How to Build a Keyword Map: Research, Cleanup and Search Intent Clustering
A useful keyword list is not a pile of phrases exported from a tool. It is a decision-making system that connects what people search for with the pages your website can genuinely offer.
The workflow is straightforward: define the scope, gather candidate queries, clean irrelevant entries, understand search intent, group queries into page-level clusters, and map those clusters to existing or planned URLs. The difficult part is not collecting thousands of terms; it is deciding which deserve a page.
Start with the business, not a keyword tool
Write down your products, services, service areas, target audiences and editorial expertise. For a hypothetical air-conditioning company, the starting categories might be installation, maintenance, repairs and buying advice.
Next, expand each category with natural modifiers: price, apartment, home, emergency, maintenance schedule, comparison and selection. Do not assume every combination is relevant. A company that only installs equipment should not build commercial pages around used-unit sales.
For a new website, sketch the likely architecture first. For an established website, export its existing URLs and annotate what each page actually covers.
Gather queries from multiple sources
Use customer questions, your own on-site search, product terminology, analytics and search performance reports. Google Search Console is useful for understanding queries already associated with your verified property. For Russian-language search research, Yandex Wordstat can help explore search phrasing.
Keep the source and collection date for each metric. Tools measure different things; a reported impression count is not automatically comparable to a keyword-volume estimate.
Collect specific long-tail phrases as well as broad topics. “Air conditioner installation” and “cost to install a split system in an apartment” are related, but the second makes the user's task much clearer.
Clean the data without deleting useful intent
Normalize whitespace, obvious spelling variants and duplicate rows. Then assess relevance: wrong geography, services you do not offer, unrelated product types and misleading informational requests.
Be careful with negative keywords. A phrase unsuitable for paid advertising may still be useful in an educational article. For example, “install AC yourself” might not belong on a professional installation landing page, but it could inspire a safety-oriented explanation of when professional work is needed.
Rather than deleting questionable phrases permanently, label the exclusion reason. This makes future reviews faster.
Identify the search intent
Four practical intent categories are:
- Informational: the user wants an explanation or instructions.
- Commercial investigation: the user is comparing options.
- Transactional: the user is ready to buy, book or request a service.
- Navigational: the user is looking for a specific company or page.
These are working categories, not rigid laws. A query such as “best air conditioner for a small apartment” may require an independent comparison, while “air conditioner installation cost” usually needs pricing context and service details.
For ambiguous terms, examine the actual search results in the target location. Are they service pages, product listings, editorial guides or aggregators? Record when you checked: search results can change.
Cluster around a page-level task
A cluster contains queries that one well-designed page can reasonably answer. Two phrases are not automatically one cluster just because they share nouns.
Consider this simplified example:
| Query | Likely intent | Possible destination |
|---|---|---|
| air conditioner installation price | Transactional | Installation service page |
| split-system installation cost | Transactional | Same page, if the offer is equivalent |
| how to choose an AC unit | Informational | Buying guide |
| air conditioner repair at home | Transactional | Repair service page |
| air conditioner maintenance | Mixed | Review the results before deciding |
These are illustrative examples, not measured search-volume data.
SERP-overlap tools can assist clustering, but a shared ranking URL does not prove that two queries should always have the same landing page. Editorial judgment, business scope and user expectations still matter.
Map clusters to URLs
Create a working sheet with the following columns:
| Field | Why it matters |
|---|---|
| Query and normalized query | Preserves the source phrase and clean version |
| Source, region and collection date | Keeps measurements interpretable |
| Search intent | Defines the user's task |
| Cluster | Groups related formulations |
| Target URL | Assigns page ownership |
| Action | Keep, improve, merge, create or exclude |
| Priority and rationale | Makes the implementation order transparent |
An existing page may already satisfy the cluster. Improve it rather than creating a competing article. If two pages address the same task, evaluate whether a merge would create a clearer resource. Do not create a separate URL for every word order or grammatical variation.
Prioritize by relevance, not volume alone
High volume does not guarantee value. Consider commercial relevance, the clarity of the user’s intent, current page quality, effort required, seasonality and how trustworthy the available data is.
A low-volume service query with an obvious customer need may be more valuable than a broad informational phrase that attracts readers outside your audience.
You can use high, medium and low priority labels with a short explanation. If you build a numeric score, document the weights; do not present it as a precise traffic forecast.
Turn clusters into useful page briefs
For each page, specify the main user question, related subquestions, content format, essential facts, internal links and review criteria.
A service page should explain the offer, eligibility, geographic scope and how to proceed. A guide should answer the question clearly, explain trade-offs and link to relevant products or services only where helpful.
Avoid writing briefs that merely prescribe keyword density. Natural language and complete answers matter more than repeated phrases.
Quality-control checklist
Before handing the map to developers or editors, verify that:
- Every retained cluster matches the actual business or editorial scope.
- Every cluster has a primary intent and one accountable target page.
- Ambiguous groupings have been checked against search results.
- Duplicates and irrelevant queries have documented decisions.
- Metrics include their source, location and date.
- Existing URLs have an explicit keep, improve or merge action.
- New pages have clear briefs and a sensible priority.
- Commercial and informational pages are connected through useful internal links.
What happens after launch?
Keyword mapping is not finished when the spreadsheet is approved. Publish or improve the priority pages, ensure they are discoverable, then observe search queries, impressions, clicks and user outcomes over time.
If a page underperforms, inspect the match between its content and the searcher's intent before adding more keyword repetitions. Review the map when products, locations or search behavior change.
The strongest keyword map is a maintained record of decisions: what users need, which page answers them, what still needs work and how progress will be evaluated.
This is an independent English adaptation of a Russian-language guide originally published on the Morlix CMS blog.
Top comments (0)