Search term harvesting using bulk sheets

The workflow that turns Auto campaigns into Keyword campaigns

If bid adjustment is the highest-frequency bulk operation, harvesting is the highest-value. The workflow goes like this: your Auto campaign (or broad-match Keyword campaign) has been running for several weeks, generating impressions on a wide range of search terms. Some of those search terms converted; many didn't. You take the converters and create explicit, tightly-bid Keyword campaign rows for them — promoting your discovered winners into managed targeting where you can bid more confidently. You take the non-converters and add them as negatives, blocking future waste. One bulk operation captures both.

This article walks through harvesting end-to-end: where the data comes from, how to identify harvest candidates, the exact bulk-sheet rows you produce, and the decision rules that make the difference between a useful harvest and a noisy one.

What "harvesting" actually means

The basic insight: Auto campaigns are a discovery tool. They expose your product to search terms Amazon's algorithm thinks might be relevant. Most won't convert; a few will. You can't predict which in advance — only the data reveals which queries actually drive your customers.

Harvesting is the act of moving those discovered winners into Keyword campaigns where you have explicit control over each one. The benefits:

• You can bid more aggressively on a known-converter than you would on an unknown term.
• You can use exact match (or phrase match) instead of broad, capturing exactly the high-intent queries.
• You can isolate the converter from the Auto campaign's broader discovery, so the Auto campaign keeps exploring while your Keyword campaign exploits.
• You build a long-term inventory of proven keywords that compound over time.

The same logic applies in reverse for negatives: search terms that triggered impressions but didn't convert are waste candidates. Adding them as negatives prevents future spending on those queries.

Where the harvest data comes from

The Customer Search Term Report is the source. It's NOT in Bulk Operations — it lives in the Reports section of Seller Central, under Advertising → Reports → Search Term Report (or similar — Amazon's UI navigation shifts occasionally).

The report shows, for each keyword (or auto-targeting group) you targeted, the actual search queries that triggered your impressions, with per-query performance: impressions, clicks, spend, orders, sales, ACoS.

Pull the report for a meaningful window — usually 30–60 days. Shorter windows have insufficient signal; longer windows include data from before any recent bid changes (which may dilute your current performance picture).

Identifying harvest candidates: positive (keyword) harvests

Filter the report to rows with clear converter signal. A starting decision rule:

• Search term has 2+ orders in the report period, OR
• Search term has 1 order AND clicks ≥ 5 with ACoS at or below your target, OR
• Search term has 1 order AND CVR (orders / clicks) > 8% (well above category average)

These thresholds are starting points. Adjust based on your category, margin, and risk tolerance. Higher thresholds = fewer harvests = more confidence per harvest. Lower thresholds = more harvests = more risk of including noisy converters.

Exclude rows where:

• The search term is already an exact-match keyword in another Keyword campaign — you'd be duplicating.
• The search term is brand-related and you have a separate branded campaign covering it.
• The search term seems unrelated to your product even though it happened to convert (sometimes long-tail oddities — "best gifts for grandma" converting because grandma needed an air fryer — that won't repeat reliably).

Identifying negative candidates: terms to block

The reverse filter for negatives:

• Search term has 0 orders AND clicks ≥ 10, OR
• Search term has 0 orders AND spend ≥ 2× your average order value's ad spend, OR
• Search term has 1 order AND ACoS > 100% (the order didn't pay for the ad spend), OR
• The search term is obviously off-topic regardless of conversion data (e.g., your product is "air fryer" and you're showing on "blender" — block immediately, don't wait for click accumulation)

Again, starting points. Categories with high CPCs and low CVRs may need more relaxed thresholds; high-CVR categories can be more aggressive.

Building the harvest bulk sheet

Once you've identified your positive and negative candidates, the bulk sheet has two sections:

Section 1: New Keyword rows (positive harvests).

For each positive harvest, you create one Keyword row in your bulk sheet — typically in an existing "harvested keywords" ad group of an existing Keyword campaign (or you create a new campaign for harvests). The row structure:

• Campaign ID — the existing Keyword campaign you're adding into.
• Ad Group ID — the existing ad group within that campaign.
• State — enabled.
• Bid — typically higher than your Auto campaign's bid for that group, because you have proven conversion data. A common starting point: 130–150% of the Auto campaign's bid for the relevant targeting group.
• Keyword Text — the search term you're promoting.
• Match Type — usually exact for the cleanest signal. Some sellers also add the same term as phrase or broad in separate rows for ongoing discovery.
• Entity — Keyword.
• Operation — Create.
• Product — Sponsored Products.

One row per harvested term. Duplicate for variations (you might harvest a single term as both exact and phrase, in different ad groups).

Section 2: New Negative Keyword rows (negative harvests).

For each negative harvest, you create a Negative Keyword (campaign-level) or Negative Keyword (ad-group-level) row in the source campaign — the one where the bad search term was triggered. Structure per the negatives article: Campaign ID, State, Keyword Text, Match Type (negativeExact or negativePhrase), Entity, Operation, Product.

One row per negative. Choose negativeExact for specific terms; negativePhrase for variations.

The both-sides upload

Both sections live in the same bulk sheet upload. You're simultaneously:

• Adding new Keyword rows (positive harvests) to your Keyword campaigns.
• Adding new Negative Keyword rows (negative harvests) to your source campaign (usually the Auto campaign).

The upload processes both. Within 30 minutes, your new keywords start serving and your negatives start blocking. The Auto campaign continues running (it's not paused), but it's no longer triggering on the harvested terms — those now get served by the dedicated Keyword campaign with higher bids.

Try it: build the positive harvest rows

Once you've decided which search terms to promote into keywords (using the decision rules above), the builder below turns your list into ready-to-upload Keyword rows. Enter the destination Campaign ID and Ad Group ID, pick a match type and bid, paste your harvest list — get a .xlsx out. For the negative side (terms to block), use the negative keyword builder; the two outputs can be uploaded together or separately.

The harvest decision tree

For any given search term in the report, the decision flow is:

1. Does the term have a clear positive signal (orders, low ACoS, high CVR)? → Harvest as positive Keyword.
2. Does the term have a clear negative signal (high clicks/spend, no orders, high ACoS)? → Harvest as Negative Keyword.
3. Is the term obviously off-topic regardless of data? → Negative Keyword.
4. Is the term ambiguous (some signal but not clear)? → Leave it alone. Let the next harvest cycle's data clarify.

The "leave it alone" decision is underrated. Most search terms in any given report won't be harvest candidates either way. Discipline about thresholds prevents premature decisions on noisy signal.

Avoiding overharvest and underharvest

Two common failure modes:

Overharvest: Setting thresholds too low. Every search term with one click and one order gets harvested as a positive keyword. Every search term with one click and no order gets negated. Result: hundreds of low-signal keywords and negatives accumulating. The Keyword campaign becomes bloated; the Auto campaign starves because all interesting search terms are blocked or duplicated elsewhere. Your account becomes noisy and hard to manage.

Defense: meaningful click thresholds (≥ 5 for negatives, ≥ 10+ for high-confidence positives) before harvesting either direction.

Underharvest: Setting thresholds too high. You never harvest until a search term has 50 clicks and 5 orders. By then, the Auto campaign has been spending on the term for months while you watched data accumulate. You miss the opportunity cost.

Defense: realistic thresholds that balance confidence against latency. The "2 orders" rule is a common compromise — fewer than 2 orders is genuinely noisy; 2+ is meaningful in most categories.

The harvest cadence

How often to harvest:

• Monthly is the standard cadence. Pull the search term report for the previous 30 days, identify harvests, upload. Most well-managed accounts run this monthly.
• Initial harvest (first time on an account, or for a new product launch): include 60–90 days of data because there's more accumulated signal to mine.
• Event-driven: After a major change (new SKU launch, big bid adjustment, seasonal shift), do an off-cycle harvest 2–3 weeks after the change so the new performance data feeds the harvest decisions.

What happens after harvesting

For positive harvests: the new Keyword campaign's exact-match keywords start serving immediately. You'll typically see meaningful click volume within a week. The ACoS on the harvested keywords usually beats the Auto campaign's average because you're targeting proven converters with higher bids.

For negative harvests: the negated terms stop triggering impressions in the source campaign. The campaign's overall spend may drop slightly (you're blocking some spend) and overall conversion rate may rise slightly (you've removed waste). Effect is usually visible within 7–14 days.

The Auto campaign continues running and continues discovering new terms. The cycle repeats next month — Auto discovers, you harvest, Keyword campaigns capture the proven winners, negatives block the proven losers. Over 6–12 months, your account converges on a tight set of converters and a comprehensive negative list.

Common harvesting mistakes

• Harvesting from a campaign with too little data. An Auto campaign that's been running for two weeks doesn't have enough signal to harvest reliably. Wait at least 30 days.
• Harvesting the same term multiple times. If a term has already been harvested into a Keyword campaign, harvesting it again creates duplicate keywords. Always cross-check against existing keywords before adding new ones.
• Negating terms that converted. Filter discipline matters. A search term with 8 clicks and 1 order at 40% ACoS may look like a candidate for negation, but it converted — and the ACoS might improve with better bid management rather than blocking.
• Forgetting to set Operation correctly. Both positive (Create) and negative (Create) operations are creates, not updates. Setting Operation to Update on a new row makes Amazon look for an existing ID, fails to find one, and errors.
• Not setting match type explicitly. Both Keyword and Negative Keyword rows require Match Type. Leaving it blank fails.

Sponsored Display and SB harvest variations

The same idea applies to other ad types, with adjustments:

• Sponsored Brands harvest: Pull SB-specific search term reports. Harvest into SB Keyword campaigns. Negatives go on the source SB campaign. Workflow is the same.
• Sponsored Display harvest: Less common because SD's audience targeting doesn't have search terms in the same sense. But you can harvest from SD's placement reports — promote converting ASINs to dedicated SD product target campaigns, negate non-converters.
• Cross-product harvest: Sometimes a search term that converts in your SP Auto campaign also makes sense as a SB or SD target. Multi-channel harvesting is more complex but can be more thorough.

Next steps

Harvesting connects to:

• SP Keyword campaigns — the destination for positive harvests.
• Negative keywords in bulk — the destination for negative harvests.
• The monthly workflow — the next article, ties harvesting into the broader optimization cycle.

And the value proposition for BulkSheet Pro is at its strongest with harvesting: applying decision rules to thousands of search terms manually is exhausting. BSP reads your search term reports, applies your harvest thresholds, and produces the upload-ready sheet — with the positives and negatives separated, validated, and ready. Try it free →