Work in four passes: harvest terms from the top competitors, group them by buyer intent, size each group by volume and relevance, then decide where each term lives in the listing or the campaigns. Score the finished list before you write anything. A list nobody graded is a guess with a spreadsheet.
The short version
- Harvest before you invent. The market has already told you what it types. Read that first.
- Group by intent, not by volume. A term someone uses while comparing is worth more than a term someone uses while browsing.
- Relevance beats size every time. Ranking well for a term that describes your product beats ranking poorly for a bigger one.
- Every term needs a home, either in the listing, in a campaign, or in the discard pile with a reason.
- Set stop rules per keyword before you spend. Deciding when to quit is easier before money is involved.
The four pass workflow
- Harvest. Take the ten to fifteen products that actually win your category and pull every term they rank and advertise for. Add your own vocabulary last, not first, because your vocabulary is the one thing you know is biased.
- Group by intent. Sort terms into three buckets. Discovery terms describe a category. Comparison terms include an attribute, a use case, or a constraint. Purchase terms name a specific configuration or specification. The middle bucket is usually where a new listing can actually win.
- Size and qualify. For each group, look at search volume, how well your unit matches the term, the price band of what currently ranks, and how many established sellers hold the top positions. A term where the top results are all rated well above you is not a first month target.
- Assign a home. Every surviving term goes to the title, the bullets, the backend fields, an exact match campaign, a phrase or broad campaign for discovery, or the discard list. If you cannot name the home, the term is not qualified yet.
Score the list before you write anything
Weight each criterion, score from 0 to 10, and multiply. A list below 70 is not ready to build a listing on.
| Weight | Criterion | What a strong list looks like |
|---|---|---|
| 20 | Relevance to the actual unit | Every term describes what arrives in the box |
| 18 | Harvested from winning competitors | Sourced from what ranks, not from a brainstorm |
| 15 | Intent grouping is explicit | Discovery, comparison, and purchase terms separated |
| 12 | Realistic competitive difficulty | You can plausibly rank for the priority set this quarter |
| 10 | Each term has an assigned home | Listing field or campaign named for every keyword |
| 10 | Backend and attribute coverage planned | Nothing important left only in the title |
| 8 | Stop rules defined per term | You know in advance what makes you drop a keyword |
| 7 | Locale correct | Researched natively in the language you are selling in |
The two heaviest weights are relevance and harvesting for a reason. Almost every bad list I have reviewed failed on one of them, usually by chasing a high volume category term that the product only loosely matches. That term brings clicks that do not convert, which drags conversion rate down and makes the listing look worse for everything else.
Stop rules, and why they matter more than the list
Here is where I learned this. Early on I poured money into a failing product for three months, convinced advertising would turn it around. It did not, and the reason it went on that long is that I had never written down what would make me stop. Every week the decision was made emotionally, and every week the answer was one more week.
That experience produced the criteria we now use to decide whether to scale, fix, or kill anything, from a product down to a single keyword. For keywords the version is simple. Before you spend, write down the number of clicks a term gets to prove itself, the cost per order that makes it worth keeping, and the conversion rate below which it is a discovery term rather than a revenue term. Then let the rule decide, not your mood at the end of a slow week.
Scale a term when it converts at or above your listing average with an acceptable cost per order. Fix a term when it converts poorly but the traffic is relevant, because the problem is usually the listing rather than the keyword. Kill a term when it has had a fair sample of clicks and the buyers arriving through it clearly wanted something else.
What most agencies will not tell you
A keyword list is easy to make impressive and hard to make honest. Most agencies will not tell you that a report of two thousand keywords is not evidence of work. It is evidence of an export, and a thousand of those terms will be irrelevant to your unit.
The question worth asking a candidate is the reverse of the one they expect. Ask what would make them recommend you stop bidding on a keyword, or stop selling a product entirely, and ask over what window they would decide. Anyone who answers cleanly has thought about your money as something that can be wasted. Anyone who only talks about growth has not.
Related answers
- Rank new Amazon listing fast without reviews
- How to test product-market fit on Amazon
- Ranking plan for Amazon Europe marketplaces
- What to prioritize: PPC vs SEO vs reviews in month one
- Amazon launch services: the complete guide
We will grade your current keyword list against this scorecard as part of the free audit at Flapen.
