B2B keyword research when most terms show 10 to 50 searches a month: five sources to mine, a scored workbook, revenue sizing and one list for SEO and paid.
Open a keyword tool with a B2B term list and every line shows a dash or a 20. Nothing is wrong with your list. That is what a B2B keyword universe looks like in a tool built to price advertising auctions. Most of the terms that turn into a demo request or a proposal show 10 to 50 searches a month, and a filter set at 100 removes them before anyone has read them. The useful answer is to change the filter, not the list: score clusters on intent, winnability and deal value, with volume as a weighting rather than a gate.
In our own Search Console this year, a growing share of the queries that reach us are full sentences an assistant typed on a buyer’s behalf: which agencies do this, in this sector, with published numbers. Every one of them shows a dash in Semrush. They are also the queries that turn into calls.
Below: why volume filters hide most of the universe, where the low volume keywords that matter sit, the five sources I mine first, how the workbook scores everything, how to size a cluster in revenue, and how one list serves SEO and paid. We made the pipeline argument in SEO for B2B; this is the keyword layer underneath it.
Why does filtering by search volume hide most of a B2B keyword universe?
Because the number you are filtering on was built to describe an advertising auction, not a B2B buyer.
Just so you know, for reference: the volume in every keyword tool starts life as Google Ads data. It is a twelve-month rolling average, rounded into buckets, with close variants merged into one line. In plain English, a term showing 20 a month might be 5 in August and 60 in the fortnight your buyer is shortlisting, and “20” is also hiding six spellings of the same question. The tools model on top of that; I covered the sources in the international keyword research piece. The number is an estimate of an average, not a count of buyers.
Now apply a filter at 100. On a typical B2B universe of 400 to 800 queries, that removes 70% to 85% of the lines. What survives is the head terms your whole category is fighting over and the definitional queries students type. What goes in the bin is “does X integrate with Sage” and “X vs Y for mid-market”. The filter keeps the traffic and throws away the pipeline.
If half your existing posts came from a volume-sorted export, that is no reflection on whoever built them: volume is the column every tool sorts on by default, and most B2B blogs were built in that order. Those posts still have a use. The fix is rarely to refresh them. Leave them live, fold the useful ones into a cluster, and put the new effort into the pages never written because the number next to them looked too small.
Where does B2B search demand actually hide?
Underneath the head term, in phrasings that describe a situation rather than a product. Five patterns, on every account.
Problem phrasings. Nobody searches “field service management software” on day one. They search “engineers double booked” or “job sheets still on paper”. Ten to 40 searches each, and the page that answers the problem is the first page in the sales cycle.
“Vs” and “alternative” queries. Typed at shortlist stage by the person who will sign. Rarely above 50 a month, almost always the highest enquiry rate on the site once built, and the pages many B2B companies hold back from writing, understandably, because they name a rival.
Integration and compliance questions. “Does X work with HubSpot”, “X Cyber Essentials”, “X SOC 2”. These come from IT, security and procurement, the people who can’t say yes but can say no. Ten searches a month, and a missing page here stalls a deal marketing already counted.
Job title plus problem. “Finance director cash flow forecasting tool”, “practice manager patient recall system”. The role tells you who is reading and what they are paid to fix. Only that role types it, which is exactly why the page converts.
The questions assistants ask on a buyer’s behalf. A buyer asks ChatGPT or Gemini for a shortlist, the assistant runs long, specific searches to build it, and those hit your site as impressions with query strings twelve words long. No tool records them. Your Search Console does. If the page that answers them does not exist, you are not in the shortlist, and you never find out.
None of these will pass a volume filter. The wider logic of intent over volume is in our commercial intent keyword strategy piece.
Which five sources should you mine before you open a keyword tool?
Your own business, in this order. The keyword tool comes last because it can only expand what you already know to look for.
1. Sales call transcripts
Pull the last 30 recorded discovery calls and read the first ten minutes, where the prospect describes the problem in their own words. Write down the nouns and the complaints. You will get 40 to 60 phrasings the tools have never heard of.
2. CRM lost-deal reasons
Every lost reason is a page. “Went with [Competitor]” is a vs page. “No integration with our finance system” is an integration page. “Couldn’t get security sign-off” is a compliance page. If your CRM does not record lost reasons as a required field, fix that before any content; it is a small change.
3. Support tickets
Tickets describe what customers struggle with after they buy, which is what prospects worry about before they buy. The top five themes are usually five objection pages.
4. Search Console queries with impressions and no clicks
Export sixteen months, filter to queries with 50 or more impressions and a click-through rate under 1%. Those are searches Google already thinks you are relevant for, on a page that does not answer them. The assistant-driven queries surface here, and it is the cheapest win in the method because the demand is already arriving.
It is also where the biggest single find I have seen came from. A client, a sensible team who knew their home market well, had several thousand monthly searches for their service with “near me” attached, sitting in their own Search Console, and not a single page for it. The data had never been put in front of them, so it had never come up in a planning meeting. Once it had, we built the pages out in priority order, and the same logic runs down to the smallest term on the sheet. As I put it on the call: even if there’s no search volume, at least we’re getting visibility for it, we’ve got landing pages for it.
5. Competitor pages that rank for nothing but convert
List three competitors’ commercial pages, not their blogs. The ones with no measurable traffic in any tool are the interesting ones: a page stays live for three years with no visible traffic because it closes deals. Copy the intent, never the page.
Only then do I open Semrush, to expand each phrasing into its variants. By then the universe is 400 to 800 queries and I know what every one is for.
How do you cluster and score a B2B keyword universe?
Cluster by the page that would answer the query, then score the cluster, never the keyword. The unit of work is a page.
One row per cluster, four columns. Cluster volume, the sum of every query in it, which can be zero. Intent weight: bottom of funnel 3, middle 2, top 1. Winnability, from reading the actual SERP rather than a difficulty score: 3 if the top results are thin or a directory; 2 if they are competitor pages you can out-answer; 1 if the first page is Gartner, a government body and two national brands. Pages needed: how many URLs you would build or rebuild to own the cluster.
The score is intent multiplied by winnability, divided by pages needed. Then everything sorts into three priorities.
- P1: bottom-of-funnel clusters with a winnable SERP: commercial pages, vs pages, pricing, integrations. First 90 days, nothing else until these are live.
- P2: middle-of-funnel clusters that feed a P1 page: problem guides, comparison explainers, “how to choose” content. Months three to nine, one cluster at a time. Get the content up first and come back to optimise.
- P3: the zero-volume pages you build anyway. Compliance answers, niche integrations, the assistant-shaped questions. No tool will ever credit them with demand, and they keep you in the shortlist. Visibility without volume still counts.
I will say this about P3, because it is the part people push back on, and fairly, since it asks for effort with no number beside it. A page nobody searches for but every buyer checks is a sales document with a URL, and it costs a day to write.
Worked example (illustrative, rounded numbers)
A UK scheduling software company, £18,000 a year per deal, two candidate clusters.
Cluster A is “what is field service management”: 1,900 searches a month, intent 1, winnability 1 because Salesforce, IBM and Wikipedia hold the first page, one page needed. Score: 1 times 1 over 1, 1.0.
Cluster B is “[Competitor] alternative UK” plus “[Competitor] vs [Competitor]” plus “[Competitor] pricing” plus “does [Competitor] integrate with Sage”: 140 searches a month in total, intent 3, winnability 2 because the SERP is a G2 listing and two thin competitor pages, four pages needed. Score: 3 times 2 over 4, 1.5.
Cluster B, at 7% of the volume, goes first. On a volume-sorted sheet it would not make the first hundred rows.
How do you size a cluster in revenue rather than sessions?
Run the cluster volume through the click share for the position you can realistically hold, then your own enquiry and close rates, then deal value. Four numbers you can defend.
Take Cluster B. 140 searches, a top-three position taking around 30% of clicks, so 42 visits a month. Bottom-of-funnel pages on B2B sites we run convert to enquiry at 3% to 6%; call it 5%, so two enquiries a month. A 25% close rate makes that half a customer a month, which at £18,000 is roughly £108,000 a year from four pages. Cluster A, sized the same way at a 0.5% enquiry rate and a 10% close rate because the readers are students, produces a quarter of a customer a month. Eleven times the traffic, half the revenue.
Those numbers sit next to every P1 cluster in the workbook, and they are what the board reads. To be fair, they are estimates and I say so on the sheet. But an estimate built from your own close rate is a plan. A session forecast is a hope, and hope is not a strategy.
One more consequence. A B2B site usually needs 30 to 60 pages built properly, not 300. If a 5,000-line keyword list lands on your desk, it is usually a tool export that has not been clustered yet, a normal first step rather than a finished plan. The question that moves it on is which 40 pages it maps to.
How does one keyword universe serve both SEO and paid?
Same workbook, one extra column: which channel owns the cluster this quarter. Ali and I are stubborn on this. Two teams with two lists are bidding against each other with the same finance director’s money.
Where a P1 cluster has winnability 1 and would take twelve months to rank, paid does the heavy lifting now and the organic page is built anyway for the hand-over later. Where organic sits in the top three, paid pulls back and spends elsewhere. Where a P3 page exists for a zero-volume query, paid runs it as a low-bid exact match, because the click is cheap and the page is the best answer. On accounts we inherit, the first pass usually moves 10% to 15% of paid budget off terms organic already owns. The longer argument is in SEO vs PPC.
Paid also gives a fast read on intent. Put £300 behind a ten-search-a-month phrase for a month and you learn whether it converts before anyone spends a week writing for it. The cheapest keyword research there is.
Questions we get asked
What is a good search volume for a B2B keyword?
There isn’t one. Ten searches a month from the person who signs is a good keyword; 5,000 from students is not. Judge the cluster on intent, winnability and deal value, and treat volume as a weighting inside that, never a gate.
Should we target keywords with zero search volume?
Yes, when a buyer or a blocker would type it at shortlist stage: integrations, compliance, “vs” and “alternative”. They sit in P3 on our workbook, cost a day each, and keep you in shortlists you would never otherwise see.
How many keywords does a B2B site need?
A universe of 400 to 800 queries, clustered into 30 to 60 pages, is typical for a mid-market B2B company with one core service line. Around 30 of those queries usually carry most of the qualified pipeline.
How do you find B2B keywords?
Start inside the business: sales call transcripts, CRM lost-deal reasons, support tickets, Search Console queries with impressions but no clicks, and competitors’ commercial pages. Then expand each phrasing in a keyword tool to catch variants and whatever volume exists.
How is B2B keyword research different from B2C?
B2C research sorts by volume, because one searcher is one buyer. B2B research sorts by intent and deal value, because six people search on one deal and most type something a tool has never counted. Fewer pages, targeted harder, scored on revenue.
What is B2B SEO?
Being found and trusted in search by companies buying from other companies, over a long sales cycle with several people involved. Keyword research decides which pages exist; the pipeline side is in SEO for B2B.
Where we fit
We build and score keyword universes for UK B2B companies whose buyers search in tiny numbers: SaaS, professional services, IT and industrial firms where the deal value makes a 20-search-a-month query worth a page. It is the first step of any B2B SEO strategy we build, sits inside our SEO strategy work and runs through our B2B and SaaS SEO service.
The proof point I use is Myagi: 38% more qualified organic leads and 115% pipeline growth from a universe built on intent rather than volume. If you want to see your own, get in touch and we’ll tell you what we need from your CRM and Search Console.
