Search guide
Keyword research for a business that sells to people, not to Google
Most keyword research produces a spreadsheet nobody opens again. This produces a page map: which terms belong together, which need their own page, and which are not worth writing for at all.
11 min read · 7 steps · By Aman Mishra · Written August 2026
What you end up with
A mapped keyword list where every term is assigned to a page, every page has one primary term, and the terms you decided against are written down with the reason.
Have these ready
- Google Search Console access for your own site
- A spreadsheet, or the keyword clustering tool
- Two hours, uninterrupted, for a small site
What it takes
11 min to read
7 steps, in the order below. Skip to the one you are stuck on.
Keyword research goes wrong in a specific and predictable way. Somebody opens a tool, types in a broad term, exports two thousand rows, sorts by volume, and hands over a spreadsheet. Nothing in that file is a decision. It is raw material presented as a conclusion.
What you want at the end is smaller and much more useful: a list of pages, each with one term it is trying to win, and a note about the terms you chose to ignore.
-
Start with what you sell, not with a tool
Write out the services you actually deliver and want more of, in the words a customer would use rather than the words on your website. This list is the spine of everything that follows.
Before any tool, write down the services you deliver and want more of. Not your full capability list. The three or four things you would like the phone to ring about.
Then write each one in the words a customer would use. This is harder than it sounds, because every industry has internal language that customers never type. A firm selling what it calls managed detection and response is being searched for as who monitors our network for hackers. A clinic offering aesthetic dermatology is being searched for by name of the treatment.
If you are not sure, three sources will tell you: your own enquiry emails, the notes from your last ten sales calls, and the questions your team answers on the phone most often. Those are customer words by definition.
-
Find what you already rank for
Export the last twelve months of Search Console queries. Anything already sitting between positions five and twenty is the cheapest ranking you will ever buy, because the page exists and Google already has an opinion about it.
Open Search Console, go to the performance report, set the range to the last twelve months, and export the query list.
This export is the single most valuable input you have, and it is free. It is not an estimate of what people search for in general. It is a record of searches where Google already showed your site to somebody.
Two things to pull out of it.
Terms in positions five to twenty. Google already considers your page a plausible answer and has placed it just outside where clicks happen. Moving a page from position 11 to position 6 usually takes less work than getting a new page to rank at all, and produces a much larger change in traffic than the numbers suggest.
Terms you did not expect. Every export contains searches nobody on your team would have thought to target. These are the most interesting rows in the file, because they are demand you discovered rather than demand you guessed at.
-
Read the results pages before trusting the volume
Search each candidate term and look at what Google chose to rank. If the first page is all national directories, or all information with no commercial page in sight, the volume figure is telling you nothing useful.
Take each candidate term and search it. Actually search it, in a private window, and look at what is on the page.
You are asking three questions.
What kind of page did Google choose? If the entire first page is blog posts, Google has decided this search is informational, and putting a service page against it will not work no matter how good the page is. If the first page is service pages, the reverse.
Who is ranking? If it is ten national directories and two marketplaces, an individual business is not going to break into that with a page. That is not a reason to give up on the term, but it does mean the route in is a listing on those directories rather than a page of your own.
Is anyone selling on this page? If the results contain no commercial pages at all, the searchers are not buyers, whatever the volume figure says.
This step is slow and it is the step people skip. It is also the one that prevents most wasted work.
-
Sort by intent rather than by volume
Split the list into people ready to buy, people comparing, and people learning. All three matter, but they need different pages and they justify different amounts of work.
Split what survives into three groups.
Ready to buy. The person has a problem now and is looking for somebody to solve it. Service words, urgency words, location words, brand-plus-price. These deserve your best pages and the most effort.
Comparing. Alternatives, versus, reviews, best, cost, how much. This person will buy, just not today. These pages convert less immediately and often influence the decision more.
Learning. How, why, what is. Mostly not buyers yet, sometimes not buyers ever. Worth writing for when the topic is close to what you sell and you can be genuinely useful, not worth writing for because the volume looked good.
A common failure is a site with fifty learning pages and two buying pages, which produces traffic charts that go up and an enquiry count that does not move.
-
Group the terms that mean the same thing
Terms that return substantially the same results belong on one page. Building a separate page for each phrasing is how sites end up competing against themselves.
Search two candidate terms and compare the results. If most of the same pages rank for both, Google is treating them as the same question, and one page should target both. If the results are largely different, they need separate pages.
This is what the keyword clustering tool does mechanically, by grouping terms that share enough words to be the same request. It runs in your browser and does not send the list anywhere. For a small list, doing it by hand and checking the results pages is more accurate.
The mistake this step prevents is worth naming. Building one page for emergency plumber and another for emergency plumbing service does not double your coverage. It splits your own signals across two weaker pages, both of which now compete for the same result.
-
Assign every group to exactly one page
Each cluster gets one page, and each page gets one primary term. Write the mapping down. This document is the deliverable, not the keyword list.
Now make the actual document.
One row per page. For each: the URL, the primary term, the secondary terms it should also cover, the search intent, and whether the page exists or needs writing.
Two rules make this document worth having.
One primary term per page. If two rows share a primary term, you have a decision to make now rather than a ranking problem to diagnose in four months.
Every group has a page. A cluster with no page assigned is either work to schedule or a decision not to bother. Leaving it unassigned is how it gets forgotten.
-
Write down what you decided against
Record the terms you rejected and why. Six months later somebody will suggest one of them again, and the reason will have been forgotten.
Last, a short list of terms you rejected, each with one line saying why. Too competitive. Wrong intent. Not something we sell. Dominated by directories.
This takes ten minutes and saves an argument. Terms that look attractive on a volume chart keep coming back, and without the note nobody remembers whether they were considered and dismissed or simply missed.
What the finished thing looks like
A small document, usually one page per twenty or thirty target terms, that says: here are our pages, here is what each one is for, here is what we are not doing.
You can hand that to a writer and get pages that make sense. You can hand the spreadsheet of two thousand rows to a writer and get nothing.
When you have done all 7
A mapped keyword list where every term is assigned to a page, every page has one primary term, and the terms you decided against are written down with the reason.