Transcript Music recommendations

Music recommendations
Bjørn Tennøe – FAST Global Services – March-October 2007
[email protected]
Use design resources to create end user value
Range of
investments
56%
Tools and
calculators
Tangibles
(usefulness)
36%
Brand
62%
18%
13%
Value
Information
content
48%
62%
Navigation
32%
Speed
Defining value for the end user
Peter Moreville
28%
Usability
18%
Help/
Tutorials
Price
16%
Admin
Fee
47%
Transaction
Fee
End users’ perception of value
Quantitative survey and regression analysis
Research by Phase 5 for undisclosed financial institution
2
Adapt to the users’ mind set
External
GUI, branding
Internal
Data model, organization
Make your service thrive
by adapting it
to the user’s mind set.
Don’t alienate users
by mirroring the service
to the organization & data model.
3
Bjørn Tennøe
Recommendations are key to online media services
– Online media services offer vast content collections.
– Most of the media collection is unknown to the consumer,
and therefore the user must get recommendations to make best use of the collection.
• Subscription services without recommendations have lower traction with consumers.
• Also media stores benefit greatly from recommendations, in addition to premium editorial content.
– Recommendations can be social, editorial or search powered (personalized). While SDP supports all
approaches, this presentation focuses on search powered recommendations.
4
Some recommendation benefits
– Drives consumers away from Top 20
• Guide towards higher margin items, avoid DRM issues
– Offers “convenience” as an alternative to “free”
• The potential of “convenience” is not fulfilled
– Introduces new service types
• Last.fm
• Pandora
5
Types of search powered recommendations
– Item to item, non personalized:
Consumers of item A
also consumed items B, C…
– Collaborative filtering, personalized:
People that have habits similar to you,
and that consumed item A,
also consumed items B, C…
– Research:
• Sequences: Given that track A is desirable,
find tracks B, C… that are appropriate to
follow.
6
Who benefits most from recommendations?
Serendipitous  Genre sensitive  Picky

Target users
for recommendations
7
Example approach: Integration with overall service
Player
Use recommendations to
 explore alternatives
Library
Playlist
Use
recommendations to
 narrow
down
Use recommendations to
 expand
– The example on the following pages are centered
around recommendations for an
online music subscription service.
– A recommendations interface with a library/playlist
structure allows seamless integration with the
overall service.
– Recommendations in the library should
 narrow down the library to something the user
can relate to.
– Recommendations in the playlist should
 expand discover something new.
– Recommendations in the player should
 offer alternatives to the current style.
8
Search powered music player
Demos available on www.tennoe.no/FAST
9
Mobile music player with recommendations
Demos available on www.tennoe.no/FAST
Start page w/ recommendations
Choose source
Expand from source
Play screen with escape paths
10
Keep the interface as simple as possible
– Avoid metadata based navigation whenever possible.
• Rather, let the user navigate by (example) content.
• Analogy: Advanced versus simple search interfaces.
– Does the user care about data type?
Can “Performing Artist”, “Conductor” and “Playlist” be presented in the same result set?
12
A note on normalization
– For music collections: Normalize and simplify the user’s track metadata for better
findability in the library.
• Sort songs by “Nelli Futhado” and “Nelly Furtado feat. Justin Timberlake” under “Nelly
Furtado”.
• A Berliner Philarmoniker CD may be found both under Karajan (the conductor), Mutter
(the violinist) and Bach (the composer). In such cases, the decision on what metadata
to display depends on frequency.
• All tracks must be represented with minimum 1 entry (artist/playlist) in the library.
13
Way forward
– FAST’s user experience team can:
• Contribute with design specifications
• Offer best practice & guidance to peer design teams
• Facilitate creative workshops & quality control activities
• Assist in prototyping
14
Thank you