Why keyword research still decides everything else
Every search engine marketing campaign is built on a list of terms someone typed into a search box. Pick the wrong terms and the rest of the work — content, pages, bids, landing pages — is effort spent in the wrong place. Keyword research is not a formality you complete before the “real” work; it is the real work.
The goal is not a big list. The goal is a short list of terms where three things line up: people actually search for them, your business can answer them better than what currently ranks, and ranking or bidding on them leads to a customer rather than a curious visitor.
Start from the business, not from the tool
Most keyword research fails because it starts inside a keyword tool. Tools are fine for scale, but they cannot tell you what your business actually sells, who buys it, and what questions they ask before buying. Answer those three questions first, in plain language:
- What exact services do we sell, and which ones pay for the rest?
- Who is the customer — a homeowner with a broken boiler, or a facilities manager with ten sites?
- What do they type when they have this problem at 11pm, in their own words?
Write those phrases down before opening any tool. They become your seed terms, and they are usually better than anything a tool would suggest, because they come from real demand rather than autocomplete patterns.
Where to find real search language
Once you have seeds, expand them from sources that reflect actual behaviour:
- Your own site search logs and the exact queries in your Search Console performance report.
- Sales and support email — the wording customers use is usually rougher and more specific than marketing copy.
- Competitor page titles and headings, especially for services you know convert well.
- Autocomplete and “people also ask” suggestions, which show the phrasing of real follow-up questions.
Sort by intent, not by volume
Search volume is the most misleading number in keyword research. A term with hundreds of monthly searches can be worthless if the searcher is only after free information, while a term with a dozen searches a month can pay your rent for a year if the searcher is ready to hire someone today.
Group every keyword by what the searcher is trying to do:
- Navigational — they are looking for a specific site or brand. Not your fight, unless it is your brand.
- Informational — they want to understand something: “how does indexing work”, “what is a meta description”. Good for articles and guides; weak for ads aimed at buyers.
- Commercial — they are comparing options: “best SEO tools”, “SEO agency pricing”. Good for content that compares honestly, weak for direct sale pages.
- Transactional — they want to act: “SEO audit service”, “hire SEM consultant”. This is where service pages and paid clicks earn their money.
A healthy campaign has all four, clearly separated. Informational terms feed the blog and build trust. Transactional terms get their own pages and, when it makes sense, ads. Mixing them — running ads to a 2,000-word explainer, or burying a buyer inside an essay — is the most common waste in SEM.
Long-tail terms carry the budget
Broad head terms (“SEO”, “marketing”) are expensive, vague and usually dominated by sites with years of authority. The long tail — “SEO audit for small e-commerce site”, “Google Ads management for local plumber” — is cheaper per click, clearer in intent, and much easier to answer well. A few strong long-tail phrases rarely beat a head term in volume, but they reliably beat it in conversion rate. If your list is 90% head terms, redo the list.
Check what you are actually up against
Before committing a page to a term, search it yourself and look at what ranks. Not at metrics — at the results:
- Are the results homepages of strong brands, or articles and niche sites? You are matching page types, not just keywords.
- Does anything rank from a small or mid-sized site? If yes, the term is winnable with good execution.
- Do the results answer the question directly? Pages that only pretend to answer are beatable.
- For paid search, run the term through a few days of impressions and clicks before judging it. Actual data beats estimated difficulty every time.
Map keywords to pages, then stop
The final step of research is a simple spreadsheet: one row per target page, one primary keyword, two to four secondary terms it naturally covers. This mapping prevents two pages from fighting each other for the same query, and it tells you exactly what still needs to be written.
Then stop researching. Sites do not fail because the keyword list was missing three more variations; they fail because the pages targeting the good keywords were thin, slow or never finished. Research is the map; the content is the road.