Causal discovery on your own tabular data

Find out what actually drives your outcomes.

UCMT Labs discovers the causal structure in your tabular data, shows each relationship it finds, and lets you test an intervention before you commit to it — so decisions rest on causes, not correlations.

Start freeRead the docs
  • No credit card required
  • Start on the free plan

From patterns to causes.

priceadsdemandrevenuechurnyour datalearns patternsexplains causeshover a causewhat drives what?

Causal discovery, step by step

+0.78+0.71+0.55DemandInventory DrawDelivery DelayChurn Risk

Illustrative example — not a benchmark result.

01 / 04

Your variables, as they are.

Point the engine at a dataset. Every column becomes a candidate cause and a candidate effect — nothing is assumed.

02 / 04

Every relationship gets tested.

UCMT evaluates each ordered pair against the data, ruling out the associations that collapse once you condition on everything else.

03 / 04

Weights, and confidence where measured.

Surviving edges carry a signed weight, and a confidence where the algorithm measures one, so you can see which links to trust and which to keep questioning.

04 / 04

Follow the chain that matters.

Trace a path end to end to see how an upstream change propagates — and what it costs you downstream.

What every discovery run gives you

  • No invented confidence

    A confidence appears only where the algorithm measures one

  • Validation gates

    Checks run before a graph is released

  • Always a valid DAG

    Acyclicity enforced on every result

The product

From a dataset to a graph you can inspect

Run discovery on a dataset, then read the graph it found and every relationship in it, with its weight and, where the run measured one, its confidence. The overview keeps your datasets and runs in one place.

The causal workspace after a discovery run on a synthetic sample dataset of retail variables: the discovered graph, each edge labelled with its weight and confidence.The causal workspace after a discovery run on a synthetic sample dataset of retail variables: the discovered graph, each edge labelled with its weight and confidence.
The discovered graphSample data
The table of discovered edges for the same sample dataset: source and target, signed weight, confidence and effect type for each of the six edges.The table of discovered edges for the same sample dataset: source and target, signed weight, confidence and effect type for each of the six edges.
Every edge, with its weight and any measured confidenceSample data
The overview page with sample data: counts of datasets and discovery runs, storage used, and a chart of runs over the last 30 days.The overview page with sample data: counts of datasets and discovery runs, storage used, and a chart of runs over the last 30 days.
Your datasets and runs at a glanceSample data

Screenshots of the UCMT Labs app, showing sample data rather than a customer’s.

Platform capabilities

From your data to the causes behind it

Causal discovery, causal inference and intervention analysis on the data you already have.

  • Causal discovery

    UCMT models and classical algorithms (PC, GES, LiNGAM and more) map directed relationships across your columns.

  • Weights and measured confidence

    Signed weights, and a confidence where the algorithm measures one: never an invented number.

  • Validation gates

    Acyclicity enforcement and pre-validation checks before a graph is released.

  • Intervention analysis

    Ask “what if we change X?” on the discovered graph and see the downstream effect.

  • Explainable results

    Export the graph and its edges for the stakeholders who need to review them.

  • Team workspaces

    Organizations with role-based access to administrative features.

How it works

From raw data to a causal graph in three steps

  1. Upload your data

    CSV or JSON. No pipeline to build.

  2. Discover the causal graph

    UCMT tests every candidate relationship and enforces a valid DAG.

  3. Decide with evidence

    Explore the graph, test interventions and export results your stakeholders can audit.

Use cases

Where causal questions matter most

The same engine applies wherever you need to know why a metric moved.

  • Financial services

    Which factors actually drive default risk?

  • Life sciences

    Which treatment variables move outcomes?

  • Operations

    What upstream change is causing the delays?

  • Marketing

    Which channel caused the lift, not just correlated with it?

Pricing

Plans for every stage of causal work

Start on the free plan. Upgrade when your team needs more data, more runs or more seats.

  • Free
  • Starter
  • Professional
  • Enterprise
Compare plans and prices

Questions

Frequently asked questions

What data can I upload?

Tabular data as CSV files, or JSON as an array of objects (one per row). Excel workbooks are not read: save the sheet as CSV first. Each column becomes a variable the engine considers as a possible cause and a possible effect.

What does a discovery run produce?

A directed acyclic graph of the relationships the engine found in your data, with a signed weight on each edge and a confidence where the algorithm measured one; runs of the gated UCMT models (v2 Global, V3, V4) also report their quality gates and REF-001 data checks. You can explore it, test an intervention on it and export it.

Do I need a credit card to start?

No. A new account starts on the Free plan, and signing up has no payment step.

Are the graph and the screenshots on this page real results?

The step-by-step graph is an illustrative example, not a benchmark result. The screenshots are of the product itself, showing sample data rather than a customer’s.

Where can I see the prices?

On the pricing page, which lists every plan with its limits and its monthly and annual price in US dollars.

Ready to understand the why?

Start mapping the causal structure in your own data — no credit card required.

Start freeContact us