Platform Features

Everything you need for causal analytics.

From raw CSV to executive insights — automated, validated, explained. No data science PhD required.

Feature 01

Automated Causal Graph Discovery

CausoAI learns the causal structure of your data automatically — no manual DAG specification required. Choose from three industry-proven algorithms, or let the platform select based on your dataset characteristics.

  • —Visual DAG editor for manual adjustments
  • —Automatic sampling up to 50,000 rows for performance
  • —Version history — compare graph iterations
  • —Export to DOT format for use with external tools

PC Algorithm

Constraint-based

Uses conditional independence tests to identify the causal skeleton, then orients edges using v-structures.

GES

Score-based

Greedy Equivalence Search — optimizes a score function (BIC) over the space of equivalence classes of DAGs.

NOTEARS

Continuous optimization

Formulates DAG learning as a continuous optimization problem, enabling gradient-based discovery at scale.

Feature 02

5-Layer Causal Readiness Score

Not all causal analyses are equally trustworthy. CRS quantifies exactly how confident you should be — and tells you what gaps need to be fixed.

L1

Data Coverage

25%

Sample size, missingness, column completeness — do you have enough high-quality data?

L2

Identifiability

25%

Temporal ordering, backdoor criterion, collider detection — is the causal effect computable?

L3

Estimation Feasibility

20%

Model assumptions, data requirements — can the estimator be applied to this data?

L4

Statistical Power

20%

Propensity score overlap, effective sample sizes — is the sample large enough to detect effects?

L5

Robustness

10%

Sensitivity to unmeasured confounders — how stable are results across methods?

Score Interpretation

80–100

High confidence

Suitable for decision-making

60–79

Directionally reliable

Inform hypotheses

40–59

Use with caution

Significant assumptions

< 40

Do not act

Insufficient for action

Feature 03

Automatic Causal Effect Estimation

CausoAI selects the statistically appropriate estimator based on your treatment type and sample size — no manual configuration required.

AIPW

Augmented Inverse Probability Weighting

Doubly Robust
When: Binary treatment + N ≥ 500

Combines outcome and propensity models for robustness against misspecification of either.

PSM

Propensity Score Matching

Matching
When: Binary treatment + N < 500

Matches treated and control units on propensity scores for reliable small-sample estimates.

CausalForest DML

Causal Forest Double ML

Heterogeneous Effects
When: Continuous treatment + N ≥ 1000

Discovers heterogeneous treatment effects across subgroups using ensemble methods.

LinearDML

Linear Double Machine Learning

Partialling Out
When: Continuous treatment + N < 1000

Uses cross-fitting to remove confounding while maintaining linear interpretability.

Also included: Heterogeneous treatment effects (T-Learner, X-Learner, DR-Learner) for segment-level analysis — identify which customer groups respond most to an intervention.

Feature 04

What-If Counterfactual Simulator

Once the causal model is built, you can run forward simulations: change the value of a treatment variable and see the predicted outcome — with confidence intervals.

  • —Continuous slider for treatment intensity (e.g., ad spend ±50%)
  • —95% confidence intervals on all predictions
  • —Dose-response curves for continuous treatments
  • —Segment-level counterfactuals for targeted decisions

Counterfactual Simulation

+20%

Predicted Outcome

+$340Krevenue lift
95% CI: $280K — $410K
$280K$340K$410K
AI Insightsclaude-sonnet-4-6
"Increasing email campaign frequency from 2x to 4x per week causally increased 30-day conversion rate by 8.3 percentage points (95% CI: 6.1–10.5 pp). The effect is driven primarily by the re-engagement segment — users inactive for 15–30 days show the strongest response (+12.7 pp)."

Feature 05

AI-Powered Insights via Claude

CausoAI integrates with Claude (claude-sonnet-4-6) to translate statistical outputs into plain-language executive summaries, assumption explanations, and actionable recommendations — cached for 7 days to control API costs.

  • —Executive summaries explaining causal effects in plain English
  • —Assumption & limitation explanations for non-technical stakeholders
  • —3+ concrete, prioritized action recommendations
  • —Natural language Q&A — ask follow-up questions about your analysis

Feature 07

Connect your data warehouse directly.

Skip the CSV export. Connect CausoAI to PostgreSQL, MySQL, or Google Sheets, write a query, preview the results, and import — all without leaving the platform.

  • —Preview 100 rows before committing to an import
  • —Read-only query validation — no accidental writes
  • —Imported data is auto-profiled and ready for analysis
  • —Credentials stored encrypted at rest

PostgreSQL

Database

Connect to any PostgreSQL instance. Write a SELECT query, preview 100 rows, and import directly — no CSV export needed.

MySQL

Database

Connect to MySQL or MariaDB. CausoAI validates your query is read-only before execution.

Google Sheets

Spreadsheet

Authenticate via service account and pull a sheet directly into your analysis workspace.

Feature 08

Advanced causal techniques — built in.

Beyond basic effect estimation, CausoAI includes a full suite of advanced causal methods that typically require specialist knowledge to run. They are automated, validated, and explained in plain English.

Mediation Analysis

Direct vs. indirect

Decompose a causal effect into its direct path and indirect paths through mediating variables. Get Natural Direct Effect (NDE), Natural Indirect Effect (NIE), and proportion mediated — essential for understanding mechanisms, not just outcomes.

Difference-in-Differences

Panel & time-series

Run temporal causal inference with entity and time fixed effects. Includes a parallel trends pre-test, event study curves, and DiD coefficient — the gold standard for measuring the impact of policy changes or interventions over time.

Subgroup Discovery & Policy Optimization

Heterogeneous effects

Automatically segment your population by who responds to treatment. CausoAI converts CATE estimates into decision-tree targeting rules and cumulative uplift curves — so you know exactly which customers to act on.

Root Cause Analysis

Upstream tracing

When a metric moves, trace its upstream contributors through the causal DAG. CausoAI weights each ancestor variable by its path contribution and returns a ranked list of root causes — not just correlated features.

Built on

Proven open-source causal inference libraries.

CausoAI wraps the best causal inference tools in a unified, automated platform — so your team gets production-grade results without needing to configure each library.

DoWhy

Causal effect estimation framework

EconML

Heterogeneous treatment effects (DML, Causal Forest)

CausalLearn

PC and GES graph discovery algorithms

CausalNex

NOTEARS continuous DAG learning

Claude API

AI-generated insights and natural language Q&A

NetworkX

Graph representation and manipulation