22. what else is there besides clicks?
this talk: bit.ly/nyt-engagement
23. what else is there besides clicks?
this talk: bit.ly/nyt-engagement
24. “engagement”: examples
if your biz model is clicks,
engagement=clicks
if your biz model is sharing,
engagement=sharing
if your biz model is time on page,
engagement=time on page
if your biz model is subscription…?
25. “engagement”: examples
if your biz model is clicks,
engagement=clicks
if your biz model is sharing,
engagement=sharing
if your biz model is time on page,
engagement=time on page
if your biz model is subscription…?
26. “engagement”: examples
if your biz model is clicks,
engagement=clicks
if your biz model is sharing,
engagement=sharing
if your biz model is time on page,
engagement=time on page
if your biz model is subscription…?
27. “engagement”: examples
if your biz model is clicks,
engagement=clicks
if your biz model is sharing,
engagement=sharing
if your biz model is time on page,
engagement=time on page
if your biz model is subscription…?
30. from “data scientists @ work”
-Caitlin Smallwood
VP, Science and Algorithms at Netflix
this talk: bit.ly/nyt-engagement
31. from “data scientists @ work”
-Caitlin Smallwood
VP, Science and Algorithms at Netflix
this talk: bit.ly/nyt-engagement
32. WWND?
if your biz model is subscription,
machine learning can help:
Balance predictive power for true KPI
(retention) with
1. interpretability
2. should be
• easy to measure,
• quick to measure,
• or both
33. ML can help!
“engagement” is hard to define. you choose:
1. poetry
2. philosophy
3. science
Wbinan of f2) which predicts 1)
34. ML can help!
“engagement” is hard to define. you choose:
1. poetry
2. philosophy
3. science
WE CHOSE SCIENCE:
• find 1) reality: KPI, preferably units of USD
• find 2) interpretable and observable features
• learn combination of 2) which predicts 1)