Amazon Keyword Research for PPC

Amazon keyword research for PPC is not the same discipline as keyword research for Google, and treating it that way is the most common structural mistake in underperforming ad accounts. On Amazon, every keyword you bid on is a purchase decision you’re paying to enter, and the data that tells you which decisions are worth entering already lives inside your account — most sellers just never rank their sources properly. This guide covers the five keyword sources in order of reliability, how to classify intent before a term ever enters a campaign, the real math behind volume-versus-relevance tradeoffs, and how to seed launch campaigns when you have no history to mine. It’s the same workflow we run inside our Amazon PPC management practice.

Rank Your Keyword Sources — They Are Not Equal

Most sellers open Helium 10, export 500 keywords, and call research done. That’s backwards. Keyword sources have a reliability hierarchy, and the order matters because each tier answers a different question.

1. Search term reports: what already converts

Your Sponsored Products search term report is the only data source on the planet that shows real customer queries matched to your actual clicks, spend, and orders. Nothing a third-party tool estimates comes close.

Pull 60 to 90 days of data and sort three ways:

• By orders, descending. These are your proven converters. Any term with 3+ orders and an ACoS below your break-even margin belongs in its own exact match campaign with a deliberate bid.
• By spend with zero orders. Anything with 15+ clicks and no sale is a negative keyword candidate — but check whether the term is genuinely irrelevant or your listing just doesn’t answer it. “Stainless steel water bottle” failing for your plastic bottle is a negation. “Water bottle for gym” failing might be an image problem.
• By click-through rate at low volume. Terms with strong CTR but under 10 clicks are early signals worth feeding budget before the data confirms them.

If you’re running auto campaigns and broad match — and you should be, as harvesting engines — this report refreshes your keyword set every single week. Weekly harvesting is the discipline; the report is just the raw material.

2. Brand Analytics and Search Query Performance: what converts for the category

If you’re enrolled in Brand Registry, the Search Query Performance dashboard shows you something search term reports can’t: the full funnel for every query where your brand appeared — impressions, clicks, cart adds, and purchases, for you and for the market total.

This is where you find asymmetries worth money. A query where the category converts clicks to purchases at 30% but your share of purchases is half your share of clicks means shoppers see you, click you, and buy someone else. That’s a price, review, or content problem — fix it before you raise bids. The inverse — your purchase share exceeding your click share — means you convert above category rate and can afford to bid aggressively into that term because your true cost per order will beat competitors’ at the same CPC.

The Top Search Terms report (also in Brand Analytics) adds click concentration: if the top three clicked ASINs for a query capture 70% of clicks and you’re not one of them, page one for that term is a knife fight. If click share is fragmented, there’s room to take position with a strong offer.

3. Helium 10 and Data Dive workflows: coverage and volume estimates

Tools earn their place at tier three — after your own data, not before. Their job is coverage: finding relevant terms your campaigns haven’t discovered yet and attaching estimated search volume so you can prioritize.

A practical Helium 10 workflow: run Cerebro on your own ASIN plus your top three to five competitors, filter to keywords where at least two competitors rank organically in the top 15, and export. That filter matters — a term one competitor ranks for might be noise; a term three competitors rank for is category-defining. Then run Magnet on your two or three root terms to catch long-tail phrasing Cerebro’s ASIN-based approach misses.

Data Dive structures the same underlying data differently: it builds a master keyword list across a competitor set and shows relevancy scoring by how many top sellers rank for each term, which makes it the stronger tool for building launch keyword lists and listing SEO in one pass. Treat every volume number from either tool as a directional estimate, not a fact — they routinely disagree with each other by 2x on the same term.

4. Competitor ASIN reverse lookup: what you’re missing

Reverse ASIN lookup deserves its own tier because of how you use it, not what tool runs it. The question isn’t “what does my competitor rank for” — it’s “what does my competitor rank for that I don’t.”

Run the lookup on the competitor whose product most closely substitutes yours, then subtract your own keyword set. What remains falls into three buckets: terms you should be targeting and aren’t (gap), terms that apply to their variant but not yours (ignore), and their branded terms (a deliberate conquest decision with its own economics — branded competitor terms often carry high CPCs and conversion rates that depend entirely on your price position against theirs).

5. Root keyword expansion: the systematic long tail

The last tier is mechanical but productive. Take your two or three root terms — the irreducible noun phrases a shopper must use to mean your product — and expand each across the standard modifier axes: use case (“for travel,” “for kids”), attribute (“insulated,” “32 oz,” “BPA free”), audience, and problem (“leak proof”). Amazon’s own search bar autocomplete, checked in an incognito session, tells you which expansions real shoppers type.

Root expansion is how you find the 40-search-a-month terms no tool bothers to surface — individually trivial, collectively often 20 to 30% of an account’s profitable spend, because long-tail exact match clicks routinely cost half of what head terms cost.

While you’re expanding roots, capture the variant forms too: singular versus plural, common misspellings, and hyphenation differences. Amazon’s matching has gotten better at collapsing these, but exact match still treats some variants as distinct auctions with distinct CPCs — “water bottles” sometimes clears 15 to 20% cheaper than “water bottle” for identical traffic quality. Free money for ten minutes of list work.

Deduplicate and Score the Master List

Five sources produce one messy spreadsheet, so before anything enters a campaign, consolidate. Dedupe on normalized form (lowercase, trimmed, plural-collapsed), then score every surviving term on three columns:

• Relevance (1–3): would a shopper typing this be satisfied landing on your product? Score it against your actual variant — size, color, material — not your product category. This is a human judgment call; no tool does it for you, and it’s the column that matters most.
• Evidence: does the term have proof behind it — orders in your search term report, purchase share in Search Query Performance, multiple competitors ranking top-15 — or is it only a tool’s volume estimate?
• Volume tier (head / mid / long-tail): rough buckets are fine; the estimates aren’t precise enough to deserve decimals.

Terms scoring relevance 3 with real evidence go straight to exact match. Relevance 3 without evidence goes to broad or phrase for testing. Relevance 2 gets tested only at discounted bids. Relevance 1 gets deleted, no matter how big the volume number next to it is — that number is the bait in the trap.

Classify Intent Before a Keyword Enters a Campaign

Every keyword that survives sourcing gets an intent label, because intent dictates which campaign it belongs in and what you’re willing to pay.

The classification prevents the classic blended-campaign failure: a category term and a competitor term in the same ad group share a budget, the competitor term’s expensive low-converting clicks drain it by noon, and your profitable category term goes dark for the rest of the day. One intent class per campaign. It also makes bid optimization tractable, because every keyword in a campaign shares roughly one conversion profile and one target.

Search Volume vs. Relevance: Do the Math, Not the Vibes

The tradeoff every keyword list forces is volume against relevance, and the resolution is arithmetic, not judgment.

Your expected cost per order on a keyword is CPC divided by conversion rate. A head term at $2.50 CPC converting at 8% costs you $31 per order. A long-tail term at $1.10 converting at 20% costs $5.50. If your product nets $12 in contribution margin, the head term loses $19 per order and the long-tail term makes $6.50 — the “small” keyword is infinitely better, it just can’t scale past its volume ceiling.

So the practical rule: relevance sets your floor, volume sets your ceiling, and you build from the floor up. Fill your exact match structure with high-relevance terms first, regardless of volume, until you run out. Only then climb the volume ladder into broader terms — and when you do, enter them at bids derived from their expected conversion rate, not the conversion rate of your good keywords. If a term is half as relevant, it deserves roughly half the bid.

The one legitimate exception is rank-building: overpaying on a high-volume term for a defined period because organic rank on that term is worth the ad loss. That’s a strategy with a budget and an end date, not a default posture.

Seeding Launch Campaigns With No History

Launches invert the whole hierarchy — you have no search term reports and no Search Query Performance, so tiers three through five become primary. Here’s the seeding sequence we use:

Week 0: Build the master list from reverse lookups on the three to five best-selling direct competitors, cross-referenced through Data Dive or Cerebro. Score each term by how many competitors rank top-15 for it.

Structure: One exact match campaign with your 10 to 15 highest-conviction terms (high relevance, mid volume — the terms you can actually win). One broad match campaign with 20 to 30 terms as a discovery engine. One auto campaign, all four match types split into separate campaigns if budget allows, bid low, purely for harvesting. One product targeting campaign aimed at competitors with worse reviews or higher prices than you.

Bidding posture: Expect to run above break-even for 30 to 45 days. A launch listing with 5 reviews converts at maybe a third of its mature rate, so the target-ACoS math will tell you to bid low — ignore it temporarily on your conviction terms, because the point of launch spend is generating the sales velocity and conversion history that lifts organic rank. We walked through exactly this sequencing in our new product launch case study, which reached page one in 60 days on a seeded structure like this one.

Weeks 2–6: The moment search term data arrives, the normal hierarchy reasserts itself. Harvest weekly, promote converters to exact, negate waste, and let the launch structure collapse into a standard mature structure by week eight. If you’re launching, the broader product launch strategy — pricing, review velocity, inventory depth — matters as much as the keyword seeding; PPC can’t carry a launch alone.

Your Keyword Data Should Also Be Writing Your Listing

Here’s the compounding effect most sellers leave on the table: the same research that builds campaigns should rebuild your listing. Every keyword you’ve proven converts through PPC is a keyword Amazon’s organic algorithm should see in your title, bullets, and backend fields — indexation is a precondition for ranking, and PPC conversion data is the strongest possible evidence of which terms deserve the limited real estate. The workflow for translating campaign data into on-page placement is covered in our guide to Amazon SEO optimization, and the backend keyword field is where the long-tail terms that don’t fit your visible copy still earn indexation.

The loop runs both directions. Better keyword placement lifts organic rank, organic rank lifts conversion rate through social proof accumulation, and higher conversion rate lowers your cost per order on every paid click. Sellers who treat PPC keyword research and listing SEO as separate projects pay for the same insight twice.

The Cadence That Makes It Stick

Keyword research on Amazon isn’t a phase, it’s a loop: harvest search terms weekly, review Search Query Performance monthly, rerun tool-based competitive research quarterly. Accounts that run this cadence look completely different after six months — spend concentrates into proven terms, wasted clicks get negated before they compound, and the long tail keeps refilling as shopper language shifts.

Running that loop properly across a real catalog takes hours every week, which is exactly the work our Amazon PPC advertising service takes over — a free audit will show you the keyword gaps and wasted spend sitting in your search term reports right now.