The short answer
Keyword research is the complete list of queries people use to find you in search, mapped onto your site's pages and grouped by meaning. In plain terms: it's a map that answers two questions — what people search for in your niche, and which page on your site should answer each of those queries.
The key word here isn't "list" but "map". A pile of phrases with no structure is useless. Value appears when queries are grouped by intent and tied to specific pages: then you can see what you already have, what's missing, and where two pages accidentally fight over the same query.
An analogy that explains the rest
Imagine you're opening a shop and have to decide which departments it will have and what to put on the shelves. You could arrange the goods however suits you. Or you could first find out what people come for and in what words — and organise the shelves around their demand.
Keyword research is that second story. First you learn what people search for and how, then you build the site structure around that demand rather than around your own idea of what's "correct". A site built without keyword research is a shop where the departments are named with the company's internal terms while the customer searches with completely different words and leaves.
Where keyword research sits in the process
In my work, an SEO campaign is split into five stages, and keyword research comes first for a reason:
- Keyword research — which queries we're working towards at all.
- Technical audit — whether the site can actually serve those queries.
- Competitor analysis — what's already at the top and why.
- On-page optimisation — pages, structure, content.
- Off-page promotion — links and mentions.
The logic is simple: until it's clear which queries we're targeting, it's also unclear what to fix first in technical SEO, which pages to compare with competitors, and what to write about at all. Keyword research sets the coordinate system for every other stage. Doing it second or third means redoing work you've already done.
What keyword research actually involves
Keyword research isn't "export keywords from a tool". It's a series of consecutive steps, and skipping any one breaks the result.
| Step | What we do | Why |
|---|---|---|
| Collection | Write down every query: how the service is searched for, synonyms, question and commercial forms | Capture the raw demand without losing anything |
| Cleaning | Remove the irrelevant, other brands, junk phrases | Keep only what's genuinely about you |
| Clustering | Group queries with a single intent | One group = one page, no cannibalisation |
| Mapping to pages | Match clusters to existing and missing pages | Understand what to write and what to rebuild |
| Prioritisation | Sort by demand and commercial value | Start with what brings enquiries fastest |
Notice the clustering step — it's the least obvious one for clients. The queries "buy a sofa" and "how to choose a sofa" look similar, but the intent behind them differs: the first is commercial, the second informational. Merging them onto one page means fully satisfying neither. In my experience, mixed intent on a page is one of the most common reasons a site is "supposedly optimised but not growing".
What intent is and why it beats search volume
Intent is what a person actually wants when they type a query. And it matters more than volume (how many times a month a query is entered).
The temptation is understandable: take the highest-volume words and write for them. But a high-volume query with the wrong intent brings no enquiries. If you sell a service and the query is informational ("what is SEO"), you bring in people who aren't going to buy anything. Traffic appears, conversions don't. That gap between impressions and enquiries often has its root right here — in a keyword map built without regard for intent.
So in the map I tag each cluster by intent type: commercial (ready to buy/order), informational (researching the topic), navigational (looking for a specific brand). That instantly shows which pages will bring enquiries and which will work for reach and trust.
Chasing volume in isolation from intent and commercial value is the most expensive mistake at this stage. A thousand unstructured queries are worse than fifty properly grouped ones.
Keyword research in Estonia: three maps, not one
A separate issue almost every business in Estonia runs into. Here search is split by language: the Estonian, Russian and English versions of the same query compete for different audiences. Which means keyword research is done separately for each language.
The main mistake here is to translate the Russian map into Estonian and call it done. It won't work: Estonian-speaking users phrase their queries with different words, and a literal translation almost always misses the real demand. The Estonian map is built from Estonian Google suggestions and Estonian competitors, not from a translation. Which language to prioritise and why you can't get away with just one, I covered in the article on which language to target for SEO in Estonia.
In practice, for a multilingual site I build three independent keyword maps. The page structure usually matches across them, but the specific queries and wording do not.
What I see in practice
Let me be honest about the limits: there's no universal figure like "keyword research adds +N% traffic" — too much depends on the niche and the state of the site. But there are recurring patterns.
First: on sites built without keyword research, the structure almost never matches demand. There are pages nobody searches for, and no pages for queries with real demand. The site looks logical from inside the company and illogical from a search perspective.
Second: most of the value of keyword research isn't in new pages but in redistribution. It often turns out the needed queries are already being served somehow, but smeared across several pages that get in each other's way. Consolidating them into one structure is often more effective than writing something new.
Third, important for expectations: keyword research on its own brings no traffic. It sets the plan. Beyond it you need the technical conditions for the site to be indexed at all, and time, because SEO results take months, not weeks. Keyword research shortens the path but doesn't remove it.
How to tell your site needs keyword research
A quick self-check, no special tools required:
- Can you say off the top of your head which query each page of your site is written for? If not — the structure was built without keyword research.
- Open Search Console, the "Performance" report. Look at which queries you're found for. Are there many for which you have no dedicated page?
- Are there two pages on your site that answer roughly the same query? That's cannibalisation — a sure sign keyword research is missing.
- Was your section structure built from demand, or from how your company is organised internally?
- For the Estonian market: did you collect queries separately for the Estonian and Russian audiences, or translate one list?
If at least two points raised doubt — that's a reason to start SEO with building a keyword map. If the site is only being planned, it's cheaper to lay out the structure around keyword research during technical build rather than rework the sections later.
What to do next
- Demand first, structure second. Don't invent sections from your head — first find out what people search for and how.
- Group by intent, not by word similarity. One cluster = one page = one intent.
- Prioritise by value, not by volume. Start with the queries closest to an enquiry.
- For Estonia — a separate map per language. Translating someone's list doesn't replace doing the research.
