Keyword research from scratch

You've decided to take search seriously, you open a blank spreadsheet, and the first question turns out to be the hard one: what do people actually type when they want something you sell? Most businesses answer it by guessing and end up with a list that is either too broad to rank for or too obscure to matter. Keyword research is a repeatable process rather than a brainstorm: start from what you sell, expand into the phrasings people really use, score each candidate, and cut to a shortlist you can win.

Start from what you sell, not from a tool

Every keyword tool needs a seed, and the seed decides the quality of everything that follows. Seeds come from your business, not from a database.

Write down, in plain words, each product or service you offer and the problem it solves. Then write down the words customers use when they call, email or leave a review. That last list matters most. A studio that rents wedding gowns but describes itself as "bridal couture" will never see the word "rental" in its own copy, and neither will any tool that reads the site. Searchers say rental.

Three places to harvest real customer language:

1. Sales and support conversations. The question a customer asks before they buy is a keyword in disguise.
2. Google Search Console, if your site has been live a while. Its Performance report lists the queries your site already appears for, including ones you never targeted.
3. Your competitors' navigation. The pages a rival chose to build are the searches they decided were worth a page.

Aim for a seed list of a couple of dozen short phrases. Don't judge them yet.

Expand each seed into the searches people type

Seeds are topics; keywords are the specific phrasings. "Boiler repair" is a seed. "Boiler repair cost", "boiler repair near me" and "boiler not firing up" are keywords, and each is a different person with a different need.

Google hands you the expansions. Type a seed into the search box and read the autocomplete suggestions, which Google's own explanation of the feature says are drawn from real searches. Run the search and look for the People Also Ask box: every question in it is something enough people ask for Google to surface it.

A suggestion tool does the same at scale, returning related phrases and long-tail keywords (longer, more specific searches that each get fewer searches but add up, and are usually easier to rank for).

Notice the modifiers that keep appearing: questions (how, what, why), comparisons (vs, best), locations (near me, a town name), price (cost, cheap) and urgency (emergency, same day). Each one changes who is searching and what kind of page can answer them. Expect a long, messy list at this point; that is what the step is meant to produce.

Score every keyword on demand, difficulty and intent

Three readings tell you whether a keyword is worth your effort.

Demand is monthly search volume. Treat it as an estimate that is right about scale and wrong about precision: it is modelled, rounded, and different in every market. A niche term that reads as zero can still be worth a page if it describes exactly what you sell and the few people searching it are buyers.

Difficulty estimates how strong the pages currently ranking are. It is comparative, not absolute: the same score means a different fight in a small local market than nationally. Compare difficulty across your own candidates rather than against a threshold someone posted online.

Intent is what the searcher wants, and you read it from the results page, not the words. Guides mean the intent is to learn; service pages or a map pack mean the intent is to hire or buy. Search intent covers reading a results page; the next lesson sorts intent by how close the searcher is to paying.

The trap is sorting by volume and taking the top. The highest-volume terms are usually the broadest, the most competitive and the least likely to convert. Modest demand, beatable difficulty and a searcher ready to buy beat far more volume with none of those three.

Cut to a shortlist you can win

A shortlist is what you can act on in the next few months, not everything you found. Apply these cuts in order:

1. Keep only keywords where your business is a genuine answer. A plumber in Leeds has no business targeting "how to become a plumber", whatever the volume.
2. Collapse near-duplicates. "Boiler repair Leeds" and "boiler repairs in Leeds" are one search to Google; keep one as the representative and note the rest as variants.
3. Favour the lower-difficulty end for a young site and park the biggest terms for later. A great deal is lost by spending a year on a keyword you can't yet win.
4. Keep a deliberate mix of keywords that bring buyers now and keywords that bring readers who become buyers later. All of one or all of the other is a strategy with a hole in it.
5. Research each market separately. Volumes, difficulty and often the keywords themselves change when the country changes.

Treat the shortlist as a draft. You'll reorder it after the next two lessons and add to it whenever the business adds a service or enters a market.

What to take away

• Keyword research starts from what you sell and how customers describe it, and a tool can only expand what you give it.
• Google's own autocomplete and People Also Ask box are a primary source for how people phrase a search.
• Every keyword gets three readings, demand, difficulty and intent, and volume on its own is the least useful of the three.
• A shortlist is relevant, winnable, deliberately mixed between buyers and readers, and researched per market.

Next

Now sort that shortlist by how close each searcher is to paying in The buying-intent ladder.