System Design: Twitter

[esthicodes]

Stuff like:

• Back-of-the-envelope calculations
• Sample data sets and data flows
• Assumptions about the product/domain/problem
• Pseudo code (or you can switch to an actual coding language)
• Links to files and other resources
• Online profiles

Snap Chat engagemnet model

1. Engagement is user activity on the app leading to daily active user counts
2. userse sending messages and videos to other users or watching stories
3. DAU is a useful metric that we can optimize when the goal is to improve engagements

Requirements

1. Improve DAU and as a seo day measure # of users engaging with videos or text messages with other users

DAU = # of users who opened the app at least once a day
DAU(Engaged) = # of users who sent at least one message or viewed at least one story

2. We should discuss which country we are designing/targeting our systemn for. Let say we pick USA

Data Needed

1. Past few months of user activity logs - # of users who have been actively sending messages and those who are not but just open the app - USA data for now

2. Snapchat ranking in app stores specifically for USA market

3. of downloads for the app for past few months (six months are good)

4. we need to compute the ratio of active users to the whole population of app downloaded users

Development Method

1. User research: Gather data on Snapchat usage in the USA, including demographics, user behavior, and preferences.

This could involve conducting surveys, focus groups, or other types of user research to gain insights into why people use Snapchat, what they enjoy about it,
and what they would like to see improved. We can mine app store reviews.

2. Data analysis: Analyze the data collected from user research to identify patterns and trends in Snapchat usage.

This could include using data visualization tools to create graphs, charts, and heat maps to show the usage of Snapchat over time,

the demographics of Snapchat users, and the types of content they engage with the most.

3. Identifying key drivers of engagement:
   Based on the data analysis, identify the key drivers of engagement, such as user experience, content, and features.

These drivers can be used to prioritize and focus on areas where improvements can have the greatest impact.

4. Develop and implement targeted solutions: Develop and implement targeted solutions based on the key drivers of engagement. This could involve launching new features or improving existing features, creating new types of content, or improving the overall user experience.

5. Monitor and measure success: Continuously monitor the impact of the solutions implemented to determine if they are having the desired effect of increasing engagement.

This could involve tracking metrics such as user engagement, active user growth, and time spent on the platform.
Examples of IMprovements

1. Age group 24-30 would love to have MTV available on snapchat - making them happy by providing this content
2. Some users would like app to be available on web - we will provide that

Mathematical Foundation:

1. We can formulate the problem to be binary classifictaion based on users who are using the app and those who dont
2. Building the feature set based on all the data that we gathered above
3. Formulating our backend model to be DNN - with output layer just having two neurons which predict one class or ther other
4. Binomial classifier would suffice

Features

1. app ranking
2. user speciifc feature
   a. gender
   b. age
   c. income
   d. location
   e. interests - music, clothers, resturants
3. content
   a. content category
   a. news
   b. music
   c. ..

i have a question:

• Binomial classifier would suffice, why and how?

could you make a design?