Skip to content

Latest commit

 

History

History
74 lines (48 loc) · 4.51 KB

README.md

File metadata and controls

74 lines (48 loc) · 4.51 KB
Raw

Documentation Status

MIXpanel Bayesian AB test Analysis tool (MixBABA)

This tool is intended to consume a JSON file containing details about a Mixpanel funnel and output the results of the analysis made whithin a Bayesian framework.

You can find details about the data processing here.

Installation

From PyPi

Just easy as:

pip install mixbaba

From sources

To install this package you have to clone the repository:

git clone https://github.com/NaturalCycles/MixBABA.git

You can run the unit tests, to ensure the tool will work:

cd MixBABA 
python setup.py test

And then you can install the tool via PIP:

pip install .

Usage

You can find the full documentation here, but if you want a short guide read the following.

To use MixBABA you have need:

  • a JSON file containing a list of the funnels together with the details about them (an example is in this repository)
  • the "API secret" to connect to Mixpanel; You can find your one in the settings dialog of the Mixpanel web application.

Then you can launch the analysis via command line:

mixbaba -f [funnel_file.json] -k [API secret]

The tool will extract the data relative to the funnel from Mixpanel, and the output will be put in the same directory as CSV files, as many as the funnels specified in the JSON file given in input.

By default no output will be sent to the console. If you want a CSV file as output you can ask it with

mixbaba -f [funnel_file.json] -k [API secret] -o csv

Example result

This is the standard output format for the analysis of a funnel

Group Control Impressions Control Conversions test Impressions test Conversions test CR improvement test Probability
All.All 34164 253 31105 284 0.232387 0.992551
goal.PREVENT 6175 25 6016 37 0.500153 0.947624
goal.PLAN 1561 5 1411 5 0.106157 0.568093
$country_code.US 16631 224 15438 242 0.163448 0.95048
$country_code.SE 8024 23 7275 35 0.654391 0.974175

Or, if you specify the option -of long at the command launch:

Discriminant Cohort Comment Control Impressions Control Conversions test Impressions test Conversions test CR improvement test Probability
None All Result for test is OK! 34164 253 31105 284 0.232387 0.992551
user.goal PREVENT Result for test is uncertain. 6175 25 6016 37 0.500153 0.947624
user.goal PLAN Result for test is uncertain. 1561 5 1411 5 0.106157 0.568093
user.$country_code US Result for test is OK! 16631 224 15438 242 0.163448 0.95048
user.$country_code SE Result for test is OK! 8024 23 7275 35 0.654391 0.974175