Amazon Ads Keyword Research: How It Actually Works

What Amazon ads keyword research actually means

Amazon ads keyword research is the process of finding the exact words and phrases shoppers type into Amazon's search bar, then deciding which of those phrases are worth a bid in Sponsored Products or Sponsored Brands. It is not the same job as listing keyword research, though the two overlap — a term that converts in ads often belongs in your backend search terms too, and a term buried in Search Query Performance data is a candidate for both.

The confusion between the two is common. Sellers often run one keyword research pass, dump the list into the listing and the campaign, and call it done. That works as a starting point. It stops working once the campaign has real data, because ads keyword research is really two jobs: finding candidate terms before you have spend data, and mining your own search term report once you do.

Where the keywords actually come from

Every keyword research method sits in one of three buckets, and they are not equally reliable.

• Amazon's own data — Search Query Performance (inside Brand Analytics, for brand-registered sellers) and your account's own Sponsored Products search term report. This is the only source that tells you what a shopper actually typed before buying your product, not a competitor's. It is the highest-trust source in the whole process.
• Amazon's live surfaces — search bar autocomplete and Product Opportunity Explorer. These show what Amazon's engine is currently suggesting at a category level. Useful for finding phrases you hadn't thought of; not proof any specific phrase will convert for your product.
• Third-party tools — reverse-ASIN lookups and estimated search volume from vendors like Helium 10 or similar. These are discovery tools, not proof. Their volume numbers are modeled estimates, and different tools disagree with each other on the same keyword by a wide margin. Use them to build a candidate list. Never use them as the reason to raise a bid.

The order matters. Start with your own converting search terms if the campaign has run long enough to have them. Only go hunting with autocomplete and reverse-ASIN tools when you're pre-launch or expanding into territory the account hasn't touched yet.

A worked example: turning search terms into keyword decisions

Say a Sponsored Products campaign for a kitchen product has been running on broad match for three weeks against a handful of seed terms. The search term report shows three phrases worth a decision:

• "nonstick frying pan 10 inch" — 40 clicks, 6 orders, ACOS well under category average. Harvest candidate: pull it into its own exact-match keyword with its own bid, because broad match is diluting a top performer.
• "cast iron skillet" — 55 clicks, 0 orders. Looks like volume, wrong intent — those shoppers want cast iron, not the nonstick pan being advertised. Becomes a negative exact, not a bid increase.
• "pan" — 12 clicks, 1 order, ACOS double the target. Too broad to fix with a bid change alone; the fix is tightening the match type, not the number.

None of that shows up in a keyword tool's search volume column. It only shows up in the account's own search term report, after real spend. That's what every keyword-volume tool leaves out: volume tells you a phrase gets typed, not that it converts for your product.

The mistakes that waste the most budget

The single most common mistake is bidding on search volume instead of search intent. A keyword tool will show "cast iron skillet" as a bigger number than "nonstick frying pan 10 inch." A bigger number is not a match for what you sell — see the example above. Relevance has to be checked by a human reading the actual top organic results for that term, every time.

The second mistake is running keyword research once, at launch, and never touching it again. A keyword list from month one goes stale. Competitors enter, seasonality shifts, and the account accumulates search term data that should be reshaping bids monthly, not once a year.

We've made this mistake ourselves: trusting a third-party volume estimate on a new launch without weighting it against the account's early search term data, and letting a bid sit too high on a term that looked strong on paper and converted at half the rate of a lower-volume phrase nearby. The fix wasn't a smarter tool. It was a standing rule: no keyword or bid change runs without a reason grounded in the account's own data, a way to measure whether it worked, and a preset point to undo it if it didn't.

When the keyword data says something you don't want to hear

Sometimes the answer is that your best-guess keyword doesn't work. If a term you were sure would convert shows real spend and zero orders after enough clicks to be meaningful — usually somewhere past 15-20 clicks depending on your typical conversion rate — the right move is to negative it, not to keep raising the bid hoping it turns around.

Sometimes the answer is that the listing is the problem, not the keyword. If Search Query Performance shows a term generating impressions and clicks but low conversion relative to category benchmarks, the keyword is fine and the listing isn't answering the question that search term implies. That's a listing fix, not a bid fix.

And sometimes the answer is that the keyword research itself was wrong — a third-party volume estimate overstated demand, or a reverse-ASIN pull grabbed a competitor's irrelevant traffic. When that happens, don't patch it with a bid adjustment. Rerun the research against your own account's data and rebuild the list.

Where Dr. PPC fits into this

Everything above is manual work: pulling search term reports, cross-checking volume estimates, deciding what's a harvest and what's a negative, then doing it again next month. Dr. PPC is built to do that reading and proposing continuously across a full account, at $300 a month plus 3% of ad spend, capped, month-to-month, with the first 30 days free. Every proposed change carries three things before it runs: the evidence behind it, a measurement plan, and a rollback trigger — the same discipline described above, applied automatically instead of manually.

The reporting layer underneath it, Orbit, includes search-term and keyword tracking alongside inventory, ASIN profitability, and BSR data, and it's included at no extra charge rather than sold as a separate subscription. It's operated by Full Circle, which has managed more than $500M in revenue across 100+ brands. None of that replaces understanding the process above — it automates it once you already do.