mitzu 0.5.5

Product analytics over your data warehouse

Mitzu is an open source product analytics tool that queries your data lake or data warehouse.

Features

• Visualization for:
  • Funnels
  • Segmentation
  • Retention
  • User Journey (coming soon)
  • Revenue calculations (coming soon)
• User Lookup (coming soon)
• Cohorts Analysis
• Standalone web app for non-tech people
• Notebook visual app
• Notebook low-code analytics in python
• Batch ETL jobs support

Supported Integrations

Mitzu integrates with most modern data lake and warehouse solutions:

• AWS Athena
• Databricks Spark (SQL)
• Files - SQLite (csv, parquet, json, etc.)
• MySQL
• PostgreSQL
• Snowflake
• Trino / Starburst

Coming Soon

• Clickhouse
• BigQuery
• Redshift

Quick Start

In this section, we will describe how to start with Mitzu on your local machine. Skip this section if you rather see Mitzu in a prepared notebook or webapp. Otherwise get ready and fire up your own data-science notebook.

---

Install the Mitzu python library

pip install mitzu

Reading The Sample Dataset

The simplest way to get started with Mitzu is in a data-science notebook. In your notebook read the sample user behavior dataset. Mitzu can discover your tables in a data warehouse or data lake. For the sake of simplicity we provide you an in-memory sqlite based table that contains

import mitzu.samples as smp

dp = smp.get_sample_discovered_project()
m = dp.create_notebook_class_model()

Segmentation

The following command visualizes the count of unique users over time who did page visit action in the last 30 days.

m.page_visit

In the example above m.page_visit refers to a user event segment for which the notebook representation is a segmentation chart. If this sounds unfamiliar, don't worry! Later we will explain you everything.

Funnels

You can create a funnel chart by placing the >> operator between two user event segments.

m.page_visit >> m.checkout

This will visualize the conversion rate of users that first did page_visit action and then did checkout within a day in the last 30 days.

Filtering

You can apply filters to user event segment the following way:

m.page_visit.user_country_code.is_us >> m.checkout

# You can achieve the same filter with:
# (m.page_visit.user_country_code == 'us')
#
# you can also apply >, >=, <, <=, !=, operators.

With this syntax we have narrowed down our page visit user event segment to page visits from the US. Stacking filters is possible with the & (and) and | (or) operators.

m.page_visit.user_country_code.is_us & m.page_visit.acquisition_campaign.is_organic

# if using the comparison operators, make sure you put the user event segments in parenthesis.
# (m.page_visit.user_country_code == 'us') & (m.page_visit.acquisition_campaign == 'organic')

Apply multi value filtering with the any_of or none_of functions:

m.page_visit.user_country_code.any_of('us', 'cn', 'de')

# m.page_visit.user_country_code.none_of('us', 'cn', 'de')

Of course you can apply filters on every user event segment in a funnel.

m.add_to_cart >> (m.checkout.cost_usd <= 1000)

Metrics Configuration

To any funnel or segmentation you can apply the config method. Where you can define the parameters of the metric.

m.page_visit.config(
   start_dt="2021-08-01",
   end_dt="2021-09-01",
   group_by=m.page_visit.domain,
   time_group='total',
)

• start_dt should be an iso datetime string, or python datetime, where the metric should start.
• end_dt should be an iso datetime string, or python datetime, where the metric should end.
• group_by is a property that you can refer to from the notebook class model.
• time_group is the time granularity of the query for which the possible values are: hour, day, week, month, year, total

Funnels have an extra configuration parameter conv_window, this has the following format: <VAL> <TIME WINDOW>, where VAL is a positive integer.

(m.page_visit >> m.checkout).config(
   start_dt="2021-08-01",
   end_dt="2021-09-01",
   group_by=m.page_visit.domain,
   time_group='total',
   conv_window='1 day',
)

SQL Generator

For any metric you can print out the SQL code that Mitzu generates. This you can do by calling the .print_sql() method.

(m.page_visit >> m.checkout).config(
   start_dt="2021-08-01",
   end_dt="2021-09-01",
   group_by=m.page_visit.domain,
   time_group='total',
   conv_window='1 day',
).print_sql()

Pandas DataFrames

Similarly you can access the results in the form of a Pandas DataFrame with the method .get_df()

(m.page_visit >> m.checkout).config(
   start_dt="2021-08-01",
   end_dt="2021-09-01",
   group_by=m.page_visit.domain,
   time_group='total',
   conv_window='1 day',
).get_df()

Notebook Dashboards

You can also visualize the webapp in a Jupyter Notebook:

import mitzu.samples as smp

dp = smp.get_sample_discovered_project()
dp.notebook_dashboard()

Usage In Notebooks

• Example notebook
• Documentation

Webapp

Mitzu can run as a standalone webapp or embedded inside a notebook.

Trying out locally:

docker run -p 8082:8082 imeszaros/mitzu-webapp

• Example webapp
• Webapp documentation

Connect Your Own Data

Mitzu is be able to connect to your data warehouse or data lake. To get started with your own data integration please read our handy docs

Contribution Guide

Please read our Contribution Guide