Case study: Gumlet turned ChatGPT mentions into 20% of inbound revenue. Read it →
B2B SaaS keyword research is the work of ranking buyer queries by the revenue they can produce, not by the traffic they can attract. DerivateX scores every cluster on six factors: buyer stage, deal value, conversion likelihood, product fit, ranking feasibility and AI-search overlap. In DerivateX’s experience, fewer than 20 clusters carry the majority of qualified pipeline at a $5M to $50M ARR company.
Revenue-weighted B2B SaaS keyword research is a prioritization method that scores each keyword cluster by the pipeline it can realistically produce, then builds the content plan in that order. DerivateX uses it because the standard alternative, sorting a keyword export by volume and difficulty, systematically pushes the highest-value commercial queries to the bottom of the list. Those queries have low volume by definition. A term searched 90 times a month by people comparing two vendors is worth more to a software company than a term searched 9,000 times a month by students, job seekers and competitors.
The unit of work is the cluster, not the keyword. A cluster is a group of queries that one page can answer completely. If two pages target the same cluster, you have created an internal competition problem that no amount of link building fixes, so the map carries a hard rule: one cluster, one owning page, named in the sheet before the brief is written.
The output is not a keyword list. It is a spreadsheet where every row carries a target page URL, a buyer stage, an estimated annual pipeline contribution, a feasibility score and an owner. DerivateX reviews that sheet monthly and re-scores it quarterly, because feasibility changes when a competitor publishes and conversion economics change when pricing changes.
It fails because the model assumes a stable relationship between rankings and revenue, and that relationship has moved. When Google AI Overviews appear, click-through rate drops by 61%, and 40% of Google queries now show them. A page can hold position 3 and lose half its sessions without anything going wrong on the page. DerivateX sees this pattern on almost every audit of a $5M to $50M ARR account: rankings flat or improving, sessions down, demo requests down more. This is a structural change in how results are delivered, not a sign that anyone’s SEO work was sloppy.
There is a second failure, older and more common. Traffic-first research produces a content calendar full of top-of-funnel guides because those clusters have volume. The guides rank, the sessions arrive, and the sales team reports that none of the leads know what the product does. DerivateX has taken over accounts with 400 published posts and 11 pages that a buyer would ever read while holding a credit card.
The fix is not fewer posts. It is a different sort order: score by revenue potential first, then let volume break ties inside a score band.
Start with the pages you already own, then the language your buyers already use, then the tool exports. DerivateX runs these three steps in that order deliberately, because starting in a keyword tool anchors the whole plan to whatever that tool happened to index.
List every page that a buyer could convert on: pricing, product pages, integration pages, comparison pages, alternatives pages, use case pages, and any high-performing bottom-of-funnel post. For each one, record current sessions, current demo requests, and the query it currently ranks for. Roughly half of what you will end up doing is strengthening pages that already exist, and that half is faster to ship and faster to show in pipeline.
Read 20 recorded sales calls, the last 50 support tickets, your win and loss notes, and the subreddits where your category argues. You are looking for the exact nouns and the exact objections. Software firms consistently name their product by a feature word while their buyers name it by an outcome word, and the tool export cannot tell you which one wins. This step also matters for AI search, since 46.7% of Perplexity’s top sources come from Reddit, meaning the phrasing used in community threads is the phrasing that gets retrieved.
Assign each cluster to one of five stages and lock the page type to the stage. DerivateX uses this mapping because it stops the most common planning error, which is answering a bottom-of-funnel query with a blog post.
| Buyer stage | Example query shape | Page type that should own it | Primary conversion event |
|---|---|---|---|
| Unaware of category | “how to reduce onboarding time” | Guide with a product-led section | Email capture or tool use |
| Problem aware | “why do estimates take so long” | Guide or teardown | Newsletter, template download |
| Solution aware | “best construction estimating software” | Listicle or category page you own | Demo request |
| Vendor aware | “[competitor] alternatives”, “X vs Y” | Comparison and alternatives page | Demo request or trial |
| Ready to buy | “[category] pricing”, “[product] integration with [tool]” | Pricing, integration, use case page | Trial signup or sales conversation |
Grow and Convert’s bottom-of-funnel content guide names comparison and alternatives keywords as one of its five buying-intent content types, alongside category keywords, side category keywords, category keywords with specificity, and jobs to be done keywords. That matches what DerivateX sees in client analytics, with one caveat worth stating: comparison pages convert well but they cap out. There are only so many competitors to compare against, which is why the map needs the other four stages to keep growing after the comparison set is built.
You weight it by multiplying realistic traffic against the conversion economics of the segment that query attracts, then discounting for how hard the ranking is and how much of the answer AI search will absorb. DerivateX runs these three steps as a single scoring pass.
Take the formula seriously and fill it in per cluster, not per site:
Estimated annual pipeline = monthly searches x expected CTR at target position x visit-to-demo rate x demo-to-opportunity rate x ACV x 12
Two of those inputs vary wildly by cluster. Visit-to-demo rate on a comparison page is often ten to thirty times the rate on a top-of-funnel guide. ACV varies because different queries attract different company sizes, and a query containing “for enterprise” brings in a different buyer than the same query containing “free”. Your pricing model shifts the weighting too: with usage-based or expansion-heavy pricing, queries used by teams already at scale outrank queries used by hobbyists, while with flat monthly subscription tiers, the volume of qualified signups carries more weight and a mid-volume commercial cluster can outperform a small enterprise-flavored one.
Domain difficulty scores from tools are a proxy, and a weak one for SaaS. Look at the actual result page instead and answer three questions. Who owns the top five, and are they vendors or publishers? Is the dominant format a listicle you cannot appear in as the subject? Has the top result been stable for 18 months? A cluster where three review sites and two funded competitors have held position for two years is not a 12-month target for a company at $8M ARR. DerivateX scores those clusters low on feasibility and routes them to a different play, usually placement inside the third-party listicles rather than a head-on page.
For each cluster, record whether it triggers Google AI Overviews, and whether the equivalent buyer prompt in ChatGPT or Perplexity returns a vendor recommendation. These are separate columns because they behave differently. Only 11% of domains are cited by both ChatGPT and Perplexity, and 80% of URLs cited by ChatGPT and Perplexity are not in Google’s top 100. A cluster can be an AI opportunity and a Google dead end at the same time, and a DerivateX map says which.
Keyword.com positions itself as a rank tracker built for both traditional SERP monitoring and AI search, with competitor tracking, citation analysis and brand sentiment analysis across AI search results. Its pricing page lists the 360° Visibility plan from $26 per month billed annually. If your team needs the visibility data in a dashboard and has the strategy handled in house, buying a tracker is the right call and hiring an agency is not.
You convert the scored sheet into a build order, assign one owning page per cluster, and set up measurement that reports demos rather than sessions. DerivateX ships these three steps together, because a scored map that never becomes a publishing queue is just an expensive audit.
Rate each factor 0 to 5, multiply by its weight, and sum. Product fit is a gate, not a factor: if the honest rating is 0, the cluster leaves the map even if everything else scores high. Ranking for a query your product does not solve produces traffic your sales team will resent.
| Factor | Weight | What a 5 looks like | What a 1 looks like |
|---|---|---|---|
| Buyer stage proximity | 3 | Query names a vendor, a price or an integration | Query is a definition someone will read once |
| Product fit | 3 (gate) | Your product is a genuine top-three answer | You would have to stretch to mention yourself |
| Conversion likelihood | 3 | Comparable page already converts above 3% | No comparable page has ever converted |
| Deal value fit | 2 | Query language matches your best-fit segment | Query attracts buyers below your floor price |
| Ranking feasibility | 2 | Vendor pages rank, top five has churned recently | Stable review-site duopoly for two years |
| AI-search overlap | 2 | Prompt returns vendor recommendations today | No AI surface answers this query at all |
Maximum score is 75. In practice, anything above 52 goes into the current quarter, 35 to 52 goes into the backlog with a named trigger, and below 35 gets archived rather than deleted, since feasibility scores move.
For each cluster in the build queue, decide one of three things: refresh an existing page, build a new one, or consolidate two competing pages into one. Refreshes come first, because pages that already have history and links reach a target position faster than new pages, so a quarter that starts with eight refreshes and four new builds usually shows pipeline movement sooner than a quarter of twelve new posts.
Internal linking is part of the map, not a cleanup task afterwards. Every supporting page in a cluster links to the owning money page with anchor text that names the cluster, and every money page links out to two or three supporting pages. This is what makes a group of posts behave as one asset instead of nine competing ones. Programmatic pages belong in the map only where the template can carry a real answer, which is why DerivateX caps programmatic builds at the number of pages where a human could defend each one to a buyer.
Build the forecast from the same per-cluster math you scored with, and publish the assumptions next to the number. A forecast with visible assumptions survives a bad month; a single traffic projection does not. If you want a starting structure for that model, the SaaS SEO revenue projection calculator lays out the inputs in the same order.
Then instrument the reporting. Sessions and positions go in an appendix, while the main report shows demo requests by cluster, opportunities by cluster, and attributed revenue by owning page. DerivateX builds this view during onboarding rather than at the first quarterly review, because retrofitting attribution to work already published is where most reporting projects die. The mechanics of tying organic work to closed revenue are covered in more depth in the B2B SaaS SEO pipeline guide.
Below is an illustrative slice of a map for a hypothetical B2B SaaS company at $12M ARR with a $9,000 ACV. Every figure in it is a modelled output from the formula in step 4, invented for this example. None of it comes from a published dataset, a client account or a measured benchmark.
| Cluster (illustrative) | Stage | Modelled monthly searches | Modelled visit-to-demo | Score /75 | Modelled annual pipeline |
|---|---|---|---|---|---|
| “[competitor] alternatives” | Vendor aware | 320 | 4.5% | 66 | $155,000 |
| “[category] pricing” | Ready to buy | 210 | 3.8% | 61 | $86,000 |
| “[category] software for [vertical]” | Solution aware | 1,100 | 1.6% | 58 | $190,000 |
| “[product] + [tool] integration” | Ready to buy | 90 | 6.0% | 54 | $58,000 |
| “how to [core job to be done]” | Problem aware | 8,900 | 0.2% | 31 | $46,000 |
Look at the last two rows. The problem-aware cluster has almost 100 times the modelled search volume of the integration cluster and produces less modelled pipeline. A volume-sorted plan would build the big guide first and reach the integration page in month nine, if ever.
Five mistakes account for most of the failures DerivateX sees when auditing existing programs at software companies, and none of them are about keyword selection skill.
It changes the unit of research from keywords to prompts, and the unit of success from position to citation. DerivateX runs both tracks off one map, because the same buyer moves between Google and an assistant inside a single evening. With 200M+ weekly ChatGPT Search users and 3 to 4 brands typically recommended per AI category query, the practical question for a software firm is whether it holds one of those three or four slots for the prompts its buyers actually type.
The overlap between the two surfaces is thinner than most teams assume. 28% of ChatGPT-cited pages have zero organic Google visibility, which means the page that gets you recommended may be one that would never have made a traffic-based content plan. Visitors arriving from AI-sourced discovery also convert at 4.4x higher rates, which is why those clusters deserve a weight in the score rather than a separate project.
A Citation Surface Map is the artifact DerivateX builds alongside the keyword map. It lists, for each priority cluster, the specific third-party pages that assistants currently cite when answering the matching prompt: the listicles, the review profiles, the community threads and the documentation. That map tells you where corroboration has to exist before your own page can be recommended. Citation Engineering is the methodology DerivateX uses to make language models recommend a brand deliberately, by building and corroborating those sources rather than by rewriting a page and hoping.
The results this produces look different from ranking results. For REsimpli, DerivateX’s work made them the most cited and recommended real estate CRM for investors in ChatGPT within 90 days. For Gumlet, more than 20% of monthly inbound revenue is attributed to AI discovery. For Verito, DerivateX moved an average position of 40 on Google to first page, and got them cited and recommended on ChatGPT and Google AI Overviews for 40 of their commercial hosting queries. That last one is the pattern most keyword maps should aim for, since the same commercial cluster paid off on both surfaces. The mechanics are laid out in the B2B SaaS AI search visibility guide.
Measure it on demo requests per owning page, opportunities created, and pipeline attributed to the cluster, reviewed at 90-day and 180-day marks. DerivateX reports on that set and keeps rankings and sessions as diagnostic inputs underneath, because rankings explain why the pipeline number moved but they are not the number anyone approves budget against.
| Timeframe | What should be true | What to report |
|---|---|---|
| Days 0 to 30 | Map scored, owning pages assigned, attribution wired | Cluster count by score band, baseline demos per page |
| Days 30 to 90 | Refreshed pages moving, first new commercial pages live | Position movement on priority clusters, first demo lift |
| Days 90 to 180 | Commercial clusters holding, AI citations appearing | Demos and opportunities by cluster, citation counts by prompt |
| Days 180 to 365 | Compounding: refreshed pages plus new ones both contributing | Attributed pipeline and revenue by owning page |
On cost, DerivateX publishes its numbers. The entry engagement, Rank and Get Found, is $5,000 retainer plus $1,000 to $1,200 off-site budget, so $6,000 to $6,200 all in, run as a 90-day pilot with no lock-in after. Own Your Category is $8,000 retainer plus $1,500 to $2,000 off-site, so $9,500 to $10,000 all in. Market Leader is $12,000 retainer plus $2,000 to $3,000 off-site, so $14,000 to $15,000 all in on a six-month minimum. There is also a $3,500 one-time diagnostic delivered in two weeks that credits in full against month one if you convert within 30 days. Full details sit on the DerivateX pricing page. These are targets and process commitments, not guarantees of rankings or revenue.
Here is where DerivateX is the wrong answer. If what you need is publishing throughput against a strategy you already trust, agencies that position themselves around high-volume editorial production, such as Animalz and Siege Media, are built for that and DerivateX is not: a scored map deliberately caps the queue at the clusters that can carry pipeline, which is fewer pages than a content-volume program will ship. If your ACV is under $2,000 and your sales motion is fully self-serve, a $6,000 monthly program is hard to justify against paid acquisition, and you would be better served buying a tracker and hiring one strong in-house writer. If your main problem is that your pages get read but nobody signs up, the constraint is conversion, not discovery, and a testing platform such as Convert, whose pricing page lists Growth from $399 per month billed monthly, will move your number faster than we will. And if you have a capable in-house SEO team that simply needs the AI-search layer, DerivateX’s engagement model for SEO teams is a better fit than a full retainer.
The best approach to B2B SaaS keyword research is to score clusters by pipeline potential before volume, then build in that order, assigning exactly one owning page per cluster and reporting on demos rather than sessions. DerivateX runs that sequence on every account: inventory the money pages, pull buyer language from calls and community threads, map queries to buyer stages, score on the six factors, then refresh before you build new.
The reason it works is ownership. Rankings are rented, and the rent went up in 2025 when Google AI Overviews expanded and answers started appearing above the results. Clusters are owned. When one page holds a commercial cluster on Google, gets cited by assistants for the matching prompt, and carries corroboration on the third-party pages those assistants read, it keeps producing demos through algorithm changes and interface changes alike. Software companies that reorder their map around pipeline usually find they need fewer pages than they thought and better ones than they have. If you want the same logic applied to your build order by an outside team, the DerivateX SaaS SEO service runs it as a scored, quarterly-reviewed engagement.
Inventory your existing money pages, pull buyer language from sales calls and community threads, group queries into clusters, then score each cluster on buyer stage, deal value, conversion likelihood, product fit, ranking feasibility and AI-search overlap. Build in score order, assign one owning page per cluster, and report on demos rather than sessions.
Fewer than most teams expect. In DerivateX’s experience, a $5M to $50M ARR company has 40 to 80 clusters worth mapping, and fewer than 20 of them carry the majority of qualified pipeline. Thousands of individual keywords roll up into those clusters, but the build queue is measured in clusters and owning pages.
BOFU keywords are queries a buyer types when they already understand the category and are choosing a vendor: competitor alternatives, X versus Y comparisons, category pricing, integration queries and free trial searches. They have low volume and high conversion. Grow and Convert’s bottom-of-funnel guide names comparison and alternatives keywords as one of five buying-intent content types.
Refreshed pages with existing authority often move within 30 to 90 days. New commercial pages typically take 90 to 180 days to hold position and produce steady demos. AI citations can appear faster on some prompts, as with REsimpli becoming the most cited real estate CRM for investors in ChatGPT within 90 days.
DerivateX runs it inside its retainers, which start at $6,000 to $6,200 all in per month for the 90-day Rank and Get Found pilot. A standalone scoped version is the $3,500 diagnostic, delivered in two weeks, credited in full against month one if you convert within 30 days.
If you want to see which buyer prompts your category already answers without you, DerivateX runs a free AI visibility audit with a 48-hour turnaround that returns your current citation position across ChatGPT, Perplexity and Google AI Overviews: request the free AI visibility audit.
VP, SEO & AI Search at DerivateX. We're a B2B SaaS SEO and Generative Engine Optimization agency that engineers AI citations in ChatGPT, Perplexity, Claude, and Gemini and connects them to demo bookings and revenue pipeline.
Apoorv Sharma is the co-founder of DerivateX, a B2B SaaS SEO and Generative Engine Optimization agency that engineers AI citations in ChatGPT, Perplexity, Claude, and Gemini and connects them to demo bookings and revenue pipeline. He is the author of the 2026 AI Visibility Benchmark Report and the Citation Engineering methodology. He's also the brain behind "Found On AI" and has sold 2 of his companies previously
