Open model data · CC BY 4.0

GTM World Model Dataset

The canonical model, its research evidence and its decision boundaries.

The dataset contains the full current GTM World Model. Download its equations, theses, evidence register, qualification protocol and cluster roadmap under CC BY 4.0. The vocabulary is a separate download.

Dataset CC BY 4.0 v3.3 Model updated 2026-10-05

Benchmark orientation examples

The historical benchmark examples below orient a subscription-business discussion. They are not a current representative sample or deployment thresholds. Use the metrics reference for definitions, source dates and limitations; define the target segment and measurement basis before comparing.

Metric	Seed	Series A	Series B	Best-in-class
MRR Growth (MoM %) directional	15–25%	10–20%	7–15%	>20% (post-PMF)
NRR (Net Revenue Retention) established	90–100%	100–110%	105–115%	>130% (Snowflake FY2022)
CAC Payback (months) established	18–30	15–24	12–20	<12 months
Magic Number established	0.5–0.8	0.6–1.0	0.7–1.2	>1.0
Rule of 40 directional	N/A (pre-scale)	20–40	30–50	>60 (top-quartile public)
Burn Multiple directional	<2×	<1.5×	<1×	<0.5×
LTV:CAC Ratio directional	2–4×	3–5×	4–6×	>5×

Confidence ratings: established = derived from multiple independent sources with consistent results; directional = single-source or limited sample; contested = conflicting data across sources.

Methodology

Source types

The benchmark reference draws on investor, vendor, operator and company reports. Source class does not by itself establish independence, causality or applicability:

Investor and vendor benchmark reports: Bessemer Venture Partners State of the Cloud 2024; SaaS Capital Annual Survey (1,000+ respondents); KeyBanc Capital Markets SaaS Survey (350+ private companies); OpenView SaaS Benchmarks 2024.
Tier 2 — Public company filings: S-1, 10-K, and 10-Q filings from Snowflake, HubSpot, Salesforce, Datadog, Veeva, and GitLab. Used for best-in-class benchmarks and regime-segmented analysis (ZIRP vs. post-ZIRP).
Tier 3 — Aggregated operator surveys: SaaStr Annual attendee surveys, RevOps Squared benchmark reports, Pavilion/GTM Alliance member surveys. Higher variance; treated as directional.
Tier 4 — Single-company disclosures: Investor presentations, blog posts with disclosed metrics, conference presentations. Used only where Tier 1–3 data is unavailable; marked contested.

Confidence ratings

Every metric in the dataset carries one of three confidence ratings:

established — Consistent across 3+ independent sources; low methodological variance.
directional — 1–2 sources, or high variance across sources; useful as orientation not target.
contested — Conflicting data or limited sample; include with caution.

Regime segmentation (ZIRP vs. post-ZIRP)

The dataset distinguishes the ZIRP era (2019–2021, characterized by hyper-growth at any cost) from the post-ZIRP era (2022–present, characterized by efficiency-first growth). Benchmarks for CAC payback, burn multiple, and S&M efficiency changed materially between regimes. Default benchmarks reflect the post-ZIRP era unless otherwise stated.

Limitations

Benchmarks depend on their sample, definitions and era. The MRR walk (MRR_t = MRR_{t-1} + New + Expansion − Contraction − Churned) reconciles a consistently defined recurring-revenue stock. It is not recognized revenue or cash flow; the stock equation requires explicit additional categories for reactivation, currency and other adjustments when applicable.

Downloads

All data is published under CC BY 4.0. You may use, redistribute, and build on this data freely — with attribution. The current model and vocabulary are available as JSON:

gtm_world_model.json — Full model, evidence and qualification vocabulary.json — GTM vocabulary (219 terms with definitions and confidence ratings)

Research and qualification: The current model includes evidence_register, qualification_protocol, cluster_roadmap and proposed_studies. Evidence records state what was studied and where transfer to a commercial decision remains unverified.

REST API access: GET /api/gtm/terms (219 terms) · GET /api/gtm/theses (32 theses) · GET /api/gtm/model (full world model). Full API documentation →

How to cite

@misc{shalvi_gtm_benchmark_dataset_2026,
  author    = {Singh, Shalvi},
  title     = {GTM World Model Dataset},
  year      = {2026},
  version   = {3.3},
  url       = {https://shalvisingh.com/gtm/dataset},
  license   = {CC BY 4.0},
  note      = {Canonical model, scoped evidence register and decision qualification protocol.}
}

APA: Singh, S. (2026). GTM World Model Dataset (Version 3.3) [Dataset]. shalvisingh.com. https://shalvisingh.com/gtm/dataset. CC BY 4.0.

This dataset carries a Schema.org Dataset JSON-LD record. Structured-data consumers (Google Dataset Search, OpenAlex, etc.) can discover it automatically.