SaaS keyword research, ICP-first and SERP-validated
Written by Olayinka Olayokun·Published ·Updated ·Verified
Keyword research is the discipline of discovering the queries an ICP actually uses, classifying them by intent and difficulty, and sequencing them into a publishing roadmap a site can realistically rank for.
Summary and key takeaways
SaaS keyword research starts from the five problems your ICP googles at 11pm, not from a tool's autosuggest. On Invoicemonk, a 12-query BOFU list built from inbound emails — average tool volume ~30/mo — produced the page that converted a day-28 trial-to-paid signup, while the high-volume 'invoicing software' page Semrush recommended ranked nowhere.
- •Start from 5–20 problems your ICP googles in their own words — sales emails and demo questions are the cheapest source.
- •Validate every seed against the live SERP — PAA, related searches, and who currently ranks tell you more than KD scores.
- •Filter difficulty against your real DR, not aspirational ones — a DR <15 SaaS site competes for KD 0–20, not KD 40.
- •Group queries into 3–6 clusters before sequencing — isolated posts in topics you don't otherwise cover rarely rank.
- •Treat zero-volume long-tail B2B queries as winnable — most keyword tools systematically undercount them.
In plain English ·SaaS keyword research is the practice of building a query list from real ICP language, validating it against the actual SERP, and filtering by what a young SaaS domain can plausibly rank for. Tool volume is a sanity check, not the input — most winnable B2B queries are undercounted or zero-volume in Ahrefs/Semrush.
- KDI 14
- Keyword Difficulty for 'saas keyword research' — realistic for a DR 20+ site with a topical map Semrush
How this compares
| Approach | Where seeds come from | Filter | Typical first-page lag |
|---|---|---|---|
| Tool-first (Ahrefs/Semrush autosuggest) | Volume slider on head terms | KD against aspirational DR | 12+ months, often never |
| AI-first (ChatGPT brainstorm) | Model's training data | None — volumes are hallucinated | Inconsistent; AI picks already-saturated SERPs |
| ICP-first (this playbook) | Sales emails, demo questions, SERP PAA | Real DR + live SERP composition | 3–6 months for the first cluster |
Keyword research is the discipline of discovering the queries an ICP actually uses, classifying them by intent and difficulty, and sequencing them into a publishing roadmap a site can realistically rank for.
Most founder-led SEO fails because the keyword list was built from a generic tool, not from the ICP. The result is a content calendar full of high-volume head terms that never convert and never rank. The fix is to invert the order: start from the customer, validate with the tool — never the other way around.
What this guide covers: seed queries, intent classification, difficulty filter, topical grouping, publishing sequence.
Start from the ICP, not the tool
Write the five problems your ICP googles in their own words. Not what they say in customer interviews — what they would actually type into Google at 11pm trying to fix a workflow that's broken. Those are your seed queries.
Write the five problems your ICP googles in their own words. Not what they say in customer interviews — what they would actually type into Google at 11pm trying to fix a workflow that's broken. Those are your seed queries.
Pull the People Also Ask block and the 'related searches' from the SERP for each seed. That's your first 30 query modifiers, drawn directly from real searcher behaviour rather than a tool's extrapolation.
If you have a sales team, read the last 50 inbound emails and demo questions. Every recurring phrasing is a seed. Founders consistently underestimate how often sales language matches searcher language word-for-word.
Filter by intent
Sort every query into informational ('how does X work'), commercial ('best X for Y'), or transactional ('X pricing', 'X login', 'X vs competitor'). A SaaS site needs all three, mapped to TOFU, MOFU, and BOFU pages.
Sort every query into informational ('how does X work'), commercial ('best X for Y'), or transactional ('X pricing', 'X login', 'X vs competitor'). A SaaS site needs all three, mapped to TOFU, MOFU, and BOFU pages.
Discard 'best X' queries you can't legitimately appear in. Roundups are won by publishers with backlinks, not by vendors. Target 'X vs Y', 'alternatives to X', 'X for [vertical]' instead — those are vendor-winnable.
Be honest about which queries are commercial vs transactional. A 'free invoicing software' query is commercial; a tutorial page won't satisfy it.
Score by ranking difficulty you can realistically beat
Use your actual domain rating, not Ahrefs' aspirational KD slider. A DR 8 SaaS site competes for KD 0–15 queries, not KD 40. Targeting KD 60 because the volume looks tempting is how you spend nine months on a page that never breaks page three.
Use your actual domain rating, not Ahrefs' aspirational KD slider. A DR 8 SaaS site competes for KD 0–15 queries, not KD 40. Targeting KD 60 because the volume looks tempting is how you spend nine months on a page that never breaks page three.
Look at who's ranking on page one, not at the difficulty score. If the top 10 are all DR 70+ publishers, your DR 8 site won't crack it. If page one has a few DR 20 sites mixed in, the query is winnable with one good page and three internal links.
Build a 90-day publishing roadmap
Group queries by topical cluster, not by traffic potential. Publishing a single high-volume post in a topic you don't otherwise cover produces nothing; publishing five medium-volume posts that complete a sub-topic produces ranking for all five plus the head term.
Group queries by topical cluster, not by traffic potential. Publishing a single high-volume post in a topic you don't otherwise cover produces nothing; publishing five medium-volume posts that complete a sub-topic produces ranking for all five plus the head term.
Sequence within each cluster: the pillar page first, then 3–5 clusters inside two weeks. Publishing the cluster while Google is still re-crawling the pillar is the cheapest way to earn topical authority on a young site.
Tool stack that's enough — and what's overkill
You need: Google Search Console (free, primary truth source), one paid tool with a keyword database (Mangools ~$30/mo for early stage, Semrush ~$140/mo or Ahrefs ~$129/mo for a maintained topical map), and the SERP itself.
You need: Google Search Console (free, primary truth source), one paid tool with a keyword database (Mangools ~$30/mo for early stage, Semrush ~$140/mo or Ahrefs ~$129/mo for a maintained topical map), and the SERP itself.
You don't need: a content optimiser, a SERP scraper, a competitive intelligence dashboard, or three keyword tools cross-referenced. Founders consistently overspend on tooling and underspend on judgement.
The checklist for saas keyword research: an icp-first playbook
- ✓20 seed queries written from ICP problems, in the customer's exact words
- ✓Each seed expanded with the SERP's People Also Ask and related searches
- ✓Every query labelled informational / commercial / transactional and mapped to TOFU/MOFU/BOFU
- ✓Difficulty band set to KD 0–20 for DR <15, KD 0–35 for DR 15–30
- ✓Page-one composition checked manually — at least one peer-sized site in the top 10 before targeting
- ✓Roundup-class queries discarded if a vendor page cannot legitimately appear
- ✓Queries grouped into 3–6 topical clusters with a pillar query identified for each
- ✓90-day publishing order that ships every pillar before its clusters
- ✓Google Search Console verified and impressions baseline captured for the 8 BOFU queries
5 chapters inside saas keyword research: an icp-first playbook
Each chapter covers one narrow piece of the guide. Read in order, or pick the one closest to your bottleneck.
Where saas keyword research: an icp-first playbook sits in the system
every other discipline — topical authority, on-page, internal linking all start from the query list.
general keyword research — SaaS prioritises intent fit over volume.
Search Console impressions and clicks once pages ship — never with tool estimates alone.
the topical map to define the central entity, sub-topics, and supporting pages. Read the topical authority for saas (without 100 articles) guide →
the site's domain rating and topical maturity — KD 40 queries are not winnable from a DR 8 site.
Quick answers about saas keyword research: an icp-first playbook
- How is SaaS keyword research different from general keyword research?
- It optimises for intent fit over volume. A 'best invoicing software' query has 5,400/mo volume but is owned by review publishers; a 'how to invoice a US client from Nigeria' query has '0' tool volume and converts a trial signup in week one. SaaS keyword research targets the second by default.#
- How do I do keyword research for a SaaS product launch?
- Before the product is live, you have no Search Console data — so seed queries have to come from customer interviews and competitor SERPs. Pull the top 10 ranking pages for your category, extract the entities and questions they cover, then narrow to the 8–12 BOFU queries that name your ICP and use-case specifically.#
- Can I use generative AI for SaaS keyword research?
- For expansion yes, for selection no. ChatGPT or Claude is useful for clustering and generating question variations from a seed list, but it hallucinates volumes and difficulty. Always re-validate AI-suggested queries against Semrush/Ahrefs and the live SERP before adding them to the roadmap.#
- Which keyword research tool should a SaaS founder pay for?
- One, not three. Mangools (~$30/mo) is enough for pre-seed and seed founders; Semrush (~$140/mo) or Ahrefs (~$129/mo) makes sense once you're maintaining a 30+ page topical map. Google Search Console is always the primary truth source — start there before you pay for anything.#
Questions about saas keyword research: an icp-first playbook
- For BOFU pages, 20 monthly searches from your ICP is plenty — those searchers are pre-qualified. For TOFU clusters, aim for 200+ total monthly searches across the cluster, not per page.
- Google's guidance to write content for people, not for keyword tools. Google Search Central — Helpful content
- Google Keyword Planner buckets volumes and is skewed toward advertised terms. Google Ads Help — Keyword Planner
Reading the guide is the start. To see which fixes inside it your specific URLs need, get a founder-grade SEO audit of your URLs. Same six disciplines, applied to the pages you actually own.
Next guides in the playbook
Keyword research is upstream of every other discipline. A topical map drawn from the wrong queries produces nothing; on-page work on the wrong page produces nothing. Get the query list right first, and the rest of the playbook has something real to operate on.
Want this applied to your URLs?
The audit reads your site against the full playbook and returns the prioritised fix list.
Get your free audit →