A CDM-based data platform toolchain

Your data model shouldn't just describe the system. It should run it.

Igloo turns one versioned model — built on Microsoft's Common Data Model language — into a working data platform: validation, connectors, deployment, and a data space you can share with partners. No separate wiki that drifts from what the database actually enforces.

Deploy anywhere — Kubernetes, on-prem, local dev, any cloud, or hosted by us. Starts from your existing finance or building CDM.

Your project layer
local overrides, implementation decisions
Domain layers
industry models — e.g. real estate CDM
Enhancement layers
validation traits, units, PII, IoT semantics
Semantic core
sharpened meanings over the base
Microsoft CDM foundations
primitives, meanings, data formats

// each layer overrides what's below it — without forking, without renaming

Schema and meaning disagree by default

Every data org runs on two artifacts that drift apart the day they're born — and every report, pipeline, and AI agent re-derives its own guess at the gap.

The schema

Enforces structure. Executes. Says nothing about what the data means, or why a field exists.

The meaning

Lives in the wiki, the glossary, tribal knowledge. Says everything. Executes nowhere.

Keep your data. Change what it can do.

Igloo works across three stages, whether you're starting from a spreadsheet, an existing database, or a blank model.

01

Model

Build a data platform from your existing finance or building CDM — or reverse-engineer one from a database nobody ever modeled.

02

Enforce

The model compiles into validation plans, generated code, and test data automatically — no hand-written checks to keep in sync.

03

Share

Expose the platform as a data space to partners, banks, or regulators — with the same rules enforced on their side as yours.

deploy

Runs wherever your data already lives

Kubernetes, on-premise, local dev, any cloud — or we host it for you. The platform is a deterministic, versioned artifact you can diff, blue/green-deploy, and roll back like code.

quality

Data quality checks, out of the box

A twelve-metric taxonomy — type, format, uniqueness, referential integrity, completeness, and more — generated straight from the model, never hand-written and never drifting from it.

connect

Straight out of Excel

Keep your team working in the spreadsheets they already use. Igloo shapes that data into the model's own compliant format — plus connectors beyond Excel as your sources grow.

evolve

Standards that improve without breaking

Layer your own decisions on top of an industry model without forking it. Every proposed change is checked against your policy — no breaking change reaches production unnoticed.

What's actually built today

We'd rather tell you exactly what's real than sell you a roadmap. Here's the honest split.

Built & hardened
Core SDK and layering engine

Full CDM language support, conformance-tested against Microsoft's own corpus, with a layering engine for identity, resolution, squash, and materialize.

Built & hardened
Data quality & test data generation

Twelve-metric validation, generated test corpora that load into a real database under full constraints, and a compiled quality plan you can embed.

Built & hardened
Forward & reverse engineering

Turn a database dump into a model, or a model back into deployable DDL — closing the schema-to-data loop end to end.

In progress
Visual modeler

The diagramming workbench for hand-authoring models. Core navigation and rendering ship today; the full editing surface is still being built.

In progress
Code generation

Typed DTOs and validators generated from the same model definitions that drive quality checks today.

Designed, not yet built
Guided reference repair

Point at a broken or partial model and get a scoped, explained path to a valid one.

See it running on your own model — not a demo dataset.