Skip to content

Retentioneering: product analytics, data-driven customer journey map optimization, marketing analytics, web analytics, transaction analytics, graph visualization, and behavioral segmentation with customer segments in Python. Opensource analytics, predictive analytics over clickstream, sentiment analysis, AB tests, machine learning, and Monte Car…

Switch branches/tags

Rete logo

Pipi version Telegram Python version Downloads Travis Build Status

Language grade: JavaScript Language grade: Python

What is Retentioneering?

Retentioneering is a Python framework and library to assist product analysts and marketing analysts as it makes it easier to process and analyze clickstreams, event streams, trajectories, and event logs. You can segment users, clients (agents), build ML pipelines to predict agent category or probability of target event based on historical data.

In a common scenario you can use raw data from Google Analytics BigQuery stream or any other silimal streams in form of events and their timestamps for users, and Retentioneering is all you need to explore the user behavior from that data, it can reveal much more isights than funnel analytics, as it will automatically build the behavioral segments and their patterns, highlighting what events and pattern impact your conversion rates, retention and revenue.

Retentioneering extends Pandas, NetworkX, Scikit-learn for in-depth processing of event sequences data, specifically Retentioneering provides a powerful environment to perform an in-depth analysis of customer journey maps, bringing behavior-driven segmentation of users and machine learning pipelines to product analytics.

Most recent is Retentioneering 2.0.0, this version has major updates from 1.0.x and it is not reverse compatible with previous releases due to major syntax changes. With significant improvements we now provided architecture and the solid ground for farther updates and rapid development of analytical tools. Please update, leave your feedback and stay tuned.

intro 0


This is new major release Retentioneering 2.0. Change log is available here.

Complete documentation is available here.


Option 1. Run directly from google.colab. Open google.colab and click File-> “new notebook”. In the code cell run following to install Retentioneering (same command will install directly from Jupyter notebook):

!pip3 install retentioneering

Option 2. Install Retentioneering from PyPI:

pip3 install retentioneering

Option 3. Install Retentioneering directly from the source:

git clone
cd retentioneering-tools
python3 install

Quick start

Start using Retentioneering for clickstream analysis

Or directly open this notebook in Google Colab to run with sample data.

Suggested first steps:

import retentioneering

# load sample user behavior data as a pandas dataframe: 
data = retentioneering.datasets.load_simple_shop()

# update config to pass columns names:
    'user_col': 'user_id',

Above we imported sample dataset, which is regular pandas dataframe containing raw user behavior data from hypothetical web-site or app in form of sequence of records {'user_id', 'event', 'timestamp'}, and pass those column names to retentioneering.config. Now, let's plot the graph to visualize user behaviour from the dataset (read more about graphs here):

                     targets = {'payment_done':'green',

intro 1

Here we obtain the high-level graph of user activity where edge A --> B weight shows percent of users transitioning to event B from all users reached event A (note, edges with small weighs are thresholded to avoid visual clutter, read more in the documentation)

To automatically find distinct behavioral patterns we can cluster users from the dataset based on their behavior (read more about behavioral clustering here):


intro 1

Users with similar behavior grouped in the same cluster. Clusters with low conversion rate can represent systematic problem in the product: specific behavior pattern which does not lead to product goals. Obtained user segments can be explored deeper to understand problematic behavior pattern. In the example above for instance, cluster 4 has low conversion rate to purchase but high conversion rate to cart visit.

clus_4 = data.rete.filter_cluster(4)
                        targets = {'lost':'red',

intro 1

To explore more features please see the documentation

Step-by-step guides


This is community-driven open source project in active development. Any contributions, new ideas, bug reports, bug fixes, documentation improvements are very welcome.

Retentioneering now provides several opensource solutions for data-driven product analytics and web analytics. Please checkout this repository for JS library to track the mutations of the website elements:

Apps are better with math!:) Retentioneering is a research laboratory, analytics methodology and opensource tools founded by Maxim Godzi and Anatoly Zaytsev in 2015. Please feel free to contact us at if you have any questions regarding this repo.