Methodology

Every public-company page computes CAC as a two-step pipeline against SEC EDGAR. There is no manual entry, no guesswork, no third-party API in between.

1. Numerator (S&M expense). Pulled from the SEC EDGAR XBRL company-facts API. Tagged us-gaap:SellingAndMarketingExpensefor every company in our cohort except Paycom, which reports only the combined SG&A line - that exception is flagged on the page.
2. Denominator (net-new customers). Customer counts are not in XBRL. They appear in MD&A narrative - "As of January 31, 2026, we had 13,328 total customers, increasing from 10,996 customers as of January 31, 2025." A Claude-driven structured-extraction prompt parses every 10-K and 10-Q for these disclosures, recording the exact sentence verbatim. The disclosure quote is shown on every page.
3. Computation.Annual CAC = FY S&M ÷ (current-year customers − prior-year customers). Trailing-twelve-month CAC = rolling 4-quarter S&M ÷ same denominator. We report both annual and TTM where the data supports them.
4. Multiple slices per company. Many companies disclose more than one customer-count slice - total customers, $100k+ ARR customers, $1M+ TTM customers, Atlas customers (MongoDB), Forbes Global 2000 customers (Snowflake). We render every slice the filing breaks out, each with its own CAC.

Industry × stage × ACV benchmark cells on our benchmarks pages only publish a number when a named primary source publishes one. The primary sources we draw from:

• First Page Sage SaaS CAC Payback Benchmarks (2024): 28-industry × 4-customer-size CAC payback table. We use this for 9 of our 12 industries: 6 direct mappings (fintech, ecommerce, edtech, healthtech, security, services) and 3 proxy mappings clearly labelled on the cell pages (devtools via Engineering, saas-b2c via Entertainment, vertical-saas via Construction).
• Optifai B2B SaaS CAC Payback Benchmark (n=939, April 2026): segmented by ACV band (SMB / Mid-Market / Enterprise) but published as ranges rather than point estimates. We use this for our saas-b2b industry. Cell pages document that the displayed median is the upper bound of Optifai's range; the best-in-class value is the lower bound.
• First Page Sage 2025 CAC by Channel: per-channel CAC dollars for B2B SaaS (paid search, LinkedIn, SEO, outbound). We cite these on the by-channel page with direct links per number.

Sources we plan to add but haven't yet: SaaS Capital's annual benchmark survey, ChartMogul's quarterly SaaS benchmarks, Bessemer's State of the Cloud public-company composites, OpenView (now retired but reports still cited), and KeyBanc Capital Markets' SaaS Survey. These reports require PDF parsing or paid access that we haven't worked through yet. Marketplaces and AI-native cells stay as stubs until cite-able primary data surfaces.

Each sourced cell shows the report name, the table or section reference, and the publication date. Cells are re-verified when the underlying reports refresh.

The benchmark grid has 180 cells (12 industries × 5 stages × 3 ACV bands). The named primary sources above don't cover all 180 - published data is concentrated in a few canonical cuts (SaaS B2B by ARR band, mostly). For the cells they don't cover, we do not triangulate, average, or impute a number to fill the gap.

Instead each unsourced cell renders a page that:

• States explicitly that no primary source covers this cut
• Links to the four nearest adjacent cells in the grid
• Links to live public-company CAC for any company in our cohort in that industry
• Explains the segment definition (what we mean by "Series B mid-market healthtech")

The page exists at its canonical URL so the long-tail search query has a destination, but it doesn't fabricate a number to fill it. This is the load-bearing principle of the publication: a missing number, surfaced honestly, is more useful than a confident guess.