On this page
Lists of keyword tools compare features. What decides a keyword plan is the order you use them in, and knowing which question each one can answer. This post sets out the workflow our SEO team runs, tool by tool, then works it through a published engagement: a UK B2B SaaS platform that grew organic traffic 127% in five months.
Which tool answers which question
No single tool does keyword research. Each answers one question well and others badly, and most bad plans come from asking a tool a question it cannot answer. This is how we divide the work.
| Tool | The question it answers | What it cannot tell you |
|---|---|---|
| Google Search Console | Which queries your site already appears for, with clicks, impressions and average position | Demand for queries you do not yet appear for |
| GA4 | Which landing pages turn visits into signups, leads or sales | Which query brought each visitor |
| Google Keyword Planner | Roughly how often a term is searched, from Google’s own data | Exact numbers: accounts without active campaigns often see broad ranges |
| A third-party database, such as Ahrefs or Semrush | Related terms, an estimated difficulty, and who ranks now | True volume: these are modelled estimates, and two tools rarely agree |
| Google Trends | Whether interest is rising, falling or seasonal | Volume: Trends shows relative interest on a 0 to 100 scale |
| The results page itself | What format wins, and whether an AI answer sits above the links | Anything on its own: it is one market and one moment |
One rule follows from the table. We never choose a target on a third-party volume estimate alone. Estimates are good for comparing terms with each other and poor for forecasting traffic.
The workflow, step by step
Before any tool is opened, we ask for read access to Search Console and GA4 and an hour with whoever knows which queries bring in money. The first two show what is already working; the third is where the plan’s priorities come from.
- Start from your own data. In Search Console, list the queries where you already have impressions, and in GA4 the landing pages that convert. Near-wins and converting pages come first.
- Collect buyer questions. From the sales conversation, write down the questions buyers ask before they choose, in their own words.
- Expand and size. Run those questions through Keyword Planner and a third-party database for related terms and rough demand. Check Trends for seasonality.
- Read the results page. For each candidate, look at who ranks, in what format, and whether an AI answer already resolves it. Mark what the site can realistically win.
- Cluster into a topical map. Group terms that one page can answer, so two pages never compete for one query. Sequence the clusters by commercial value and effort.
- Brief each page. Every brief names the target query, the question the page must answer in its opening lines, and the internal links it needs.
- Measure on the business line. Report rankings, sessions and signups together, and agree up front when each will show movement.
Worked example: a SaaS platform with no organic pipeline
A UK B2B SaaS platform came to us directly with nearly every qualified lead arriving through paid or outbound. The site had thin content, no coherent topical structure and inconsistent naming, and the product did not appear in classic results or in AI answers.
A keyword tool, asked for opportunities, would have pointed at the highest-volume category terms. We rejected that. For a site with no authority in the category, those terms were held by sites that had it, and the effort would have produced pages stuck far down the results.
We built a topical map from the questions buyers ask, clustered into topics the site could credibly own, and sequenced them by commercial value.
We also chose fewer, deeper pages over a faster calendar, each opening with a direct answer an engine could quote, alongside schema and entity cleanup. Reporting ran weekly in Prism, from rankings to sessions to signups, with AI citations tracked beside classic positions.
Organic traffic grew 127% in five months, with citations in Google AI Overviews on commercial queries. That figure belongs to the whole programme, not to keyword choice alone. The full record is in the SaaS case study, and the wider approach is on our SaaS SEO page.
Where keyword research goes wrong
Three mistakes are worth checking your own plan for:
- Volume first. A high-volume informational term can bring thousands of visits that never buy. Our post on the keyword that is costing you money shows how to spot one in your own analytics.
- One tool’s numbers treated as fact. Difficulty scores and volumes are each vendor’s model. Use them to rank options, not to promise traffic.
- A keyword list instead of a map. A list produces one page per term, and pages that compete with each other. A map assigns each cluster to one page.
When keyword research is not the bottleneck
Research is the wrong place to spend when:
- You need leads this month. Organic work is slow, and paid search is the honest answer to an urgent gap.
- Your category is new enough that nobody searches for it yet. Search follows demand; it does not create it.
- Pages are not indexed, or the site has technical faults. Fix those first, or new pages will not rank whatever their keywords.
- Nobody can spare time to review content for accuracy. Pages written to a perfect keyword brief without subject review are the thin kind this method avoids.
