Skip to content

Mainakkundu/streamlit-docker-deploy

Repository files navigation

A/B Test Stat Engine - Streamlit App

Streamlit App for measuring the effectiveness of A/B testing scenario in 3 different verticals. This app focous on only continious KPI.

Components:

  1. Sample Size Estimation: This is the tab for pre-experiment analysis, where you can input base-mean,base-standard-deviation,average-traffic on your websites or store, and incremental lift you are anticipating and it will throw you the required sample size and duration for a test to go live.

  2. Stat Base Measurement: In this tab, you can measure the Test results is it significant or not. It can measure with 3 views:

Control Vs Test:

Measuring the incrementality between Test & Control group using T-Test and also tied the Confidence Intervals and p-value for significance.

CUPED:

Microsoft's approach to reducing the variance and increase the power of the test using pre-experiment data. Much more applicable if a metric is highly correlated with pre-behavior.

Non Parametric Test:

Measuring the incrementality between Test & Control group using Non-Parametric Method.

Pre Vs Post:

Measuring the effectiveness of Test after and before the intervention(offer day, campaign day, UI change any intentional change). This approach is highly recommended for the scenario where there is a no-hold out set aka Control group. This methodology will help to understand what is the cumulative incremental effect and relative effect on Pre and Post.

  1. Analysis & Recommendation: In this tab, you can measure the effectiveness of the Test on a day level which helps to understand the seasonality effects & novelty effect and how the overall test converges or not.

About

Docker deployment of ab testing app

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors