Quick Start Tutorial
Your First Analysis
This tutorial walks through all seven steps of the RootCause workflow using sample data. Download the two CSV files below — they represent customer demographics and subscription information for a fictional telecoms company.
Step 1: Connect Data
Go to Sources and click Import data. Select Local File Upload, drag both CSV files onto the drop zone, and click Upload. RootCause will process the files and begin building the ontology automatically.

Full details: Connect Data
Step 2: Build Ontology
Go to Ontology. RootCause has already scanned both datasets and built a semantic map — identifying that Customer Id appears in both files and linking them. Review the network to confirm the connections look right. No action is required to proceed.

Full details: Build Ontology
Step 3: Build 360 Table
Go to Ontology and look at the Recommended Datasets section in the right panel. Click the recommendation to create a joined view of both datasets in one click. RootCause generates the join recipe automatically — you can inspect and edit it in the Operations editor.

Full details: Build 360 Table
Step 4: Build Causal Graph
Causal discovery runs when you build a Digital Twin. Go to Digital Twins and click New Digital Twin (or, on the Ontology page, click Create & build twin on the recommended dataset — it opens the same builder with the view pre-selected). Keep the default method — Discover from data — and pick your joined Data View with Static as the type. The result of this run — a directed graph of cause-and-effect relationships — will land on the twin's Relationships tab, where you can click any node to see its top drivers and path analysis.

Full details: Build Causal Graph
Step 5: Build Digital Twin
In the builder's final Review & run step, click Discover & train. This one action runs causal discovery and trains the model. Training runs a sequence of stages and takes a few minutes. When it completes you'll see a model fit score and a set of suggested simulations.

Full details: Build Digital Twin
Step 6: Run Simulations
From your Digital Twin, click the Simulations tab and then New Simulation. Try typing a plain-language question — for example, "What happens to churn if we give away free tech support?" — and click Generate Scenario. Review the generated configuration and run it.

Full details: Run Simulations
Step 7: Produce Reports
Go to Reports, click New report, and choose Generate with AI. Describe the report you want — for example, "Causal analysis of churn drivers and the effectiveness of service interventions" — and use @ to reference your Digital Twin so the draft is grounded in it. RootCause produces a structured report with an Executive Summary and evidence-linked findings. Export to PDF to share it.

Full details: Produce Reports
What success looks like
If you followed all seven steps, you now have: two datasets linked through a shared Customer Id in the ontology, one joined Data View, and a trained Digital Twin with a model fit score on its Evaluation tab. Your simulation run shows a Completed status with an AI summary of the effect and the detailed result sections below it — and it stays in the Simulations tab's run list to revisit or export. From here, the same loop applies to your own data: connect it, review the ontology, build the view, train the twin, and start asking what-if questions.
Last updated

