Demand
Does the phrase appear in at least two independent places, such as autocomplete and a competitor heading?
Pass: two sources. Fail: only a tool's estimate.Four columns, one verdict per query
What Is Keyword Research? A Board Of Real Queries And Four Checks
Keyword research is the process of collecting the exact phrases people type into search, judging each one on intent, competition and business value, and turning the survivors into a list of pages to build. It is not a volume report. It is a decision record: this query deserves a page, this one deserves a section, this one deserves nothing. The work is finished when every row has a verdict and a destination, not when the export has forty thousand rows.
- 4judgment columns
- 1verdict per query
- 0pages without a reason
What does keyword research actually produce?
It produces a page list with a reason attached to every line. Each row records the real question, the intent behind it, the strength of the pages already ranking, and a verdict: build a page, add a section, or skip it. The output is a build order you can hand to a writer, not a spreadsheet of traffic estimates.
The board
Five queries, read across four columns
A board is easier to defend than a stack of numbers, because every verdict is visible next to the evidence. Here is a working board for a freight and container business. Read a row left to right and you can see why each query got the answer it did.
Real questionIntentCompetitionWorth it
"how much does a 40ft container cost"
Commercial investigation · comparing options
Medium · freight blogs, no rate tables
Page · cost guide
"40ft container internal dimensions"
Informational · needs a spec
Low · thin pages, dated tables
Section + table
"container shipping company china"
Commercial · building a shortlist
High · marketplaces and directories
Only with proof
"what is a demurrage charge"
Informational · avoiding a cost
Low · glossary definitions
Section in glossary
"buy used 20ft container"
Transactional · ready to act
High · local yards with stock
No · stock page only
Verdicts are decisions, not destinations. A "section" verdict still needs a URL and an owner.
Column by column
The four judgments, and what each one is for
1. The real question
Record what people type, not what your category is called internally. The phrase in the box is short, misspelled, lowercase and often a full question. If your board says "industrial filtration solutions" and the query says "what filter fits a 3 inch pipe", the board is already wrong. Pull real wording from autocomplete, related searches, your support inbox and the questions sales hears twice a week.
2. Intent
Intent decides the page type before it decides anything else. Someone comparing costs wants a table and a range. Someone who has decided wants stock, a lead time and a way to order. Someone learning wants a definition and a next step. Put the same paragraph in front of all three and two of them leave.
3. Competition
Open the results page and read the top ten as a buyer. Are they answering the question, or only mentioning it? Is the information current, specific and verifiable? Competition is not a single score from a tool. It is a judgement about whether your page can be meaningfully better within a reasonable build time.
4. Whether it is worth it
The last column is commercial, not technical. Could a visitor who lands on this page become a customer in the next ninety days? If the answer is no for every plausible version of that visitor, the query is a distraction, however easy it looks to rank.
Classification
Four intent types and the page each one needs
Most arguments about a keyword are really arguments about intent. Sort the query into one of these four, and the page type, the call to action, and the success metric follow without further debate.
| Intent | How it sounds | What they want next | Page type | Primary action |
|---|---|---|---|---|
| Informational | what, why, how, can, does | To understand a term or process | Guide, glossary section | Read the related page |
| Commercial | best, vs, cost, supplier, alternatives | To compare before choosing | Comparison, service page | Request a quote |
| Transactional | buy, order, price list, in stock, near me | To act now | Product, category, contact | Add to cart or call |
| Navigational | brand names, login, tracking | To reach a known place | Home, account, tracking page | Complete the task |
Two warnings. First, one query can carry mixed intent; pick the dominant one and let the page acknowledge the other in a section. Second, intent can shift by country and season, so date the verdict and revisit it.
Thresholds
Pass or fail: the rules we apply
Thresholds turn opinion into a repeatable decision. Ours are deliberately simple, because a rule nobody remembers is not a rule.
Intent
Can you name the next action a reader should take after this page?
Pass: one clear next step. Fail: "learn more" is the only answer.Competition
Can you list three specific things your page will do better than the current top three?
Pass: three concrete gaps. Fail: "ours will be nicer".Value
Could a visitor from this query become a customer without leaving your niche?
Pass: plausible buyer path. Fail: curiosity traffic only.Mistakes
Five habits that break a query board
- Exporting first, thinking second.A tool answers the question you typed into it. If the seed list came from a category page and nothing else, the board is a mirror, not a map.
- Sorting by volume on day one.High volume with vague intent produces visits that never convert. Sort by intent, then by whether you can win, then by volume.
- Never opening the results page.A keyword you cannot beat is a keyword you should not build for. Read the top ten before you brief a writer.
- One page per variant.Plural, singular and two spellings of the same term belong on one page. Splitting them creates pages that compete with each other.
- Filing the board and never reopening it.Intent shifts, competitor pages get rewritten, and last year's easy win becomes this year's wall. Date every verdict and review the board twice a year.
Want your queries judged and ranked?
Board to build order
How the four columns become pages
A finished board collapses into four buckets. Queries marked "page" get a brief each. Queries marked "section" attach to an existing URL with a named subheading. Queries marked "later" wait for a page that does not exist yet. Queries marked "no" are closed with a reason, so nobody reopens the argument in three months.
Then count. A site publishing four pages a month should carry no more than twelve approved page verdicts in the queue at any time. A queue of eighty is not a plan, it is a wish list that guarantees the last twenty rows get written badly.
A short example
A container trader used the board for one product line. Of ninety collected queries, eleven passed all four columns: four cost and sizing guides, three port-specific pages, two comparison pages and two glossary sections. Two months later the sizing table was the most cited page on the site because it was the only one with internal dimensions in both metric and imperial. Nothing about that win required a bigger budget. It required a board that said "section" out loud instead of building a thin page for every question.
If you want the definition that sits above this work, read Whats Is SEO?. To have the whole cycle run for you, from collection to verdicts, see Keyword Research Services.
FAQ
Questions about keyword research
How many keywords should end up on the board?
Collect a few hundred, approve a few dozen. The board is a filter, not an archive. If every row passes, the thresholds are too soft.
Do I need a paid tool?
Not to start. Autocomplete, related searches, competitor headings and your own support inbox will fill the first pass. A paid database helps with estimates, not judgement.
What if two queries seem identical?
Check the results page. If the same pages rank for both, they are one query and one page. If the results differ, the intent differs and you probably need two.
How long does a board take?
Two weeks for a focused market, including the results-page review. Most of the time goes into reading what already ranks, not into collecting more rows.
Is keyword research still useful with AI search?
More so. Answer engines quote pages that answer a specific question clearly. The board tells you which question to answer, and in which words.
Who should own the verdicts?
One person with commercial authority, not a committee. Boards fail when every row needs a meeting, because the verdicts arrive after the quarter has already been planned.