# Learn (In ≪2 Mins) Ab...
## Metadata
- Author: [[@bandanjot on Twitter]]
- Full Title: Learn (In ≪2 Mins) Ab...
- Category: #tweets
- URL: https://twitter.com/bandanjot/status/1497675344739979269
## Highlights
- Learn (in <2 mins) about FOUR types of A/B experiments you can run for your products/features: 🧵 ([View Tweet](https://twitter.com/bandanjot/status/1497675344739979269))
- 1. A/A Experiment for Data collection
In this case A and B are both exactly same versions.
Run such experiments to collect baseline data. But also to check there whether B is exactly same as A (if it is not , then you will see difference in metric performance of A vs B) ([View Tweet](https://twitter.com/bandanjot/status/1497675345524314113))
- 2. A/B experiments to check for improvements
Run this experiment to check whether variation B improves your primary and supporting metrics over A and proves your hypothesis that you set before starting experiment. ([View Tweet](https://twitter.com/bandanjot/status/1497675346468130822))
- 3. A/B1/B2/... experiment to check for performance of multiple variants
Here you want to check more than one variation (B1, B2.. and so on) and see which one performs better as compared to base (A). ([View Tweet](https://twitter.com/bandanjot/status/1497675347235676161))
- Summary of four types of A/B tests:
1. A/A tests for data collection
2. A/B tests to check for improvement
3. A/B1/B2.. to check multiple variants and their performance against the base (A)
4. A/B no-harm test (to ensure B does no harm over A) ([View Tweet](https://twitter.com/bandanjot/status/1497675348909117440))
- If you learnt something new about experimentation , do retweet the first tweet:
https://t.co/sfQeK7C98X ([View Tweet](https://twitter.com/bandanjot/status/1497680356610293762))