Key takeaways
- Volume is the least useful column in your export. Intent and winnability decide what gets published.
- Classify intent by reading the SERP, not by pattern-matching the phrase.
- Positions 8–30 in your own Search Console data are usually the cheapest wins available.
- One keyword, one page: map before you write, or you will cannibalise later.
- Re-score quarterly — SERP intent shifts more often than most teams check.
Start with the data you already own
Before touching a paid tool, open Google Search Console and filter to queries where your average position is between 8 and 30 with at least a few hundred impressions over three months.
Those queries are proof: Google already considers your site a plausible answer. Moving from position 12 to position 5 typically multiplies clicks by four or five, and usually requires editing an existing page rather than creating a new one.
Three filters worth running:
- High impressions, low CTR, decent position — a title and meta description problem, fixable in an afternoon.
- Position 8–20, rising impressions — a content depth problem; the page is close and needs expansion.
- Queries with no matching page — genuine content gaps where you rank accidentally on a tangential page.
Harvest broadly, then stop
Expand from every source — competitor exports, autocomplete, People Also Ask, forums, review sites, your own site search, sales call transcripts — until new sources stop producing new ideas. That saturation point usually arrives faster than people expect, at a few thousand rows.
Two habits improve this stage enormously. First, harvest the questions, not just the noun phrases; questions carry intent on their face. Second, record where each keyword came from. A term that appeared in three sales calls deserves different treatment from one that appeared only in a tool export.
Classify intent from the live SERP
Open the results page for every candidate you are serious about. Incognito, correct country, no personalisation. Then record what you see:
- Dominant content type in the top ten (guide, listicle, product page, tool, video, forum)
- SERP features present (featured snippet, People Also Ask, shopping, local pack, video carousel)
- Format signals — average length, whether pages are dated, whether originals or aggregators rank
- Whether any result is weak enough to displace
Two rules come out of this every time. First, match the format the SERP rewards or accept that you will not rank. Second, if the SERP is split — half listicles, half product pages — there are two intents and you may be able to serve the underserved one.
A client insisted on targeting a term with 12,000 monthly searches. The top ten were entirely software directories. We published the guide anyway, as a test. Eight months later it had never passed position 30. The lesson cost us the argument and them a quarter.
Score winnability, not difficulty
Difficulty scores measure backlinks. Winnability measures whether you can win the term this year. Score each candidate from 1 to 5 on four factors and take the average:
| Factor | Score 5 | Score 1 |
|---|---|---|
| SERP strength | Weak pages, forums, outdated posts in the top ten | All major publishers with hundreds of referring domains |
| Format match | We can produce exactly the format that ranks | The SERP demands a tool or dataset we do not have |
| Topical authority | We have or plan a full cluster on this topic | Completely new territory for the domain |
| Business value | Buyers search this before purchasing | Interesting but unrelated to what we sell |
Publish only 4s and 5s in year one. Keep the 3s for when your cluster is established, and be honest that 1s and 2s may never be worth it.
Map one keyword to one page
Before anything is written, build a mapping table: keyword, intent, winnability, target URL (existing or new), cluster, funnel stage, CTA, status.
The mapping table is the artefact that prevents cannibalisation, duplicated commissioning and the awkward quarterly discovery that two writers covered the same topic from opposite ends of the backlog. It is also the fastest way to onboard a new writer: they can see what exists, what is planned and where their piece fits.
When two keywords keep fighting for the same page, that is the mapping telling you to merge. Merge early; it is cheaper than a redirect chain later.
Prioritise by expected value, not by volume
Rank the shortlist by a rough expected-value calculation rather than raw volume:
expected clicks = volume × realistic CTR at target position × probability of reaching it
Then weight by commercial value. A term with 200 searches a month at 30% CTR with a high chance of ranking and direct purchase intent beats a 10,000-search head term you have a 10% chance of cracking and that brings students and job seekers.
Sequence the resulting list with dependencies in mind — commercial pages before the blog, supporting articles before pillars, refreshes before new pieces.
Re-score every quarter
Search results move. Intent shifts — a term that returned guides last year may return product pages today, usually because a category matured. Competitors publish. Your own authority changes.
Once a quarter, re-check the SERPs for your top twenty target terms, re-score winnability, promote anything that became reachable, and retire targets whose intent has moved away from what you sell. It takes half a day and routinely reveals two or three opportunities that did not exist three months ago.
If you would rather not maintain this yourself, it is exactly what our keyword research service delivers — including the quarterly re-score.
Frequently asked questions
Search intent is the goal behind a query: to learn something, to compare options, to buy, or to reach a specific site. It is determined by looking at what currently ranks, because the results page is Google's own statement about what it believes searchers want.
Yes, and more than ever for mid-authority domains. Long-tail queries have clearer intent, less competition and higher conversion rates. They also aggregate: fifty long-tail terms in one cluster often exceed the traffic of the head term.
Look at who ranks, not at the difficulty score. If every result is a major publisher or a site with hundreds of referring domains to that page, and none of them is weak, the term is out of reach for now. If there is a forum thread or a 2019 blog post in the top five, there is an opening.
One primary keyword and as many close variants as share the same SERP — often ten to forty. If two terms return substantially different results, they need different pages.
Google Search Console (free and irreplaceable), one paid tool for volume and backlink data, and a browser in incognito mode for SERP review. Everything beyond that is convenience.
Silvia Moreno
Co-founder & Content Strategist
Silvia spent eight years in newsrooms before moving into SEO. She builds the keyword maps and content strategies behind every Witty Content Writers engagement.
