Retentioneering is a Python framework to explore, grow and optimize your product based on deep analysis of user trajectories. Using Retentioneering you can vectorize clickstream logs and cluster user trajectories to automatically identify common successful or churn patterns. You can explore those patterns using our tools such as graph visualizer, step matrix, classifier and many others.
All you need to get started with Retentioneering is clickstream log of events from your web-site or app: {user_ID, event_ID, timestamp} (or use provided sample datasets in .csv format). You can vectorize individual user trajectories in dataset and plot all your users logs on 2D map using TSNE or UMAP projection:
data.retention.learn_tsne(plot_type='targets');
Users with similar patterns will appear as close dots at such map. Group of users who do not reach specified target event represent some systematic problem: usage pattern which systematically does not lead to product goals. Next you can segment users based on their behavior in the product.
Obtained user segments can be explored with graph visualizer or step matrixes or clustered again:
(data.retention.filter_cluster(4)
.retention.plot_graph(thresh = 0.05))
Plot reverse step matrix where rows correspond to events and columns show event position in the trajectory. Numbers show fraction of users having corresponding event at corresponding step:
(data.retention.filter_cluster(4)
.retention.get_step_matrix(reverse='neg'))
To explore more features please see the documentation
Firstly, you need to install python and Jupyter. We support only python 3.6 or higher versions. For quick start better to install Anaconda.
- You can install our package using pip:
pip3 install retentioneering
- Or directly from the source:
git clone https://github.com/retentioneering/retentioneering-tools
cd retentioneering-tools
pip3 install .
Explore example notebooks to get started or go through documentation pages:
- First steps Configuration, preparing your data graph basics.
- Step Matrix, Clustering Clustermap, TSNE projections, target events in clustering, cluster graphs.
- Supervised classifier Supervised learning, sklearn-api.
- Analysis Step matrix and clustering.
- Predict application remove
- Packages and Subpackages
This is community-driven open source project in active development. Any contributions, new ideas, bug reports, bug fixes, documentation improvements are very welcome.
Feel free to reach out to us: retentioneering[at]gmail.com