Complicated TV made easy, again
Personalizing video entertainment

in the age of abundance

(of content, technology and business models)
ACM RecSys Conference

Vienna 2015
ContentWise learns user’s taste and habits and surfaces relevant content
The ContentWise software suite blends the power of
recommendation algorithms with the convenience of
• editorial and operational tools
• assisted content curation
• business rules
• analytics
• a/b testing
• metadata enhancement
Complicated TV made easy, again
History of TV (super-simplified)
Early Broadcasting Thematic Channels
Early On-demand

DVR & Home Video
Internet Streaming
Unilateral
programming
curation
Abundance of
channels
Narrow audience
segments
“TV on my terms”
Program the DVR
Rent/buy DVDs
Overwhelming offer
Aggregation of
dispersed audiences
Long-tail effects
Hard to find relevant
content
Too many channels for
surfing
Where’s the TV Guide?
Content looks like a moving target
Live Airing
Restart TV Lookback EPG
Video
Recording
Catchup
(free for X days)
SVOD
Catalog
(older stuff?)
TVOD

($$$)
OTT Apps
(scattered
content)
?! Where is the
next episode?
GOAL
Predict the next user’s action and user’s intent
WHY?
Reduce TIME-TO-CHOICE
Content offering

is huge
Screens

are small
Attention span

is short
Did you say “scroll”?Did you say “search”? I’m sorry, what did you say?
(otherwise you lose the user)
Traditional “taste profile” has limitations
No adaptation to
user’s lifestyle
No adaptation to
context
No user’s intent
Taste
Profile
Very valuable, sophisticated; but almost static
Types
Topics
Actors
Genres
Directors
Storytelling
…
User’s lifestyle
Habits
Consumption patterns
User’s intent, in session
Let’s consider additional profiling dimensions
DEVICE TYPE
TIME
LOCATION
WEATHER?
NEWS?
HOLIDAYS?
…
Unified Profiling
Linear TV On-demand
Profile
Find a way to create a
unified profile from
heterogeneous interaction
schemes and data models
Prediction Discovery
Surfacing the next available
episode for each series the
user is already following
Surfacing Next Episodes + Suggesting New TV Series
Suggesting pilots from series
the user may like to try or
promoted series
Surface what’s relevant, for me, on this device, here, now
Danke schön!
pan@contentwise.tvYes, we’re
hiring!
www.contentwise.tv

Complicated TV Made Easy, Again

  • 1.
    Complicated TV madeeasy, again Personalizing video entertainment
 in the age of abundance
 (of content, technology and business models) ACM RecSys Conference Vienna 2015
  • 2.
    ContentWise learns user’staste and habits and surfaces relevant content The ContentWise software suite blends the power of recommendation algorithms with the convenience of • editorial and operational tools • assisted content curation • business rules • analytics • a/b testing • metadata enhancement
  • 3.
  • 4.
    History of TV(super-simplified) Early Broadcasting Thematic Channels Early On-demand
 DVR & Home Video Internet Streaming Unilateral programming curation Abundance of channels Narrow audience segments “TV on my terms” Program the DVR Rent/buy DVDs Overwhelming offer Aggregation of dispersed audiences Long-tail effects Hard to find relevant content Too many channels for surfing Where’s the TV Guide?
  • 5.
    Content looks likea moving target Live Airing Restart TV Lookback EPG Video Recording Catchup (free for X days) SVOD Catalog (older stuff?) TVOD
 ($$$) OTT Apps (scattered content) ?! Where is the next episode?
  • 6.
    GOAL Predict the nextuser’s action and user’s intent
  • 7.
    WHY? Reduce TIME-TO-CHOICE Content offering
 ishuge Screens
 are small Attention span
 is short Did you say “scroll”?Did you say “search”? I’m sorry, what did you say? (otherwise you lose the user)
  • 8.
    Traditional “taste profile”has limitations No adaptation to user’s lifestyle No adaptation to context No user’s intent Taste Profile Very valuable, sophisticated; but almost static Types Topics Actors Genres Directors Storytelling …
  • 9.
  • 10.
    Let’s consider additionalprofiling dimensions DEVICE TYPE TIME LOCATION WEATHER? NEWS? HOLIDAYS? …
  • 11.
    Unified Profiling Linear TVOn-demand Profile Find a way to create a unified profile from heterogeneous interaction schemes and data models
  • 12.
    Prediction Discovery Surfacing thenext available episode for each series the user is already following Surfacing Next Episodes + Suggesting New TV Series Suggesting pilots from series the user may like to try or promoted series
  • 13.
    Surface what’s relevant,for me, on this device, here, now
  • 14.