Hyperautomation and AI/ML: A Strategy for Digital Transformation Success.pdf
Kontagent Fb Developer Garage Final Jeff
1. Metrics for Viral Tuning
By: Jeffrey Tseng
Kontagent
Facebook Developer Garage SF 2009
2. PHAME
Growing the application efficiently
Problem
Virality can be engineered
Hypothesis
Testing variants, iterating quickly
Action
Viral factor, conversions
Metrics
Different contexts, call to actions, messaging
Experiments
8. Recipient View
Application
Sender Message
Acceptance
Installs
Msg % new users Conversion to
Metrics
Conversion invited Installs
9. Application View
A/B Test
# of Messages Social
Avg msgs
Call to Action
Sent Context
sent/event
Sender
Application
Recipient
Msg
Acceptance Message
Conversion
A/B Test
% new users Conversion to
Installs
invited Installs
10. The Metrics
Avg msgs
• Average message/event sent/event
• Msg conversion rate Msg
Conversion
• % of New Users Invited
% new users
• Conversion to Installs invited
Conversion to
Installs
• Repeat visits to the event!
Repeat Visits
to the Event
12. The Viral Co-efficient
Day 1 Day 2 Day 3
3 1
1
1
1
1
2
2
Simply the “average branching factor”
13. Viral Rate Co-efficient
average branching factor
average response time
0.5 1.0 0.5
days days days
14. Viral Rate vs. Viral Co-efficient
• Viral co-efficient
– Has no time dimension
– Can be used to tell if an app is viral
– Can be used for viral tuning (trending)
– Cannot be used to compare growth rate
• Viral rate
– How fast does the app grow?
– Can be used to compare growth rate of diff apps
– Can be used for projections
15. Why Track the Viral Tree?
• Absolute long-term measure of effectiveness
• Lifetime Network Value
• Tracking sources of installs (attribution)
– Paid source vs. organic
• Identify the most socially active users
– Message or treat them differently
20. Demographics for Viral Tuning
• Demographics
– Distribution does not equal behavioral distribution
– Measure the behavior of the
demographic/segment
• User segmentation is useful
– You can’t test EVERYTHING
– Segment users allows you to focus your testing
21. Final Notes on Viral Metrics
• Viral Metrics CAN
– Provide framework for test and experiment
– Allow you to iterate quickly
– A/B testing for small changes
Viral Metrics CANNOT build you a good product
The hard work is being creative
22. Long Term vs. Short Term
• Short term
– A/B for copy is a short-term local metrics
• Optimize for long-term metrics
– Time on site,
– LTV
– LNV
– ARPU