Summary

Retentioneering-tools is a Python toolkit for reproducible, auditable behavioral analytics on event data, targeting data scientists, analysts, and AI agents. It transforms raw clickstreams and event logs into actionable insights on user behavior, conversion paths, and segmentation, operating locally without data exfiltration. Version 5.0 introduces significant performance and feature enhancements.

How It Works

Version 5.0 is a ground-up rewrite featuring a DuckDB-backed Eventstream for speed, coupled with new open-source anywidget-based interactive visualizations. An integrated MCP server allows LLM agents to query the eventstream, enabling reproducible analysis. Data processing uses a chainable API for flexible event stream manipulation, from filtering to sessionization.

Quick Start & Requirements

Requires Python 3.11+. Install via pip install retentioneering. Basic usage involves loading event data (user ID, event, timestamp) into a pandas DataFrame, initializing an Eventstream, and calling methods like transition_graph() for interactive notebook visualizations. A bundled synthetic e-commerce dataset is available. Documentation: https://retentioneering.com/docs.

Highlighted Details

• Interactive widgets (Transition Graph, Funnel, Step Matrix, etc.) visualize user journeys and behavior.
• "Diff mode" enables direct comparison of behavior between any two user segments.
• Chainable data processors offer flexible event stream cleaning, sessionization, and enrichment.
• Headless widget twins provide raw data for programmatic analysis and ML feature engineering.
• MCP server allows LLM agents to explore data and generate validated HTML reports.

Maintenance & Community

Active, community-driven open-source project. Contributions (bugs, features, docs) welcomed via issues/PRs. Contact: retentioneering@gmail.com. Key contributors: Maxim Godzi, Vladimir Kukushkin, Anatoly Zaytsev.

Licensing & Compatibility

Licensed under Apache License, Version 2.0, permitting commercial use, modification, and distribution. Excludes rights to Retentioneering's name, trademarks, hosted services, or proprietary cloud infrastructure not in this repo.

Limitations & Caveats

Version 5.0 is a major rewrite from the legacy 3.x engine (maintained separately). While core tools are Apache 2.0, other Retentioneering libraries, services, and enterprise integrations are proprietary and subject to separate commercial terms.