GASP (Generally Accepted SaaS Principles) is an open standard for SaaS metrics: definitions, formulas, relationships and a digital twin that AI agents can read instead of guessing. The formulas compile to SQL for Postgres, Snowflake, BigQuery and Databricks. The gasp-standard-mcp package serves it to AI agents and the same standard is published as Apache Ossie (incubating) files.

How many SaaS metrics does GASP define?

GASP defines 313 metrics across 13 departments including Core, Sales, Marketing, Customer Success, Support, Onboarding, Product, Engineering, Finance, RevOps, People, Partnerships and Professional Services.

What is the difference between GRR and NRR?

GRR (Gross Revenue Retention) measures revenue retained from existing customers excluding expansion. It can only be 100% or lower. NRR (Net Revenue Retention) includes expansion revenue, so it can exceed 100%. Best-in-class SaaS companies target GRR above 90% and NRR above 110%.

How is the Rule of 40 calculated?

The Rule of 40 equals ARR Growth Rate (Year-over-Year) plus EBITDA Margin. A score above 40% indicates a healthy balance of growth and profitability. GASP uses ARR growth rather than revenue growth and EBITDA margin rather than FCF margin, so two companies' scores compare like for like.

The metrics layer for the AI era

GASP is the open standard for SaaS metrics. Every metric has one definition and an exact formula that compiles to SQL, so your agents answer from the definition rather than a guess.

See the evidence Connect your agents Take the test

metrics: 313
departments: 13
in the graph: 69
relationships: 210

The evidence

Without the definition, a model guesses the metric

We asked two Claude models 24 Core metric questions over a SaaS warehouse and ran the SQL they wrote. From the schema alone they were right 60% and 70% of the time. With the GASP definition, both 100%.

Schema only

The table definitions and the question

Claude Sonnet 5.5: 60%
Claude Opus 5.5: 70%

Schema and the GASP definition

The same, plus the metric’s canonical formula and the fields it reads

Claude Sonnet 5.5: 100%
Claude Opus 5.5: 100%

Asked for Customer Acquisition Cost, the models answered 3,350 in 8 of 10 schema-only runs. The standard's value is 3,000. Both are a defensible CAC. The schema does not say which one the company reports, so the model chose.

Read the benchmark

Every question, the SQL behind each miss and the method.

Choose your path

One standard, three jobs

{ }

Engineer

Wire your warehouse to a metric model your agents can query


"mrr": "fact_subscription.mrr"

"customer_id": "dim_customer.id"

"invoice": "fact_invoice.total"

Build the adapter

Analyst

Standardise every metric definition across the company

NRR Cohort method, T12M

LTV ARPA × Margin / Revenue Churn

R40 ARR Growth + EBITDA Margin

See dashboard examples

⇄

Operator

Run your business on a graph your agents can navigate

Change one input, see what moves

Churn ↓ → NRR ↑ → ARR ↑

Cut Churned MRR 10% and NRR moves with it

Try the simulator

The map

The key metrics and how they connect

Every line is a relationship defined in the standard. Your agents walk the same graph through the MCP server when they trace a change back to its cause.

Metrics in the graph: 69
Entities: 28
Concepts: 7
Relationships: 210

Explore the graph

Engineer Analyst Operator Drag a node

The simulator

Move a lever, watch it ripple

Pick one decision, such as cutting Churned MRR by 10%. The TWIN Modifier shows how it moves every metric downstream of it as you drag the slider.

Try the TWIN Modifier

Lever

Churned MRR −10%

Impact

NRR: 108%; →; to 112%; +4pp
ARR: $12.0M; →; to $12.5M; +$0.5M
Rule of 40: 38%; →; to 42%; +4pp
LTV:CAC: 3.2×; →; to 3.6×; +0.4×

Illustrative. Propagation is simplified for demonstration, not financial planning.

The data layer

Wire your warehouse to the standard

Map your warehouse columns to the GASP ontology, then check which metrics your data covers and generate SQL for each of them in your warehouse's dialect.

Every formula is written in a grammar that compiles to SQL for Postgres (Redshift reads it), Snowflake, BigQuery and Databricks. The test suite runs every compiled query. Breakdowns by segment, plan, channel or cost centre are declared on the metric itself, so two teams splitting MRR by plan get the same rows.

01Your warehouse

fact_subscription
dim_customer
fact_invoice
dim_account
…

Postgres, Redshift, Snowflake, BigQuery or Databricks, with your own table names.

02adapter.json

{
  "mrr":         "fact_subscription.mrr",
  "customer_id": "dim_customer.id",
  "invoice":     "fact_invoice.total",
  "customer_status": "dim_customer.status"
}

280 fields, validated against the adapter schema.

03GASP ontology

Metrics in the graph: 69
Entities: 28
Concepts: 7
Source categories: 22
Query templates: 286

One versioned model for any warehouse.

Download the adapter template Read the adapter spec

The standard also exports as Apache Ossie (incubating) files that pass the upstream validator: ontology, semantic model, mapping and coverage.

For your agents

One MCP server for every metric

Listed in the Claude connectors directory, so Claude users connect in one click. Everyone else installs it with one command. When someone asks Claude why ARR is down, the agent looks up the definitions, formulas and relationships in the standard rather than making them up.

Connect in Claude Install via Claude Code

Hosted at mcp.gaspwiki.com for any MCP client, no sign-in. Read the MCP docs.

terminal

$ npx -y gasp-standard-mcp

claude · with gasp-standard-mcp

user

ARR is down. why?

claude

→ tool: infer_drivers("arr", "down")

Top candidates (of 36):

1. MRR ↓ (direct component, score 0.900)

2. Net New MRR ↓ (2 hops, score 0.675)

3. Churned MRR ↑ (3 hops, score 0.608)

Structural reasoning over the canonical relationship graph. No magnitude estimates.

For your board

Every dashboard, one set of definitions

Sales, finance and the board read Net Revenue Retention (NRR) from the same canonical definition, so the meeting starts with the number instead of "which NRR is this?"

NRR best-in-class

112%

Net Revenue Retention

GRR healthy

94%

Gross Revenue Retention

R40 above target

42%

Rule of 40

ARR Δ strong

+24%

ARR Growth (YoY)

Illustrative figures.

See example dashboards

Self-check

Could your numbers survive a board question?

Nine questions, from how your team defines churn to who owns each metric, show where your reporting differs from the standard.

9 questions 2 minutes no signup

Take the 9-question check

GASP Workspace

Your company's own metrics layer, run for you

GASP Workspace keeps your mapping to the standard, the decisions behind it with an owner on each, your check results and your readiness level over time. It compiles your metrics into your warehouse's own engine and serves them to every agent your company runs. Before a raise, a sale or a credit facility it gives you GASP certification: a reviewed, dated readiness report. We are building it now with three design partners.

Become a design partner

Built for: The data or AI platform lead at a SaaS company running agents over its metrics
Built on: The open GASP Standard, its exports and the gasp-standard-mcp package, all free
Status: In design, with three design partner places open

GASP ecosystem · AICF

Govern the agents

GASP says what metrics exist and AICF says how the AI systems reading them are governed: 201 canonical controls mapped across 12 frameworks including SOC 2, ISO 42001, NIST AI RMF and the EU AI Act, scoped by nine named profiles.

Visit aicf.gaspwiki.com

Canonical controls: 201
Source frameworks: 12
Named profiles: 9
Framework mappings: 2,989