Keyword Research: A Practical Step-by-Step Guide

Keyword research is the process of finding out which terms your customers search for on Google, scoring them by value and chance of ranking, and deciding which page answers each search. The result is not a long list of words but a simple map: term, intent, matching page, and priority. Without it, you write about what you think people care about, not about what they actually search for.

This guide is the one we apply in practice, in five steps. It does not assume expensive tools, just an order of work: start from the business, expand the list, evaluate, group by page, and prioritize. It is the foundation of on-page SEO, because every text, title, and internal link sits on top of this map. Search intent has its own guide, so we summarize it here and point you onward.

What should keyword research produce?

Successful research ends with a working document, usually a spreadsheet, where each row is a search that matters for your business. For each row, the document states:

• the term and its close variants;
• the intent of the person searching (to learn, to compare, to buy);
• the page that will serve it: an existing one or one to be created;
• the priority: what you do now, what you do later, what you ignore.

Without this document you risk two costly outcomes. Either you write articles nobody wants, or you create several pages for the same search and end up with cannibalization. Keyword research is not a one-time exercise at the start of a project; it is an instrument you keep refining.

Step 1: start from the business, not from the tool

There will always be tools. What is missing at the start is an understanding of the customer. Before opening any app, answer a few questions on paper:

1. What exactly do you sell? List your products or services under the names you use and the names your customer would use.
2. What problem do you solve? People rarely search for “your solution”; they search for their problem: “water in the basement,” “accountant for freelancers,” “how to get rid of mold.”
3. Who buys, and at what stage? A customer who is just learning what exists searches differently from one ready to sign.
4. What questions do you get by phone, email, or chat? These are searches in their raw form, written by real customers.
5. How do local customers find you? With the city or the neighborhood? With “near me”?

The output of this step is a seed list of 10–30 terms. You do not need more to get started. This is the part where the person who knows the business matters most: a tool does not know that in the local market “AC” and “air conditioning” are used interchangeably.

Step 2: expand the list from real sources

With the seed list in hand, you look for variants, questions, and adjacent topics. The sources below are free or cheap and complement each other:

For question ideas and variants in Romanian, the question and keyword generator builds a map from a single term, which you can then check in Search Console or Keyword Planner. If you are a foreign business entering the Romanian market, note that people often search in Romanian even when your product page is in English; check both languages before choosing.

One simple Search Console trick: in the Performance report, filter by your page and sort queries by impressions. Queries that sit at positions 8–20 with many impressions are often the quickest opportunity, because Google already associates your page with those searches. For more, see our guide to Google Search Console.

Step 3: score each term on three axes

Once you have a long list, you shrink it. A term is judged on three criteria, and none is enough on its own:

• Volume: roughly how many searches it gets. The easiest to see and the most misleading.
• Realistic chance of ranking: which pages sit on page one now and whether your site can compete with them. A new site does not immediately take on large portals.
• Commercial value: what a visitor is worth to your business. A search made 30 times a month that brings in a customer worth thousands of lei (the Romanian currency, RON) can beat one made 3,000 times a month that brings in nothing.

A simple system with no complicated formulas: rate value and winnability from 1 to 5, multiply them, and use volume only as a tiebreaker. Here is an illustrative example for a fictional air-conditioning company in Cluj:

The numbers are working conventions, not real data. What matters is the reasoning: the expensive, competitive term is not the first battle, while the modest but precise term can be.

Search intent: the criterion that decides the format

Two searches with the same words can call for different pages. “Air conditioner” might mean someone who wants to buy, learn how it works, or repair one. Before choosing the page, open Google, search for the term, and see what dominates the first page: guides, product categories, service pages, comparisons, local maps. The format that appears there is the one Google believes fits.

The practical rule: do not attack a commercial term with a blog article, or the other way around. The full guide to reading intent types is in the article on search intent; for now, just remember the manual check of the first results page, which takes five minutes per term.

Step 4: long-tail terms, clusters, and the page map

Long, specific terms (long-tail) usually have lower volume but clear intent and less competition. A new site typically wins its first positions here, and the authority it builds then helps with broader terms. This is not a guarantee, just a prudent strategy, which we also discuss in the plan for a new website in its first 90 days.

Then you group terms by topic into clusters. A cluster has a pillar page (the broad topic) and satellite pages (subtopics), linked with descriptive anchors. That way Google sees you cover the topic in depth and the user can find the way. How to distribute those links is covered in the guide to internal linking and site architecture.

The essential rule: one primary term, one page. For each group, choose the page that will be the “winner” and fold the remaining variants into it. If two pages target the same search, they get in each other's way; to find and fix that, see keyword cannibalization.

An example map for part of the same fictional company:

Step 5: prioritize and schedule

A good map is not executed all at once. Split it in three:

1. Now: terms with high value and good winnability, plus existing pages that can be improved quickly.
2. Next: high-value but difficult terms that need authority; these are worked in parallel with external links.
3. Later or never: high-volume, low-value terms and “vanity” terms unrelated to sales.

Set a realistic pace (for example two or three new pages a month) and revisit the map every quarter. To see how this choice ties into business goals, read what SEO is and how it works.

Notes for Romanian-language research

A few practical points for the Romanian market:

• Diacritics. Many users type without diacritics (“cat costa”). Google generally treats versions with and without diacritics as very close, but results can differ slightly; check both forms in Search Console and write content with correct diacritics.
• Inflection. Romanian has cases, numbers, and articles: “montaj,” “montajul,” “montajului.” You do not need to include every form, but natural text will contain many of them.
• English loanwords. Terms like “seo,” “backlink,” “crawl,” and “cloud” circulate in Romanian next to native alternatives. Check which form people use in Autocomplete.
• Small volumes. On a smaller market, many valuable topics have modest volumes. Do not reject a term just because it has a few dozen searches a month if its commercial value is high.
• Local. For local services, add the city and neighborhood; see also the local SEO guide.

Common mistakes

• Chasing volume. Terms with thousands of searches look attractive, but they are often vague, competitive, and unrelated to sales.
• Ignoring intent. The wrong format for the right term: an offer page for a “how to” search.
• A list with no prioritization. Hundreds of terms in a spreadsheet with no decisions help no one.
• Several pages for the same term. This creates cannibalization and does not bring extra traffic.
• Blind trust in volumes. They are estimates; use them comparatively.
• Vanity terms unrelated to your products or services.
• Copying competitors. A competitor may target the wrong thing; check the first results page every time.
• Doing the research once. The market changes and the map gets stale.
• Ignoring customer vocabulary. The technical term you use may differ from the one in actual searches.

When this does not apply, and the limits

• New topics may have no data. If you launch an entirely new product, tools will show zero; research the problem it solves and the questions in communities instead.
• Volumes are estimates. They differ between tools and guarantee no traffic; a term with a seemingly small volume may have many close variants.
• Position is never guaranteed. Research shows opportunities, not results. No one can promise a specific ranking.
• Conversational and AI search. Users ask AI assistants long questions that do not appear in volume tools. For these, rephrase your research as real questions and complete answers.
• Very local businesses have small volumes but high value; do not apply the minimum-volume thresholds used for national sites.
• Research does not replace good content. If the page is not genuinely useful, a well-chosen term will not save it.

Conclusion and next steps

Keyword research is a decision discipline, not a word-collecting one: you choose which searches deserve pages, in what format, and in what order. Start from the business, expand from real sources, score by value and winnability, group by page, and prioritize.

A plan for this week:

1. Write a seed list of 20 terms, using customer questions as well.
2. Pull from Search Console the terms with many impressions at positions 8–20.
3. Manually check the first results page for your 10 most important terms.
4. Build the table: term, intent, page, priority.
5. Pick the first three pages to create or improve.

If you want to turn the map into a full plan with texts, titles, and internal links, on-page SEO covers exactly that step from map to published pages.

Researching keywords for more than one country? Our international SEO services cover market-by-market keyword and content strategy.