Keyword Research With Little Effort: Which Words Customers Really Use

Why is a good page with the wrong words not found?

A business calls its service a “compressed air audit”. Its customers search for “compressor losing air” or “find compressed air leaks”. The page is thorough, but it speaks a language hardly anyone types into a search box. It therefore rarely appears, and when it does, for the wrong queries. The problem is neither the technology nor the length of the text but the vocabulary.

Professional keyword research tools quickly cost several hundred euros a month and deliver thousands of terms. For a small business with five to ten services, that is too much and at the same time too little, because the tools do not know the conversations with the business’s own customers. That raises one question: How do you find out with little effort which words your customers really use, and which of them deserve a page of their own? To answer it, we first clarify the terms and a model from research on human-computer communication. Then we look at what a study shows about specific and generic search terms, go through the free sources and end with a three-column method.

The search engine does not find what you offer. It finds what you call it, in the way your customer calls it.

Terms and thinking model: the vocabulary problem

A search term (keyword) is the word or phrase someone types into a search engine, such as “compressor maintenance contract cost”. Search volume is the estimated number of searches for this term per month. The long tail refers to the many longer, more specific terms that are rarely searched individually but together account for a large share of all searches. Keyword research is the work of finding these terms, sorting them and assigning them to the pages of the website.

Why this work is necessary is shown by a classic study. Furnas, Landauer, Gomez and Dumais (1987) had people spontaneously choose words for objects and actions in five application domains, such as commands in a text editor or categories in a directory. The variety was surprisingly large. In every domain, the probability that two people favored the same word was below 20 percent. The authors called this the vocabulary problem. According to their simulations, a system that knows only its designer’s favorite word fails in 80 to 90 percent of attempts in many common situations. Their recommendation was to allow many alternative words.

The study is old and concerns computer commands, not search engines. The pattern can be transferred, however: the business is the designer, its website is the system, and the customer types in their own word. Modern search engines now recognize many synonyms. They still prefer pages that clearly treat the subject in the searchers’ words. So the website does not have to solve the vocabulary problem on its own, but it should not make it bigger.

Two views: tool data or customer conversations?

The first view says that search terms should be measured, not guessed. Only tools with search volume show which words are frequent. The second view says that tool data is imprecise for small markets, and that sales hears the words with buying intent every day. Both have a true core. Tools show how many people use a word; conversations show whether they are the right people. I therefore consider the order important: first collect the words from conversations, then check with search data which of them are searched. Anyone who starts the other way round easily optimizes for terms that many search for and nobody buys.

Generic or specific terms: what a study shows

Lehmann-Zschunke and Ampler (2026) examined which properties of a search term make clicks and purchases more likely. Their data come from a German online retailer that supplies major customers in the clothing industry, so from business-to-business trade. They analyzed 823 search terms from search ads between September 2019 and February 2022, which together achieved 753,965 impressions and 98,909 clicks.

The clearest result concerns the specificity of a term, meaning how narrowly it describes a product, for example with a brand and further details. More specific terms had a clearly higher chance of a click and a higher chance of a purchase. The mere length of the term in characters, by contrast, had no demonstrable influence on click probability. And too many product details in one term had a negative effect, presumably because hardly anyone searches that way. The authors interpret this with the buying decision in phases: those further along search more concretely.

The limits are clear. It is a single retailer, paid ads rather than the normal search results, and products rather than services. The direction, however, fits what the article Understanding search intent describes: generic terms bring many searches, specific terms bring the people who are close to a decision. A term with 50 searches a month that describes exactly your service is often worth more to a business than one with 5,000 searches that is mostly driven by curiosity.

Which free sources show the customers’ words?

Five sources are enough for research that suffices for a small business. They cost no money, only a few hours of time spread over two weeks.

Search Console is the most valuable data source because it shows real queries for your own website. How to read the performance report is described in the article Understanding Google Search Console. Queries with many impressions and few clicks are particularly revealing: someone is searching for your topic there, but your page does not answer with the right words.

The Keyword Planner is part of Google Ads and can be used free of charge once an account is set up. Accounts without active ad spend, however, usually see search volume only in ranges, such as “100 to 1,000”. That is enough to know whether a term is searched at all. It is not enough for fine comparisons. Anyone who runs search ads anyway gets more precise values and, in addition, real search terms from the ads report, as the article Search ads with a small budget describes.

The three-word check: customer word, search word, page word

From the vocabulary problem I derive a simple method, the three-word check. For each service, three words are placed side by side: the word customers use in conversation, the word that is searched, and the word that is on your own page today. Where all three match, there is nothing to do. Where they diverge, that is where the work lies.

1. List the services

   Write down the five to ten services the business earns money with, starting with the one that brings the best orders.

2. Collect customer words

   For each service, enter the words customers use in conversations, inquiries and emails, word for word and with variants.

3. Check search words

   Check each customer word in search suggestions, Search Console and Keyword Planner: is it searched, and which spelling is more common?

4. Add the page words

   Note which term is in the title and main heading of the matching page today.

5. Decide

   Put the searched customer word into the title and heading. If there is no page for an important query, mark it for the content plan.

The check is deliberately kept small. It does not replace a list of thousands of terms, but it answers the question that matters to a business: does the page for my most important service speak the language of those who want to buy it? For five services it takes about half a day.

Which terms deserve a page of their own?

At the end of the research there are usually 30 to 60 terms. Not every one needs a page. Two tools already described on this website help with the selection. The first are the four search types from the article SEO for SMEs: knowledge question, comparison question, service plus location and company name. Each term is assigned to a type, and each type calls for a different kind of page. The second is the results page test: what Google shows at the top for a term today reveals which kind of page is expected.

After that, the question-page principle applies: one question, one page, the best answer. Terms that express the same intent end up on the same page, even if they are worded differently. “compressor maintenance cost” and “what does it cost to service a compressor” are two wordings of one question. Two pages for it would weaken each other.

For the order, the same rule applies as for pages in general: first the terms close to an order, meaning service plus location and specific comparison questions, then the knowledge questions. How to rate knowledge questions by frequency, search demand, closeness to an order and effort is described by the topic grid in the article Editorial plan for articles. Short questions that do not need a page of their own belong on the FAQ page.

What follows beyond the individual case

Keyword research is often seen as a technical task for agencies. At heart it is listening. A business that knows its customers’ words writes not only better pages but also better quotes, ads and conversation guides. Conversely, a large gap between page word and customer word usually shows that the business describes what it does from its own point of view. This is the same observation that the article B2B website copy describes as the difference between in-house jargon and customer language.

The limits are clear. The study by Furnas and colleagues concerns computer commands from the 1980s, the study by Lehmann-Zschunke and Ampler paid ads of a single retailer. Search volumes from the Keyword Planner are estimates, very rough for small terms. And with AI search, queries become longer and more conversational, as studies on search intent suggest. That does not speak against the research but for its foundation: anyone who knows their customers’ questions in their words is also prepared for longer queries.

Conclusion and a first step

The question was
to find out with little effort which words your customers use and which deserve a page of their own. The answer: by first collecting the words from conversations and inquiries, checking them with free search data and placing them next to the words on your pages in the three-word check. Pages are deserved by terms that express their own intent, are searched and are close to an order. The vocabulary problem explains why this is necessary: people name the same thing very differently, and a website that knows only its own word is not found by many.

• Write down the five to ten most important services.
• Collect customer words from conversations and inquiries word for word for two weeks.
• Note the queries with many impressions and few clicks in Search Console.
• Check each customer word in search suggestions and the Keyword Planner.
• Fill in the three-word check for the most important service.
• Change the title and main heading of that page to the searched customer word.
• Add terms without a matching page to the content plan.

A business’s best keyword list is not in a tool. It is in the notes from its last twenty first conversations.