Product Management from Facebook to GitHub: Using data to understand - - PowerPoint PPT Presentation
Product Management from Facebook to GitHub: Using data to understand - - PowerPoint PPT Presentation
Product Management from Facebook to GitHub: Using data to understand users & build products Identify Define Creating Opportunities Success Experiences Who am I? Cristina Fernndez Counsyl (genetics) GitHub Facebook Identify Define
Identify Opportunities Creating Experiences Define Success
Who am I?
Cristina Fernández
Facebook Counsyl (genetics) GitHub
Identify Opportunities Creating Experiences Define Success
Identify Opportunities Creating Experiences Define Success
Initial Assumptions (@FB)
- Blocked users are spammers
- People should only be friending people they know
But when we dug into the data the profiles didn’t look like spammers
- Established accounts
- With 100+ of friends
- Who posted regularly
So why are the users being blocked?
Supplement data patterns with user interviews
- Wanted to explore the world - meet
new people!
- Didn’t understand the “rule” about
- nly friending people you know
(cultural expectation)
- Most people liked receiving friend
requests from new people
What did we do?
1. Trained algorithm on localized data 2. Changed communication around WHY people were being blocked
Results:
- Less people actually getting blocked (happy users)
- More friend requests sent & accepted (FB growth)
Data served a starting point for discovering an opportunity
Identify Opportunities Creating Experiences Define Success
#vanitymetrics
New Feature: Require 2fa
- User requests
- Manual workaround
- Improve security
How many Organizations can we expect to enable this?
Focus on Organizations that are already requiring 2fa
Estimation parameters
- Organizations that are actively developing code
- >10 Users
- >80% of users have 2fa enabled
Define a goal that reflects the user needs being addressed
Identify Opportunities Creating Experiences Define Success
Artificial Intelligence
Starting from scratch meant we needed to build all the building blocks
Develop ML and NLP algorithms Create a production ready database with Topic suggestions Make Topic suggestions accessible for users Build in feedback mechanism to improve the algorithm
We needed lots of different teams and skillsets
- Data Scientists (Machine Learning)
- Data Engineering (Data Infrastructure)
- Application/Front-End Engineers (user facing features)
- Design (UX)
Cross functional work allows you to bring all the pieces together
Identify Opportunities Creating Experiences Define Success