what was harvested, what was negated, and what was optimized, so the automation stays transparent instead of becoming a black box.

Frequently asked questions

How long does keyword harvesting take to show results?

The first AUTO to PHRASE promotions happen after 7 to 14 days of data collection. PHRASE to EXACT promotions follow 14 to 30 days later. Plan on 60 to 90 days to build a solid EXACT keyword portfolio with a proven track record. The process is continuous: new terms are discovered all the time by the AUTO campaigns, so harvesting never really stops.

Should I keep the AUTO campaigns running after harvesting?

Yes. AUTO campaigns stay active to keep discovering new search terms. Shopper habits shift, new terms emerge, and seasonal trends come and go. An AUTO campaign with its negatives kept current is your permanent radar for catching fresh opportunities before competitors do.

How do I avoid cannibalization between campaigns?

The rule is simple: a term should be active in only one campaign at a time. Every promotion (AUTO to PHRASE, PHRASE to EXACT) is paired with a negation in the source campaign. Use negative exact at the campaign level for a complete block, and periodically verify that no term is live in two campaigns at once.

What is the difference between keyword harvesting and keyword research?

Traditional Amazon keyword research starts from guesses, competitor analysis, and keyword tools before you have spent anything. Keyword harvesting starts from your own live conversion data: it surfaces the exact terms real shoppers used to buy your product. Both have a place, but harvested keywords carry proof of intent that researched keywords cannot match.