Start with a sheet of paper, not a tool
Before you open anything, write down what you sell - not the way you name it internally, but the way a customer would. You call it "installation of supply and exhaust ventilation systems"; the person searching types "apartment ventilation done for me". You need both, but the second one brings in enquiries.
Pull words from three sources you already have access to: chat threads with customers, call recordings or your sales notes, and reviews. That's where people describe the problem in their own words, and that's the foundation of your list.
Write down separately what you don't do. If you only install your own stretch ceilings and never repair someone else's, then "stretch ceiling repair" is a negative keyword. That list saves money on SEO now and on ads later.
Where the queries come from: the minimum toolkit
Start with a free keyword planner: type in your service and look at what people search around it. The most useful parts are the related-queries view and the location filter - search volume nationally and in your city differ by an order of magnitude, and you're paying for customers in your city.
The second source is the search box itself. Type your service, then add the letter "a", then "b", then "c", and let autocomplete show you the continuations. Those are live phrasings that real people typed, not a machine export.
Third: the competitors ranking in the top ten for your main query. Ignore the design, look at the structure - which service pages they have, what they're called, what made it into the headings. If three competitors all have a dedicated page for "emergency call-out", there's demand for it.
Paid tools speed things up and show you which words already send competitors traffic. For a 200-300 query list they're optional; for a store with thousands of products they save weeks.
Filtering out the junk: three passes
After collecting, you'll have a dump of thousands of rows and most of it is useless. Run the list through these filters in this order.
- Location. If you work locally, keep your own region plus queries with no city in them. "Order a website in Chicago" is no use to you if you're in Denver.
- Intent. Split queries into "I want to buy" ("hire a web developer", "landing page cost") and "I want to understand" ("what is a landing page"). The first group goes to service pages, the second to the blog. Mixing them on one page makes no sense - they need different answers.
- Junk and freebies. "DIY", "download", "free", "jobs", "employee reviews", other companies' brand names - all negative keywords. One exception: if you're an authorised dealer for a brand, keep its name.
Grouping: one group, one page
Next, sort the queries into groups so that each group answers a single question. "Order a landing page", "landing page development cost", "build a landing page" - that's one page. "Order an online store" is a different one, even though the service looks similar.
The check is simple: search both queries and compare the top ten. If the results mostly overlap, the queries belong in one group - search treats them as the same task. If the results differ, you need two pages.
Your site structure grows straight out of those groups. Each group gets its own section or article, with an H1 built from the group's main query and written like a human wrote it. If you end up with forty groups and five pages on the site, you've just found your priority list.
The deliverable is a spreadsheet, not a text file. Columns: query, volume, group, page URL, status. That's the document the copywriter and the developer will both work from.
When you don't need keyword research at all
If you run on word of mouth and referrals and almost nobody searches for your service - narrow B2B, tender-based contracts, one-off bespoke projects - don't spend money on this. Check the volume first: if your main query gets dozens of searches a month in your region rather than thousands, organic traffic won't pay the work back.
If the site is a month old and has three pages, a 2,000-query list is premature too. First a working site you're not embarrassed to send ads to, then expansion to match demand.
And if you need enquiries this quarter, a keyword list won't deliver them on its own - SEO takes months. The sensible move is a compact list of core queries, run ads against it, see which words actually produce enquiries, and build the full list on that data.
Three ways to check your contractor's work
Don't ask for "the keywords", ask for the spreadsheet with grouping and a page assigned to every query. An export from a keyword tool with no page mapping is raw material, not finished work.
Ask about negative keywords. If there aren't any, nobody cleaned the list by hand. Then ask to see 20 random rows from the middle of the table - that's usually where the junk shows up.
And check whether your real phrasings from customer conversations are in there. If the contractor never asked how your customers describe your service, they built the list for a generic business.
At EFIMOV DEV we start with a brief and a page structure: there's no point building a site before the keyword picture is clear, because you'll only rebuild it later. Starting-from prices for landing pages, corporate sites and stores are published on our site, the exact quote comes after the brief, and you talk directly to the person running your project.
Keyword research isn't an export from a tool, it's a spreadsheet where every query is tied to a specific page and stripped of everything irrelevant. Start with how real customers phrase things and check the volume in your own region - you may not need thousands of words, just five well-built pages.
Frequently asked
How many queries should a keyword list have?
As many as you have pages, plus the pages you plan to build. A landing page is fine with 20-50 queries in 2-3 groups; a services site with 10 sections needs 200-400; an online store runs into the thousands because of categories and filters. A 5,000-row list for a five-page site is money spent on nothing.
Can I do the research myself, without a contractor?
Yes, if the scope is small. A free keyword planner, search autocomplete and a Google Sheet cover the basics in an evening or two. The difficulty starts with grouping large lists and with the technical side - that's where manual work stops paying for itself.
How often should the list be updated?
Every six months, and always when you add a new service or product category. On top of that, check your search query reports in analytics and search console once a quarter: they show which words are already bringing people in, and you'll often find phrasings that aren't in your list.
Is the keyword list for SEO the same as for ads?
The base is shared, the use is different. Ads have stricter requirements for negative keywords and lean on the commercial queries; SEO also values informational queries for the blog. A practical order: build one master list, then cut two selections out of it.