Short answer: Keyword research is finding the exact words and phrases your customers type or say before they buy, then sorting them by relevance, competition, and how close the searcher is to a decision. In practice: list the questions customers actually ask, expand each one with Google autocomplete and the People Also Ask box, check what already ranks for it, and keep only the phrases that match what you sell and that you could realistically outrank the current top results for. It takes one afternoon and no paid subscription to do a usable first pass.
Fatima runs a small catering business out of a rented kitchen in Nairobi. For two years her website carried one page, titled exactly the way she would describe the business to another caterer: “premium bespoke culinary solutions.” Nobody searched that. Nobody says that.
Her cousin, who “knows computers,” had built the site and picked the wording. It looked sharp. It ranked for nothing, because it answered a question nobody was asking.
When she finally wrote down what customers actually said on the phone — “caterer for a small wedding in Nairobi,” “how much does a caterer cost for 50 guests” — and built pages around those exact phrases, the site started showing up within weeks. The business had not changed. The words describing it had.
Illustrative composite — not a client, not a testimonial
That gap between how a business describes itself and how a customer actually searches is the entire problem keyword research solves. Everything below is the concrete version of closing it, without assuming a budget for a paid keyword tool.
What keyword research actually is
Keyword research is the process of finding the exact words and phrases your customers type or say before they buy, and sorting them by how much traffic and business each one is realistically worth chasing. It is not a list of clever synonyms for your own product name — it is a record of someone else’s vocabulary, gathered on purpose instead of guessed at.
The output is not a spreadsheet you admire once. It is the input to every page title, heading, and opening sentence you write afterward. Skip it, and you are writing to please yourself; do it properly, and you are writing to answer a question someone is actually asking.
Where to find real keyword ideas
Four free sources cover most of what a paid tool would show you: Google’s autocomplete, the People Also Ask box, your own Search Console data, and the exact phrasing customers already use in reviews, DMs, and WhatsApp threads.
Type a seed phrase into Google and read the autocomplete suggestions — that is Google showing you, unprompted, what real people finish that sentence with. Open the People Also Ask box on the results page and keep expanding it; each question you click reveals more. Then check Search Console, if the site has any history at all — it shows the exact queries the site already gets impressions for, which is the most honest data available, free or paid.
Paid tools add estimated search volume, difficulty scores, and bulk export, which saves time once a content plan is running at real scale. They also come with dashboards colourful enough to make a modest local business look like it is launching a rocket, which is marketing for the tool, not evidence you need it yet.
| Source | Cost | What it is good for |
|---|---|---|
| Google autocomplete + People Also Ask | Free | Real phrasing, unlimited seed keywords, no account needed |
| Google Search Console | Free | Exact queries your own site already gets found for |
| Paid keyword tool (Ahrefs, Semrush, etc.) | $50–200+/month | Estimated volume, difficulty scores, bulk export at scale |
| An agency doing it for you | Built into a retainer | All of the above, prioritised and turned into a content calendar |
How to read search volume and competition
Search volume tells you how often a phrase is searched in a given month and market, and for most African cities that number is honestly low — often single digits to a few dozen — because the tools sample a whole country, not a city, and a single African market rarely produces the volumes a US-centric example would.
Do not discard a keyword just because the number is small. A phrase searched fifteen times a month by people actively choosing a caterer in Nairobi is worth more than a phrase searched fifteen thousand times a month by people in a country you do not serve. Competition matters more than volume at this scale: check who already ranks, and be honest about whether a small business site could realistically outrank them within a year.
It also helps to know why the number is small in the first place, rather than assuming the tool is simply wrong. Keyword-volume tools sample the clickstream and panel data available to them, and that data is thinner across most African markets than it is across the US or UK — fewer connected devices reporting in, more search happening through voice and WhatsApp shares that never register as a typed query. The number you see is a floor, not a ceiling. Treat it as a rough sort order between your own candidate keywords, not as a forecast of how much traffic any one of them will bring.
Matching keywords to search intent
Search intent is the reason behind the search — informational, commercial, or navigational — and a keyword is only worth targeting once you know which one it is, because the wrong page for the intent will not rank no matter how well it is written.
“What does a caterer do” is informational — the searcher wants an explanation, and a service page trying to sell them something too early reads as pushy and underperforms. “Caterer for a wedding in Nairobi” is commercial — the searcher is comparing options, and this is exactly where a clear service page belongs. “Fatima’s Kitchen Nairobi” is navigational — the searcher already knows the business and wants its site or its number, and the only job that page has is to load fast and be found; writing a persuasive pitch for someone who already decided is wasted effort. Confusing the categories is the single most common reason a page that looks fine still does not rank: it answers the wrong kind of question for the intent behind the search.
Prioritizing which keywords to actually target
A keyword is worth prioritizing when it scores well on three things at once: relevance to what you actually sell, low competition from sites you could realistically outrank, and a searcher who is close to a decision.
Score a shortlist against those three, drop anything that scores poorly on two or more, and target what is left first. This is also where the honest limit sits: if your business is one clear service in one town with a handful of regular customers, you likely need three well-chosen keywords, not thirty. A fourteen-day keyword research sprint for a business that size would be solving a problem you do not have yet.
A five-step process you can run in an afternoon
Keyword research for a small business site fits into five steps that take an afternoon, not a subscription.
- List what customers actually ask you. Write down the exact questions customers ask before buying, in their own words, from memory, WhatsApp threads, and reviews — not the words you would use to describe your own business.
- Expand each seed with autocomplete and People Also Ask. Type each phrase into Google, note every autocomplete suggestion, and open the People Also Ask box for related questions worth adding to the list.
- Check what already ranks for each phrase. Search the phrase and read the top three results to see what kind of page currently wins — a blog post, a product page, a directory listing — and whether you could realistically compete.
- Sort by relevance, competition, and buying stage. Score each keyword on how closely it matches what you sell, how weak the current top results are, and how close the searcher is to a decision, then keep only the keywords that score well on all three.
- Group survivors into one page per cluster. Group the remaining keywords by shared intent so each cluster becomes one page that answers the question thoroughly, instead of a separate thin page for every near-duplicate phrase.
None of this requires software beyond a browser and a spreadsheet. It requires the patience to write down what customers actually say instead of what sounds professional in a pitch deck — a discipline that, notably, no tool sells you, because there is nothing to charge for.
Turning a list into a content plan
A keyword list becomes a content plan once you group related phrases under one target page instead of writing a separate thin page for every variant. “Caterer for a small wedding,” “wedding catering cost,” and “how many caterers do I need for 50 guests” are one cluster with one strong page, not three weak ones competing with each other.
The same discipline that shapes a good general small business SEO planapplies here: one keyword answered thoroughly beats five answered thinly. If part of the cluster is local — “caterer near me,” “caterer in [city]” — that work folds into the same local SEO checklist rather than a separate research pass.
At a slightly larger scale, this grouping becomes a small topical map: one core page that owns the money keyword — the service itself — surrounded by a handful of outer pages that answer the supporting questions (cost, timeline, comparisons, how-to content like this one) and link back to the core page. A site with one page and no supporting content reads to both a person and a crawler as thin coverage of the topic, even if that one page is well written. A handful of pages that visibly cover the whole question reads as a business that actually knows the subject.
The mistakes that waste the research
The research gets wasted in three predictable ways: chasing volume over intent, doing it once and never again, and building a page for every keyword instead of a cluster.
Chasing volume means picking the biggest number on the list regardless of who is searching it or why — usually a keyword too broad and too contested to ever rank for. Doing it once means the list quietly rots as customers’ language shifts, new competitors appear, and an old assumption about what people search stays baked into every new page. And splitting one cluster into ten near-identical pages, an old habit from an era when more pages seemed like more visibility, mostly just competes against yourself. It is also, worth noting, the exact tactic anyone promising “guaranteed page one rankings” tends to lean on — volume of pages standing in for quality of any one of them.
Keyword research is also not a decision layer on its own. It tells you what to write about; deciding whether to fund that work through SEO or pay for clicks with Google Ads is a separate call, and whether the phrase worth chasing is what a human types or what a person now asks an AI assistant instead is another layer entirely — the underlying discipline of finding real customer language is the same one feeding both.
None of this is exotic. It is a list, honestly gathered, sorted by three plain criteria, and revisited before it goes stale — the same unglamorous groundwork every page on this site was built on.