How to Research Keywords for SEO: Work the Four-Column Board
Keyword research for SEO is a filtering job, not a brainstorming session. You begin with a handful of phrases you already understand, expand them into the language buyers actually type, then cut hard until a short list remains that you can defend to a client or a founder. The board below is the one we use: four columns, each with an exit condition, so the work stops when the list is good enough instead of when everyone is tired.
Bring five seed phrases. We will tell you what they are worth.
How Do You Research Keywords for SEO, Start to Finish?
Write five seed phrases that describe what you sell, expand each one with questions, modifiers, and competitor terms, then filter every candidate on three tests: real intent, realistic ranking difficulty, and a plausible path to revenue. Finish with twenty to forty terms you can assign to pages this quarter.
The Board, Column by Column
Each column has a different job and a clear exit condition. Skip one and you either drown in rows or write pages nobody searches for.
Five to ten phrases in the customer's words. Pull them from sales calls, the service list, and the words buyers use before they know your product name.
Write what the buyer says, not what the brochure says.
Add questions, modifiers, comparisons, and place qualifiers. Autocomplete and related searches do most of this for free.
Breadth first. Volume estimates can wait until filtering.
Cut on intent, then difficulty, then relevance to revenue. Anything you cannot answer with one honest page leaves the board here.
If the intent is unclear, the keyword is not ready.
Group by intent so one page owns one cluster. Assign an owner, a quarter, and the page type before anything gets written.
Twenty good terms beat two hundred rows nobody opens.
The Filter: Three Tests Every Keyword Must Pass
This is where most research goes wrong. People sort by volume because volume is the number a tool puts first. Intent is what decides whether the traffic can ever convert.
| Test | What a pass looks like | What to reject |
|---|---|---|
| Search intent | One job per query. The searcher wants to learn, compare, or buy, and the top results agree on which one it is. | Mixed SERPs. Half shopping results, half long guides. You would be guessing which page type wins. |
| Ranking difficulty | Visible gaps. Thin pages, outdated pages, or forums ranking on page one, and your site already covers the topic. | Fortress SERPs. Marketplaces and national brands holding every slot with pages you cannot outbuild this year. |
| Commercial relevance | One step from money. The reader's next question is your service, your product, or your calculator. | Curiosity traffic. Interesting, unqualified, and impossible to route anywhere useful. |
Board full, shortlist still empty?
Send us your seed list and your three closest competitors. We will return a filtered shortlist with intent labels and page assignments.
The Final Move: The Shortlist Check
Before a term leaves the board, it has to survive six questions. Answer them in writing, in the same row, so the next person can pick up the sheet without a meeting.
- ✓One primary intent. You can name the job in six words or fewer.
- ✓One owning page. No other URL on the site already targets it.
- ✓Top three analysed. You know what each result does well and where it is thin.
- ✓Answerable in one page. The topic does not need a microsite to be useful.
- ✓Internal links planned. Two or three existing pages can point at it with a reason.
- ✓Owner and quarter. A name and a date, or it is not on the board.
Doing the whole board without paid tools
You do not need a subscription to run this process. You need a browser, a spreadsheet, and the discipline to stop at the exit condition of each column.
- Autocomplete. Type each seed, then every letter of the alphabet after it.
- People also ask. Open three questions per seed and follow one level deeper.
- Search Console. Filter queries with impressions but a weak average position: those are pages close to breaking through.
- Competitor sitemaps. Read post slugs, not just titles; they show the structure behind the traffic.
- One volume check. A single free lookup per shortlisted term is enough to rank order.
Four mistakes that break the board
None of these are tool problems. They are sequencing problems, and they show up in every review where the shortlist looks busy but the content plan stalls.
- Sorting by volume first. Big numbers with no clear intent produce traffic that never converts.
- One page per variant. "Cost", "pricing", and "price" are usually one page, not three.
- Research without a decision. A sheet with three hundred rows and no owner is an archive, not a plan.
- Ignoring the page you already have. Refreshing an existing page often beats publishing a fourth version of it.
Where We Take Over
If you want the board run properly, with intent labels, difficulty calls, and page assignments you can hand to a writer, our Keyword Research Services cover the whole pass: seed discovery, expansion, filtering, and a shortlist mapped to URLs. If you already have the shortlist and need pages written against it, SEO Content Writing Services turn each keyword card into a brief, a draft, and a page that matches the intent you approved.
The board is deliberately low-tech. What makes it work is that every column ends with a decision, so research produces a plan instead of a spreadsheet. Most teams we audit are not short of keywords. They are short of a filter and a date.
Frequently Asked Questions
How many keywords should I research at once?
Twenty to forty terms per quarter is enough for most small and mid-sized sites. More than that and the writing queue becomes the bottleneck, which is the same as having no plan.
Should I start with head terms or long tail terms?
Long tail first. They reveal intent faster, convert sooner, and tell you which head terms your site can realistically contest later. Head terms are a destination, not a starting point.
How accurate do search volumes need to be?
Accurate enough to rank order, not to forecast. Volume bands are directionally useful; single-number precision is not, especially for small markets and B2B niches.
What if two keywords mean the same thing?
They are one page. Group variants under the phrasing that matches the dominant intent, then mention the alternatives naturally in the copy instead of building near-duplicate pages.
How often should the board be refreshed?
Quarterly for most businesses, monthly if you sell in a fast-moving category. Between passes, watch Search Console for queries you rank for without ever targeting them.
Do I need keyword research if I already have a content calendar?
Yes, but lighter. Audit the existing calendar against intent first; half the time the plan is sound and the page types are simply mismatched to what searchers want.