Keyword research is not collecting search volumes. It is working out which searches are worth pursuing, which you could realistically win, and what pages should exist as a result. This guide covers the process you can follow yourself.
Step 1: Start from the business
Before touching a tool, write down what you sell, who buys it, what problems bring them to you, and what a customer is worth.
That last figure decides everything. If a customer is worth €5,000, a term with thirty searches a month can be extremely valuable. If they are worth €30, it cannot.
Step 2: Gather the raw list
Sources, roughly in order of usefulness:
- Search Console, if you have traffic — real queries you already appear for, including ones you never targeted
- Your sales team or inbox — the words customers actually use, which are frequently not your marketing words
- Autocomplete and related searches — free, and reflects real behaviour
- Competitor pages — what they have built pages for
- Keyword tools — for volume estimates and expansion
Treat tool volumes as approximate. They are estimates, rounded, and often wrong for low-volume terms.
Step 3: Sort by intent
Group terms by what the searcher wants.
- Informational — "what is", "how to". They want an explanation.
- Commercial — "best", "comparison", "review". They are evaluating.
- Transactional — "buy", "agency", "near me", "price". They are ready to act.
- Navigational — a specific brand or site.
Check the current results for each term. If every result is a guide, that query has been judged informational and a service page will not break in, however well optimised.
Step 4: Judge whether you can win
For each promising term, look at who currently ranks. Are they large established sites, or ordinary businesses like yours? Is the content thorough or thin?
If the first page is dominated by major national players and your site is new, choose a more specific term. Long-tail queries convert better anyway, and they are winnable.
Step 5: Map terms to pages
Group related terms into clusters, then assign each cluster one page. This is the actual output.
The rule that matters: one page per intent. If two of your pages target the same query they compete with each other and both underperform. This is called cannibalisation and it is extremely common on sites that published without a plan.
Step 6: Check what you already have
Before writing anything new, check whether an existing page already targets the term. Improving a page that already has some authority is usually faster than starting over.
German-language specifics
German research is not translated English research. Compound nouns change how people search — one German word may cover an English phrase. Singular and plural behave differently. Umlaut and non-umlaut spellings both occur and should both be checked.
How many keywords per page?
One primary intent, plus the natural variations that belong to it. Not a list to insert mechanically.
Do I need paid tools?
Helpful, not essential. Search Console, autocomplete and actually reading the results pages get you most of the way.
How often should this be redone?
A full review annually, and whenever the offer or market changes materially.