Winner Strip

Amazon Keyword Harvesting: What It Is, How It Works, and How to Do It Right

Amazon Keyword Harvesting: What It Is, How It Works, and How to Do It Right


Definition

Amazon keyword harvesting is the process of identifying search terms from auto or broad match campaigns that are generating conversions, then adding those terms as exact match keywords in a dedicated manual campaign. The goal is to move your best-performing queries out of automatic bidding, where Amazon controls placement and cost, and into precise, intentionally-bid exact match campaigns where you control both.


 Your auto campaigns are a keyword research tool. Most sellers don't think of them that way; they run auto campaigns as a hands-off way to generate sales and leave it at that. But every converting search term in your auto campaign is data telling you exactly what real Amazon shoppers type when they're ready to buy your product. Harvesting that data and acting on it is one of the highest-ROI activities in Amazon PPC.

The numbers make the case plainly. Campaigns using a structured keyword harvesting workflow reduced average CPC by 14% while maintaining equivalent conversion rates over 90 days in accounts managed through Hector AI, not because the keywords changed, but because the same terms moved from uncontrolled automatic bidding into exact match campaigns where bids could be set precisely against the actual conversion data (Hector AI Internal Data, Q4 2025).

"Most sellers are sitting on a goldmine of converting search term data and not acting on it. Their auto campaigns run, terms convert, and nothing happens, the budget just keeps flowing into Amazon's automatic bidding instead of into a controlled exact match campaign with a deliberate bid," says Meher Patel, Founder & CEO of Hector AI and Amazon Ads Top 20 Globally Partner.

This guide covers why harvesting works, the exact process for doing it properly, and the mistakes that undermine results.

 

The System That Makes Keyword Harvesting Work

To understand why harvesting matters, you need to understand the difference between how auto campaigns and exact match campaigns operate.

In an auto campaign, Amazon decides which search queries to match your ad to, how prominently to display it, and what bid to apply. You set a default bid and a budget, and Amazon takes it from there. This is great for discovery; Amazon's algorithm is genuinely good at finding relevant queries you wouldn't have thought to target. But it's lousy for control. You can't set different bids for different queries, you can't cap spend on individual terms that are converting but expensive, and you can't prioritise your best terms over your mediocre ones.

Exact match campaigns are the opposite. You're bidding on one specific query, at one specific bid, with no ambiguity about what triggers the ad. Every optimization decision, bid level, budget allocation, and Amazon Sponsored Ads dayparting applies precisely to that term.

Harvesting combines both: you use the auto campaign as the discovery mechanism, then move proven performers into exact match campaigns where you can optimize them properly. The auto campaign keeps finding new terms. The exact match campaigns concentrate your best spend on what's already working.

 

Key Criteria for Keyword Harvesting in Amazon PPC

Not every converting term deserves its own exact match keyword. Harvesting indiscriminately creates bloated campaign structures that are hard to manage. There are two criteria a search term should meet before you harvest it:

        It's generated at least 1 conversion, ideally 2–3, to reduce statistical noise. A single conversion from a single click could be a coincidence. Two or three conversions from a term with reasonable click volume is a real signal.

        Its ACoS is at or below your target. If a term is converting but at 80% ACoS and your target is 25%, you don't want to give it its own campaign with a higher bid. You'd harvest it at a reduced bid or leave it in auto.

 

There are also terms worth excluding even if they meet both criteria:

        Terms with very low search volume that may not scale (5 clicks over 90 days is hard to optimise)

        Misspellings or oddly phrased queries, these often don't repeat reliably

        Terms that overlap too closely with keywords already in your exact match campaigns

 

Quick filter

Sort your Search Term Report by conversions descending. Focus on terms with 2+ conversions and ACoS at or below target. Those are your harvest candidates. Everything else either needs more data or isn't worth moving.

 
The keyword harvesting process, step by step

Step 1: Pull your Search Term Report

Go to Seller Central → Reports → Advertising Reports. Select 'Search Term Report' and download for the last 30–60 days. Longer windows give you more data but may include performance from a period when your bids or listings were different. 60 days is usually the sweet spot.

Important: the Search Term Report shows actual search queries that triggered your ads. This is different from the keyword report, which shows how your targeted keywords performed. You want the Search Term Report.

 

Step 2: Filter for harvest candidates

Open the report in a spreadsheet. Add filters, then:

1.     Filter by campaign type, focus on auto campaigns and broad/phrase match campaigns first. Your exact match campaigns should already be well-targeted, so they're less useful for discovery.

2.     Sort by conversions descending.

3.     Keep rows with 2+ conversions and ACoS at or below your target.

4.     Remove any terms already in your exact match campaigns (use VLOOKUP or a quick manual check).

 

What you're left with is your harvest list.

 

Step 3: Create a dedicated exact match campaign

Don't add harvested keywords to an existing campaign; create a new exact match campaign specifically for harvested terms. This keeps your campaign architecture clean and makes it easy to see how harvested keywords perform as a group.

Campaign setup:

        Bidding strategy: Down only (not Dynamic bids, up and down, until you've built a track record for these terms)

        Budget: Start conservatively, $20–40/day is enough to gather data without overexposing new terms

        Targeting: Exact match only

        Name convention: Something like 'SP, Exact, Harvested, [Month]' so you know exactly what this campaign contains

 

Step 4: Set bids using the target ACoS formula

Don't just copy the bid from your auto campaign. Calculate the correct bid based on what the term has actually delivered:

 

Bid formula

Max CPC = (Target ACoS % x Conversion Rate % x Average Order Value) / 100. Example: Target ACoS 25%, term converted at 12%, AOV $35. Max CPC = (25 x 12 x 35) / 100 = $10.50

 Use the conversion rate from this specific search term's data in the Search Term Report, not your overall account average. If the term has only 2–3 conversions, you might not have a statistically reliable CVR yet; in that case, use your product's typical conversion rate as a starting point and adjust after 2 weeks.

 

Step 5: Add the harvested terms as negatives in the source campaigns

This step is the one most sellers skip, and it's why their harvesting doesn't work properly. If you add a term to an exact match campaign but don't add it as a negative in your auto or broad campaign, both campaigns will keep bidding on the same query. Now you're competing against yourself, and whichever campaign has the higher bid wins the auction, regardless of which one is more efficient.

Add each harvested term as an exact match negative in the auto and broad campaigns it came from. This funnels all traffic for that query to your exact match campaign, where you control the bid.

 

Step 6: Keep the auto campaign running

Don't pause your auto campaign after harvesting. Its job is discovery; it should keep finding new converting terms for future harvesting cycles. The goal is a continuous loop: auto campaign finds terms → you harvest the winners → you negate them in auto → auto keeps looking for the next batch.

A well-managed auto campaign running at a controlled bid alongside a growing library of exact match keywords is one of the most efficient PPC structures on the platform.

 

Choosing the Right Harvesting Interval

For most sellers, every two to four weeks is the right cadence. Too frequent (weekly), and you don't have enough conversion data per term to make reliable decisions. Too infrequent (quarterly) and you're leaving months of converting queries earning you money in uncontrolled auto bids when they could be in a precise exact match campaign.

A practical trigger: when your auto campaign has generated 100+ conversions since the last harvest cycle, it's worth pulling the report. For high-volume accounts, that might be weekly. For newer products with lower traffic, it might be monthly.

 

Mistakes that undermine the results

Harvesting too aggressively

Some sellers harvest every term with even one conversion. The problem is that a single conversion is statistically almost meaningless; the next 10 clicks on that term might not convert at all. Harvesting low-confidence terms fills your exact match campaign with keywords that look good on paper but don't actually perform reliably, and makes your campaign data harder to read.

Not negating in the source campaign

Already covered above, but worth repeating because it's so consistently skipped. If the same term runs in both auto and exact match campaigns simultaneously, you're bidding against yourself and diluting your data. Neither campaign's performance accurately reflects what that term would do in isolation.

Setting the same bid as your auto campaign

Your auto campaign uses a default bid that applies to all queries equally. Some of those queries deserve higher bids (high conversion rate, high AOV), and some deserve lower ones. When you harvest a term into an exact match, you have the data to set the right bid. Use it.

Harvesting from broad match without adding phrase negatives

Broad match campaigns match a wide range of query variations. If you harvest 'wireless earbuds for running' and add it as an exact match keyword, but don't add appropriate negatives in your broad campaign, the broad campaign will keep matching to variations of that phrase. For clean data, add phrase match negatives for the core terms you harvest.

 

Automated Harvesting: Process and Workflow

Manual harvesting works well at 5–10 campaigns. At 50+ campaigns, it doesn't, you're reviewing search term reports across dozens of ad groups, calculating bids for each harvested term, and manually adding negatives in source campaigns. The process is the same, but the time involved becomes unsustainable.

Automated harvesting runs the same logic on a schedule: pull the Search Term Report, filter for terms meeting your conversion and ACoS thresholds, calculate target bids using the formula, and create the new exact match keywords, all without manual intervention. You review exceptions (terms flagged for unusual behaviour) rather than processing every term individually.

Hector AI's keyword harvesting workflow automates this across all Sponsored Products campaigns simultaneously, using AMC data to connect search term performance to the full conversion path rather than just last-click attribution. See how it works: hectorai.live/demo

 

Manual vs automated harvesting: What changes at scale

Factor

Manual harvesting

Automated harvesting

Frequency

Every 2–4 weeks

Continuous or weekly

Data lag

Days to weeks

Hours

Bid setting

Manual formula calculation

Rule-based, per term

Negative addition

Manual, often missed

Automatic on harvest

Scales to

10–20 campaigns comfortably

Hundreds of campaigns

Error rate

Human error on negatives

Systematic, consistent

 

The bottom line

Keyword harvesting is the bridge between the discovery phase of Amazon PPC, where you're testing broadly and letting Amazon's algorithm find relevant queries, and the optimization phase, where you're managing a curated set of high-intent exact match terms with precise bids.

Brands that run Amazon DSP and auto campaigns indefinitely without harvesting are leaving real money on the table. The converting terms are there in the Search Term Report; they're just not being managed. Moving them into exact match campaigns, setting bids based on actual conversion data, and negating them in source campaigns consistently produces lower CPCs, better ACoS, and a keyword library that compounds over time.

Hector AI automates keyword harvesting across all Sponsored Ads campaigns, pulling Search Term Reports, filtering by your thresholds, and setting bids via formula.

hectorai.live/demo

Frequently Asked Question

Traditional keyword research uses external tools to find search terms based on estimated volume and relevance. Keyword harvesting uses your own campaign data, specifically your Search Term Report, to find search terms that actual Amazon shoppers have already used to buy your product. Harvested keywords aren't estimates; they're proven converters. That's what makes them more valuable than most keyword research outputs.

Yes, for a different reason. Even if your current ACoS is good, money is still flowing through auto campaigns at uncontrolled bids for terms that could be bid more precisely in exact match. Harvesting isn't just an ACoS-reduction tactic. It's a margin-optimization tactic. Better bid control on your best terms means you can allocate more budget to them profitably, which grows revenue as well as improves efficiency.

Sponsored Brands campaigns generate their own Search Term Report, and the same harvesting logic applies: identify converting queries, and move them to exact match. The main difference is that Sponsored Brands' exact match campaigns tend to have lower conversion rates than Sponsored Products for the same terms, because Sponsored Brands lead to your brand store or a custom landing page rather than directly to a product listing. Factor that into your bid calculations.

This happens, and it usually means one of a few things: the conversion data from auto campaigns reflected a period of particularly strong competition or pricing, your bid in the exact match campaign is wrong (too high or too low), or there genuinely wasn't enough data behind the harvest decision. Review the term after 3–4 weeks. If it's consistently missing target ACoS after 10+ clicks, reduce the bid significantly or pause it and leave the traffic to auto.

Two to three conversions are a reasonable minimum. One conversion could be noise. Two or three from the same term, at reasonable ACoS, is a strong enough signal to act on. If the term has 5+ conversions, you can calculate a relatively reliable conversion rate and set a precise bid; the more data, the more confidence.

Post Comments

Book A Demo