CausalML is Uber’s open-source Python library for uplift modeling and causal inference with machine learning. It has gathered about 6,000 stars on GitHub and is actively maintained, with updates in 2026.
Standard predictive models answer “who is likely to convert”, but businesses often need the harder causal question: “who will convert because of our intervention”. CausalML implements the algorithms for that job, including meta-learners (S-, T-, X-, R- and DR-learners), uplift trees and forests, and TMLE-style estimators, so data scientists can estimate heterogeneous treatment effects from experimental or observational data. It also ships evaluation metrics for uplift curves and sensitivity analysis tools, and it grew out of Uber’s production experience with large-scale experimentation and marketing optimization.
Beyond model fitting, CausalML helps teams decide what to do with the estimates: its feature-importance methods for uplift models explain which variables drive treatment-effect heterogeneity, and its interpretability tooling supports the conversations analysts must have with stakeholders before acting on a targeting policy. The library integrates with the wider Python ecosystem, accepting NumPy arrays and pandas DataFrames and interoperating with scikit-learn estimators, which makes it straightforward to drop into existing experimentation pipelines at scale.
CausalML is free and open source, released by Uber on GitHub for research and commercial use. There is no paid plan or hosted service; you install the library and run it in your own Python environment.
CausalML is for data scientists and experimentation teams in marketing, product and pricing who need to estimate the true incremental impact of treatments rather than simple correlations.