How to Run pSEO Keyword Research: From Modifier Map to Demand Forecast

What is programmatic SEO keyword research?

Programmatic SEO keyword research is the process of identifying a repeatable keyword pattern (head term + modifier dimensions), estimating search demand across every combination, and forecasting revenue at the row level before a single page is built. It treats keywords as a matrix, not a list.

The difference from traditional keyword research is scale and structure. Traditional research finds 50 keywords for 50 articles. pSEO research finds 2 dimensions that combine into 5,000 URLs from one template. Zapier built 7,000+ integration pages on a single [App A] + [App B] pattern and pulls 6.3M monthly organic visits from it. Wise runs 8.5M currency converter pages on [Currency A] to [Currency B].

The research process answers four questions before you build:

• Which modifier dimensions actually have search demand?
• How much volume does the full matrix represent?
• What's the intent mix (info vs commercial vs transactional)?
• What's the projected revenue if you capture 30% of that demand?

If you can't answer those four questions in a spreadsheet, you don't have a pSEO project. You have a content gamble.

What's a modifier in programmatic SEO?

A modifier is a variable that combines with a head term to create a long-tail query. The head term is the broad, high-volume root ("CRM software"); the modifier is the qualifier that narrows intent ("for real estate agents", "with QuickBooks integration", "under $50/month").

Modifiers fall into named dimensions. Each dimension is a column in your matrix:

Dimension type	Example head term	Example modifiers
Industry / vertical	expense management software	for SaaS, for restaurants, for agencies
Use case	project management tool	for remote teams, for client work, for sprints
Integration	CRM	with Slack, with Gmail, with HubSpot
Location	coworking space	in Austin, in Berlin, in Lisbon
Role	OKR examples	for engineers, for marketers, for sales
Price tier	accounting software	free, under $20, enterprise
Comparison	Notion vs	Asana, ClickUp, Coda

A single head term with 2 dimensions of 50 modifiers each produces 2,500 URLs. With 3 dimensions of 30 modifiers, you get 27,000. The skill is picking dimensions whose values have independent search demand -- meaning the long-tail query exists in the wild, not just on paper. We validate that in step 3.

How do I find keywords for programmatic SEO? (The 5-step method)

Programmatic SEO keyword research follows five sequential steps. Skipping one breaks the forecast: you can't price a page you haven't classified, and you can't classify a keyword you haven't found.

1. Head-term clustering. Group your business's revenue-generating queries into URL patterns.
2. Modifier dimension extraction. Mine SERPs, autocomplete, and competitor sitemaps to find the dimensions that drive demand.
3. Volume estimation. Bulk-pull search volume for every cell in the matrix via DataForSEO and validate the top 10% against Search Console.
4. Intent classification. Tag each row as informational, commercial, or transactional and weight CVR accordingly.
5. Demand forecast. Multiply volume x CTR x CVR x ACV per row to get a revenue projection that decides which clusters to build.

The rest of this article walks each step with the tools, the inputs, and the output you should have in your spreadsheet by the end.

Step 1: How do you cluster head terms for pSEO?

Head-term clustering means grouping queries by URL pattern, not topic. Each cluster becomes one page template. If two queries can't share a template, they belong in different clusters.

Start with 5-15 candidate head terms drawn from your product's revenue surfaces: features, integrations, use cases, customer segments. For each candidate, validate three criteria from Findable's pSEO playbook:

• Business relevance. Does someone searching this convert on your product?
• Baseline volume. Does the head term itself have at least 100 monthly searches as a sanity check?
• Combinatorial fit. Can the term combine with at least 50 modifiers in a single dimension?

Reject any head term that fails on combinatorial fit -- those become regular blog posts, not templates.

Output: A shortlist of 3-5 head terms, each mapped to one URL pattern. Example for a B2B SaaS expense tool:

• /expense-management-for-[industry]
• /expense-software-vs-[competitor]
• /expense-policy-template-[role]

Each pattern is one row in your project plan. We're going to fill the modifier matrix for one of them in step 2.

Step 2: How do you extract modifier dimensions from SERPs and competitor sitemaps?

Modifier dimension extraction pulls real modifiers from three sources: Google's autocomplete, the People Also Ask box, and competitor sitemaps. You don't brainstorm modifiers. You scrape them.

Source 1: Google autocomplete A-Z scrape. Run your head term + each letter of the alphabet through autocomplete (expense management software a, expense management software b...). Tools like Keyword Tool or a Python script with googleautosuggest automate this. Cluster the results -- you'll see dimensions emerge (industry, integration, price).

Source 2: People Also Ask + Related Searches. Both surface real long-tail variants. Pull them with DataForSEO's SERP API or a Screaming Frog SERP scrape.

Source 3: Competitor sitemap mining. This is the highest-yield source. Find a competitor already running pSEO on your head term, pull their sitemap.xml, and extract modifiers from URL slugs. For Zapier-style integration pages, a single sitemap can yield 7,000+ modifier values. Screaming Frog crawls sitemaps in minutes.

Cluster modifiers into dimensions. If you collect 800 raw modifiers, group them by type. "For SaaS", "for ecommerce", "for agencies" are one dimension (industry). "With Slack", "with Gmail" are another (integration). Stop at 2-3 dimensions per template -- adding a fourth multiplies pages but rarely multiplies demand.

Output: A modifier matrix, one tab per head term, columns = dimensions, rows = modifier values. This is the input to step 3.

Step 3: How do I estimate traffic for thousands of unranked pages?

You estimate pSEO traffic by pulling search volume for every cell in your modifier matrix, then validating the top decile against your own Search Console data. The cheapest path is a bulk volume API. Manual lookups don't scale past 100 keywords.

Bulk volume via DataForSEO. DataForSEO's Bulk Clickstream Search Volume endpoint returns volume for up to 1,000 keywords per request at $0.0001 per keyword + $0.01 per task. A 25,000-row matrix costs $2.75. Ahrefs Keywords Explorer and Semrush Keyword Magic Tool also offer bulk lookups but with tighter row caps and higher per-keyword cost.

Generate the cartesian product first. In Sheets, use =ARRAYFORMULA(A2:A & " " & B2:B) or a Python script to combine head term + modifier dimensions into the full keyword list. Feed that list to the API.

Validate with Search Console + clickstream. Volume APIs return Google Ads-style estimates that bucket into ranges (10, 50, 100, 500). For the top 10% of rows, cross-check actual impressions in your Search Console for similar queries. If your existing pages on the head term get 4x the API's estimated volume, apply that multiplier to the matrix. Clickstream-based volume tends to be more granular than Google Ads buckets.

Expect a long tail. Successful pSEO projects often target keywords with 10-100 monthly searches each, banking on volume across thousands of variations. Don't reject low-volume rows -- the matrix sums into traffic.

Output: Your matrix with a monthly_volume column populated for every row.

Step 4: How do you classify search intent for pSEO keywords?

Intent classification tags every row in your matrix as informational, commercial, or transactional. The tag drives your CVR assumption in step 5 -- a transactional query converts 5-10x harder than an informational one, so mixing them in one forecast inflates the projection.

The four intents (Semrush's framework):

• Informational. "What is expense management." User wants to learn. Low CVR, high top-of-funnel value.
• Navigational. "Expensify login." User wants a specific brand. Skip these for pSEO.
• Commercial. "Best expense software for SaaS." User is comparing options. High CVR, prime pSEO target.
• Transactional. "Buy expense management software." User is ready to act. Highest CVR, lowest volume.

Classification at scale. Three options, in order of cost:

1. Modifier-based rules. Tag rows by modifier prefix: best [x] = commercial, what is [x] = informational, buy [x] = transactional. Hits ~80% accuracy for free.
2. SERP-feature classification. Pull SERP features via DataForSEO's SERP API -- shopping results signal transactional, knowledge panels signal informational, comparison snippets signal commercial.
3. LLM batch classification. Run rows through GPT-4 or Claude with a prompt that returns one of the four labels. Costs ~$1 per 1,000 rows and hits >95% accuracy.

Anti-pattern. Don't build pages for navigational queries (you won't outrank the brand) or for purely informational queries with zero buying signal. Concentrate the build budget on commercial and transactional rows.

Output: An intent column on every row. You're ready to forecast.

How do I forecast revenue from a pSEO project?

A pSEO revenue forecast is a row-level calculation: monthly_volume x position_CTR x conversion_rate x ACV = projected_monthly_revenue. Sum the column, divide by build cost, and you have a payback period for the cluster. The forecast lives in the same sheet as the keyword matrix.

The four inputs:

• Position CTR. Pick a realistic ranking position. With AI Overviews now appearing on roughly a third of searches, First Page Sage's 2026 CTR data shows position 1 at 39.8% on clean SERPs but 10-12% when AI Overviews are present. Use a conservative blend: 8% for position 5, 15% for position 3, 22% for position 1.
• Conversion rate. Tier by intent. Informational: 0.3-0.5%. Commercial: 1.5-3%. Transactional: 3-6%. Use your existing Google Ads data on the same query family to calibrate.
• Average contract value (ACV). Your average revenue per customer for the segment.
• Capture rate. What share of total demand do you realistically capture? Year 1: 10-20%. Year 2: 30-40%. Apply this as a multiplier on the volume.

Worked example. A B2B expense SaaS with $1,200 ACV, targeting expense management for [industry] with 80 industry modifiers. The matrix sums to 24,000 monthly searches. At a year-1 capture rate of 15% (3,600 sessions), 2% commercial CVR, and $1,200 ACV, projected MRR is 3,600 x 0.02 x $1,200 = $86,400/month -- before accounting for retention or expansion. If the build costs $30K, payback is under two weeks of ranked traffic.

The kill threshold. Reject any cluster with less than 3x projected revenue per dollar of build cost in year 1. Anything below that is a dressed-up content marketing project, not a pSEO investment.

Output: A single cell -- projected MRR for the cluster -- that justifies (or kills) the build.

What tools do I need for pSEO keyword research?

You need four tools to run the full 5-step methodology. Everything else is optional.

Tool	Step	What it does	Cost
DataForSEO Bulk Search Volume	3	Pulls volume for 1,000 keywords per request	$0.0001/keyword
Screaming Frog	2	Crawls competitor sitemaps to extract modifiers	Free up to 500 URLs, £199/yr unlimited
Ahrefs or Semrush	1, 2	Head-term volume, SERP overlap, competitor keyword export	$129+/mo
Google Sheets or Airtable	1-5	Holds the matrix and runs the forecast formulas	Free / $20/mo

Optional but high-leverage:

• Google Search Console -- free, ground-truth volume validation for any query you already rank for
• OpenAI API or Anthropic API -- batch intent classification, ~$1 per 1,000 rows
• Keyword Insights -- automated keyword clustering if your matrix exceeds 50K rows

The entire stack runs under $300 for a one-off project of 50K keywords. Any pSEO consultant quoting you $20K for keyword research alone is selling tool markup.

Worked example: pSEO keyword research for a B2B expense SaaS

Here's the methodology applied end-to-end on one head term.

Head term: expense management software URL pattern: /expense-management-software-for-[industry]

Step 1 -- Cluster. Head term passes all three criteria: business-relevant, 14,800 monthly volume on the root, fits a one-dimension industry modifier matrix.

Step 2 -- Modifier extraction. Scraped Expensify, Ramp, and Brex sitemaps via Screaming Frog. Extracted 73 industry modifiers after dedup. Cross-referenced against Google autocomplete -- added 9 more (e.g., "for nonprofits", "for law firms", "for ecommerce"). Final dimension: 82 industries.

Step 3 -- Volume estimation. Pushed all 82 keywords through DataForSEO Bulk endpoint. Total cost: $0.0182. Distribution: 4 rows >1,000/mo, 18 rows 100-1,000, 41 rows 10-100, 19 rows below 10. Matrix sums to 9,420 monthly searches.

Step 4 -- Intent classification. Ran modifier-based rules: 100% commercial intent (every row is [product] for [industry], a classic comparison pattern). No navigational or pure-informational rows to drop.

Step 5 -- Forecast. Inputs: $1,200 ACV, 2% CVR (commercial), 15% capture rate year 1, blended 12% CTR (mid-page-1 position).

• 9,420 searches x 12% CTR = 1,130 sessions/month at full capture
• 1,130 x 15% capture = 170 sessions/month year 1
• 170 x 2% CVR = 3.4 customers/month
• 3.4 x $1,200 ACV = $4,080 net new MRR/month from one cluster

At $15K to build the template + 82 pages, payback is ~3.7 months. Build it.

This is the entire calculation. Five steps, one spreadsheet, one go/no-go decision.

What's the spreadsheet template for pSEO keyword research?

The template glues all five steps into one workbook. Five tabs, one per step:

1. head_terms -- Candidate head terms, business relevance score, baseline volume, combinatorial fit (Y/N), URL pattern.
2. modifier_matrix -- One column per dimension, one row per modifier value. Dropdown menus enforce dimension boundaries.
3. keyword_universe -- Generated cartesian product (head term + modifiers). Auto-fills from tab 2 via ARRAYFORMULA. Columns: keyword, head_term, modifier_1, modifier_2, monthly_volume.
4. intent_classification -- Adds an intent column via rule-based VLOOKUP or external LLM API call. Includes assumed CVR per intent tier.
5. forecast -- Pivots tab 4 into projected revenue per cluster. Inputs (ACV, capture rate, CTR, build cost) live in a single named-range block at the top so you can sensitivity-test.

The template enforces the methodology -- you can't fill tab 5 without populating tabs 1-4 in order. That's the point. pSEO projects fail when teams skip to volume estimation before validating the head term.

Grab the template at the link below and clone it into your Drive. Replace the example head term with yours. The forecast cell at the top of tab 5 is your go/no-go number.

Step	Action	Tool	Output
1. Head-term clustering	Group revenue queries by URL pattern	Ahrefs / Semrush + Sheets	3-5 head terms with URL patterns
2. Modifier extraction	Scrape SERPs, autocomplete, competitor sitemaps	Screaming Frog + autocomplete scraper	Modifier matrix (2-3 dimensions)
3. Volume estimation	Bulk-pull volume; validate via GSC	DataForSEO Bulk endpoint ($0.0001/kw)	Volume column on every row
4. Intent classification	Tag each row info / commercial / transactional	Modifier rules + LLM batch	Intent column with CVR tier
5. Demand forecast	Volume x CTR x CVR x ACV per row	Sheets with named-range inputs	Projected MRR + payback period