Last.fm - Lessons from building the World's largest social music platform

Slideshow Transcript

1. Slide 1: Lessons from Building The World’s Largest Social Music Platform Matthew Ogle & Anil Bawa Cavia FoWA 2007
2. Slide 5: How does it work?
3. Slide 6: How does it work? Listen - to your own music - to Last.fm radio
4. Slide 7: How does it work? Listen - to your own music - to Last.fm radio Share -your taste instead of files - automatic, via “scrobbling”
5. Slide 8: How does it work? Listen - to your own music - to Last.fm radio Share Discover -your taste - via your friends instead of files - via people with - automatic, via similar taste “scrobbling”
6. Slide 9: How does it work? Listen - to your own music - to Last.fm radio Share Discover -your taste - via your friends instead of files - via people with - automatic, via similar taste “scrobbling”
7. Slide 11: 15 million tracks scrobbled / day
8. Slide 12: 15 million tracks scrobbled / day (175 scrobbles per second)
9. Slide 13: 15 million tracks scrobbled / day (175 scrobbles per second) Over 6 billion tracks scrobbled since 2003
10. Slide 14: 15 million tracks scrobbled / day (175 scrobbles per second) Over 6 billion tracks scrobbled since 2003 10 million artists
11. Slide 15: 15 million tracks scrobbled / day (175 scrobbles per second) Over 6 billion tracks scrobbled since 2003 10 million artists 70 million tracks
12. Slide 16: 15 million tracks scrobbled / day (175 scrobbles per second) Over 6 billion tracks scrobbled since 2003 10 million artists 70 million tracks 700k tracks streamable on Last.fm radio
13. Slide 17: 15 million tracks scrobbled / day (175 scrobbles per second) Over 6 billion tracks scrobbled since 2003 10 million artists 70 million tracks 700k tracks streamable on Last.fm radio 17 million items tagged
14. Slide 18: 15 million tracks scrobbled / day (175 scrobbles per second) Over 6 billion tracks scrobbled since 2003 10 million artists 70 million tracks 700k tracks streamable on Last.fm radio 17 million items tagged 145k artist wikis
15. Slide 19: Last.fm World Domination Headquarters Whitechapel, London, 2004
16. Slide 20: Early Growth Lessons
17. Slide 21: Early Growth Lessons Don’t overextend – scale with your growth, not before
18. Slide 22: Early Growth Lessons Don’t overextend – scale with your growth, not before Make sure revenue sources scale with increased usage
19. Slide 23: Early Growth Lessons Don’t overextend – scale with your growth, not before Make sure revenue sources scale with increased usage Google ads
20. Slide 24: Early Growth Lessons Don’t overextend – scale with your growth, not before Make sure revenue sources scale with increased usage Google ads User subscriptions
21. Slide 25: Early Growth Lessons Don’t overextend – scale with your growth, not before Make sure revenue sources scale with increased usage Google ads User subscriptions Involve users in your web application’s story
22. Slide 26: Early Growth Lessons Don’t overextend – scale with your growth, not before Make sure revenue sources scale with increased usage Google ads User subscriptions Involve users in your web application’s story Make growth a selfish aim for existing users
23. Slide 27: Early Growth Lessons Don’t overextend – scale with your growth, not before Make sure revenue sources scale with increased usage Google ads User subscriptions Involve users in your web application’s story Make growth a selfish aim for existing users Be as open as you can afford
24. Slide 28: Audioscrobbler.com homepage, late 2004
25. Slide 29: Openness and Growth
26. Slide 30: Openness and Growth In 2004, we had a cool service, but needed data
27. Slide 31: Openness and Growth In 2004, we had a cool service, but needed data Audioscrobbler Protocol 1.0
28. Slide 32: Openness and Growth In 2004, we had a cool service, but needed data Audioscrobbler Protocol 1.0 Webservices for accessing users’ music profile data
29. Slide 33: Openness and Growth In 2004, we had a cool service, but needed data Audioscrobbler Protocol 1.0 Webservices for accessing users’ music profile data Together, provided “round trip” incentive for plugin developers
30. Slide 34: Openness and Growth In 2004, we had a cool service, but needed data Audioscrobbler Protocol 1.0 Webservices for accessing users’ music profile data Together, provided “round trip” incentive for plugin developers Winamp, iTunes, WMP, Amarok
31. Slide 35: Openness and Growth In 2004, we had a cool service, but needed data Audioscrobbler Protocol 1.0 Webservices for accessing users’ music profile data Together, provided “round trip” incentive for plugin developers Winamp, iTunes, WMP, Amarok Foobar, QCD, mpd, XMMS, mpg321
32. Slide 36: Openness and Growth In 2004, we had a cool service, but needed data Audioscrobbler Protocol 1.0 Webservices for accessing users’ music profile data Together, provided “round trip” incentive for plugin developers Winamp, iTunes, WMP, Amarok Foobar, QCD, mpd, XMMS, mpg321 Noatun, Quod Libet, Jinzora, Ampache
33. Slide 37: Openness and Growth In 2004, we had a cool service, but needed data Audioscrobbler Protocol 1.0 Webservices for accessing users’ music profile data Together, provided “round trip” incentive for plugin developers Winamp, iTunes, WMP, Amarok Foobar, QCD, mpd, XMMS, mpg321 Noatun, Quod Libet, Jinzora, Ampache (and dozens more)
34. Slide 38: Openness and Users
35. Slide 39: Openness and Users Involve non-technical users as well
36. Slide 40: Openness and Users Involve non-technical users as well Promote a community around your application
37. Slide 41: Openness and Users Involve non-technical users as well Promote a community around your application Talk to your users (bad news > no news)
38. Slide 42: Openness and Users Involve non-technical users as well Promote a community around your application Talk to your users (bad news > no news) More likely to tolerate growing pains…
39. Slide 47: Moving into the new office Old Street, London, May 2006
40. Slide 48: Growing up
41. Slide 49: Growing up process
42. Slide 50: Growing up process product
43. Slide 51: Growing up process product
44. Slide 52: Growing up • process 20 to 40+ employees in 6 months product
45. Slide 53: Growing up • process 20 to 40+ employees in 6 months • “people trump process” product
46. Slide 54: Growing up • process 20 to 40+ employees in 6 months • “people trump process” product • simple tools are effective tools
47. Slide 55: Growing up • process 20 to 40+ employees in 6 months • “people trump process” product • simple tools are effective tools • radiate information across different channels in the company
48. Slide 56: Growing up process excerpt from #last.fm irc channel
49. Slide 57: Growing up process excerpt from #last.fm irc channel irccat: SVN commit by norman (23872) 'randomSplitter: splits data into train and test sets randomly' (changeset: https://admindev.last.fm/ trac/changeset/23872)
50. Slide 58: Growing up process excerpt from #last.fm irc channel irccat: SVN commit by norman (23872) 'randomSplitter: splits data into train and test sets randomly' (changeset: https://admindev.last.fm/ trac/changeset/23872) 3:40 PM
51. Slide 59: Growing up process excerpt from #last.fm irc channel irccat: SVN commit by norman (23872) 'randomSplitter: splits data into train and test sets randomly' (changeset: https://admindev.last.fm/ trac/changeset/23872) 3:40 PM irccat: *** jonty is refreshing webnodes now: 'Fix for group owners'
52. Slide 60: Growing up process excerpt from #last.fm irc channel irccat: SVN commit by norman (23872) 'randomSplitter: splits data into train and test sets randomly' (changeset: https://admindev.last.fm/ trac/changeset/23872) 3:40 PM irccat: *** jonty is refreshing webnodes now: 'Fix for group owners' mischa: jonty: memcache key should be set to: java-playlist-10093 where 10093 ==userid.
53. Slide 61: Growing up process excerpt from #last.fm irc channel irccat: SVN commit by norman (23872) 'randomSplitter: splits data into train and test sets randomly' (changeset: https://admindev.last.fm/ trac/changeset/23872) 3:40 PM irccat: *** jonty is refreshing webnodes now: 'Fix for group owners' mischa: jonty: memcache key should be set to: java-playlist-10093 where 10093 ==userid. irccat: Trac: ticket #1779 (http://support.last.fm/trac/ticket/1779) changed by julian, Comment: Fixed for the next release..
54. Slide 62: Growing up process excerpt from #last.fm irc channel irccat: SVN commit by norman (23872) 'randomSplitter: splits data into train and test sets randomly' (changeset: https://admindev.last.fm/ trac/changeset/23872) 3:40 PM irccat: *** jonty is refreshing webnodes now: 'Fix for group owners' mischa: jonty: memcache key should be set to: java-playlist-10093 where 10093 ==userid. irccat: Trac: ticket #1779 (http://support.last.fm/trac/ticket/1779) changed by julian, Comment: Fixed for the next release.. irccat: number of anon flash streams is 1440, number of registered flash streams is 764
55. Slide 63: Growing up process excerpt from #last.fm irc channel irccat: SVN commit by norman (23872) 'randomSplitter: splits data into train and test sets randomly' (changeset: https://admindev.last.fm/ trac/changeset/23872) 3:40 PM irccat: *** jonty is refreshing webnodes now: 'Fix for group owners' mischa: jonty: memcache key should be set to: java-playlist-10093 where 10093 ==userid. irccat: Trac: ticket #1779 (http://support.last.fm/trac/ticket/1779) changed by julian, Comment: Fixed for the next release.. irccat: number of anon flash streams is 1440, number of registered flash streams is 764 felix: hey abc when did you put the adsense leaderboard on bottom cat pages live?
56. Slide 64: Growing up process excerpt from #last.fm irc channel irccat: SVN commit by norman (23872) 'randomSplitter: splits data into train and test sets randomly' (changeset: https://admindev.last.fm/ trac/changeset/23872) 3:40 PM irccat: *** jonty is refreshing webnodes now: 'Fix for group owners' mischa: jonty: memcache key should be set to: java-playlist-10093 where 10093 ==userid. irccat: Trac: ticket #1779 (http://support.last.fm/trac/ticket/1779) changed by julian, Comment: Fixed for the next release.. irccat: number of anon flash streams is 1440, number of registered flash streams is 764 felix: hey abc when did you put the adsense leaderboard on bottom cat pages live? abc: friday
57. Slide 65: Growing up process excerpt from #last.fm irc channel irccat: SVN commit by norman (23872) 'randomSplitter: splits data into train and test sets randomly' (changeset: https://admindev.last.fm/ trac/changeset/23872) 3:40 PM irccat: *** jonty is refreshing webnodes now: 'Fix for group owners' mischa: jonty: memcache key should be set to: java-playlist-10093 where 10093 ==userid. irccat: Trac: ticket #1779 (http://support.last.fm/trac/ticket/1779) changed by julian, Comment: Fixed for the next release.. irccat: number of anon flash streams is 1440, number of registered flash streams is 764 felix: hey abc when did you put the adsense leaderboard on bottom cat pages live? abc: friday 3:45 PM
58. Slide 66: Growing up process excerpt from #last.fm irc channel irccat: SVN commit by norman (23872) 'randomSplitter: splits data into train and test sets randomly' (changeset: https://admindev.last.fm/ trac/changeset/23872) 3:40 PM irccat: *** jonty is refreshing webnodes now: 'Fix for group owners' mischa: jonty: memcache key should be set to: java-playlist-10093 where 10093 ==userid. irccat: Trac: ticket #1779 (http://support.last.fm/trac/ticket/1779) changed by julian, Comment: Fixed for the next release.. irccat: number of anon flash streams is 1440, number of registered flash streams is 764 felix: hey abc when did you put the adsense leaderboard on bottom cat pages live? abc: friday 3:45 PM mokele: ? lookup track 11082618
59. Slide 67: Growing up process excerpt from #last.fm irc channel irccat: SVN commit by norman (23872) 'randomSplitter: splits data into train and test sets randomly' (changeset: https://admindev.last.fm/ trac/changeset/23872) 3:40 PM irccat: *** jonty is refreshing webnodes now: 'Fix for group owners' mischa: jonty: memcache key should be set to: java-playlist-10093 where 10093 ==userid. irccat: Trac: ticket #1779 (http://support.last.fm/trac/ticket/1779) changed by julian, Comment: Fixed for the next release.. irccat: number of anon flash streams is 1440, number of registered flash streams is 764 felix: hey abc when did you put the adsense leaderboard on bottom cat pages live? abc: friday 3:45 PM mokele: ? lookup track 11082618 irccat: track.id(11082618) Zetan Spore ??? Subspace Distortion http://www.last.fm/music/Zetan+Spore/_/Subspace+Distortion (lastfm=t)
60. Slide 68: Growing up process excerpt from #last.fm irc channel irccat: SVN commit by norman (23872) 'randomSplitter: splits data into train and test sets randomly' (changeset: https://admindev.last.fm/ trac/changeset/23872) 3:40 PM irccat: *** jonty is refreshing webnodes now: 'Fix for group owners' mischa: jonty: memcache key should be set to: java-playlist-10093 where 10093 ==userid. irccat: Trac: ticket #1779 (http://support.last.fm/trac/ticket/1779) changed by julian, Comment: Fixed for the next release.. irccat: number of anon flash streams is 1440, number of registered flash streams is 764 felix: hey abc when did you put the adsense leaderboard on bottom cat pages live? abc: friday 3:45 PM mokele: ? lookup track 11082618 irccat: track.id(11082618) Zetan Spore ??? Subspace Distortion http://www.last.fm/music/Zetan+Spore/_/Subspace+Distortion (lastfm=t) pete_bug: jonty, can you please suspend PP campaign 3670?
61. Slide 69: Growing up process excerpt from #last.fm irc channel irccat: SVN commit by norman (23872) 'randomSplitter: splits data into train and test sets randomly' (changeset: https://admindev.last.fm/ trac/changeset/23872) 3:40 PM irccat: *** jonty is refreshing webnodes now: 'Fix for group owners' mischa: jonty: memcache key should be set to: java-playlist-10093 where 10093 ==userid. irccat: Trac: ticket #1779 (http://support.last.fm/trac/ticket/1779) changed by julian, Comment: Fixed for the next release.. irccat: number of anon flash streams is 1440, number of registered flash streams is 764 felix: hey abc when did you put the adsense leaderboard on bottom cat pages live? abc: friday 3:45 PM mokele: ? lookup track 11082618 irccat: track.id(11082618) Zetan Spore ??? Subspace Distortion http://www.last.fm/music/Zetan+Spore/_/Subspace+Distortion (lastfm=t) pete_bug: jonty, can you please suspend PP campaign 3670? jonty: pete_bug, sure one sec.
62. Slide 70: Growing up process excerpt from #last.fm irc channel irccat: SVN commit by norman (23872) 'randomSplitter: splits data into train and test sets randomly' (changeset: https://admindev.last.fm/ trac/changeset/23872) 3:40 PM irccat: *** jonty is refreshing webnodes now: 'Fix for group owners' mischa: jonty: memcache key should be set to: java-playlist-10093 where 10093 ==userid. irccat: Trac: ticket #1779 (http://support.last.fm/trac/ticket/1779) changed by julian, Comment: Fixed for the next release.. irccat: number of anon flash streams is 1440, number of registered flash streams is 764 felix: hey abc when did you put the adsense leaderboard on bottom cat pages live? abc: friday 3:45 PM mokele: ? lookup track 11082618 irccat: track.id(11082618) Zetan Spore ??? Subspace Distortion http://www.last.fm/music/Zetan+Spore/_/Subspace+Distortion (lastfm=t) pete_bug: jonty, can you please suspend PP campaign 3670? jonty: pete_bug, sure one sec. irccat: Trac: ticket #1526 (http://support.last.fm/trac/ticket/1526) "group recommendations are slow" created by muz.
63. Slide 71: Growing up process excerpt from #last.fm irc channel osmotic communication
64. Slide 72: Growing up process product
65. Slide 73: Growing up • process 15 million unique users a month product
66. Slide 74: Growing up • process 15 million unique users a month • product plan for going global
67. Slide 75: Growing up • process 15 million unique users a month • product plan for going global • affiliations should enhance the user experience
68. Slide 76: Growing up • process 15 million unique users a month • product plan for going global • affiliations should enhance the user experience • harness critical mass
69. Slide 77: Growing up • process 15 million unique users a month • product plan for going global • affiliations should enhance the user experience • harness critical mass • embed your service in others
70. Slide 78: Growing up • process 15 million unique users a month • product plan for going global • affiliations should enhance the user experience • harness critical mass • embed your service in others • make the transition from service to platform
71. Slide 79: Growing up product harness critical mass
72. Slide 80: Growing up product harness critical mass
73. Slide 81: Growing up product harness critical mass user-generated content
74. Slide 82: Growing up product harness critical mass network effects of socialization user-generated content
75. Slide 83: Growing up product harness critical mass network effects of socialization affiliate services user-generated content
76. Slide 84: Growing up product harness critical mass network effects of socialization user moderation affiliate services user-generated content
77. Slide 85: Growing up product harness critical mass user-generated content drives feature design
78. Slide 86: Growing up product
79. Slide 87: Growing up product embed your service in others
80. Slide 88: Growing up product embed your service in others quilts
81. Slide 89: Growing up product embed your service in others quilts charts
82. Slide 90: Growing up product embed your service in others quilts charts radio
83. Slide 91: Growing up product embed your service in others allow users to export
84. Slide 92: Growing up product
85. Slide 93: Growing up product transition from service to platform
86. Slide 94: Growing up product transition from service to platform
87. Slide 95: Growing up product transition from service to platform
88. Slide 96: Growing up product transition from service to platform
89. Slide 97: Growing up product transition from service to platform openness is key to web platforms
90. Slide 98: Attention and “Myware”
91. Slide 99: Attention and “Myware”
92. Slide 100: Attention and “Myware” When you pay attention to something (or don’t), data is created
93. Slide 101: Attention and “Myware” When you pay attention to something (or don’t), data is created Myware: spying on yourself
94. Slide 102: Attention and “Myware” When you pay attention to something (or don’t), data is created Myware: spying on yourself Attention economy alters traditional meaning of “active user”
95. Slide 103: Attention and “Myware” When you pay attention to something (or don’t), data is created Myware: spying on yourself Attention economy alters traditional meaning of “active user” Social web + attention aggregators = where the action is
96. Slide 106: Attention data Monetizing attention
97. Slide 107: Attention data Monetizing attention 1 - Microchunk it - Reduce the content to its simplest form.
98. Slide 108: Attention data Monetizing attention 1 - Microchunk it - Reduce the content to its simplest form. 2 - Free it - Put it out there without walls around it or strings on it.
99. Slide 109: Attention data Monetizing attention 1 - Microchunk it - Reduce the content to its simplest form. 2 - Free it - Put it out there without walls around it or strings on it. 3 - Syndicate it - Let anyone take it and run with it.
100. Slide 110: Attention data Monetizing attention 1 - Microchunk it - Reduce the content to its simplest form. 2 - Free it - Put it out there without walls around it or strings on it. 3 - Syndicate it - Let anyone take it and run with it. 4 - Monetize it - Put the monetization and tracking systems into the microchunk.
101. Slide 111: Attention data Monetizing attention 1 - Microchunk it - Reduce the content to its simplest form. 2 - Free it - Put it out there without walls around it or strings on it. 3 - Syndicate it - Let anyone take it and run with it. 4 - Monetize it - Put the monetization and tracking systems into the microchunk. Fred Wilson, The Future of Media, November 2005
102. Slide 112: Attention data Monetizing attention
103. Slide 113: Attention data Monetizing attention Powerplay is sponsored airtime
104. Slide 114: Attention data Monetizing attention Powerplay is sponsored airtime Individualized targeting based on attention history
105. Slide 115: Attention data Monetizing attention Powerplay is sponsored airtime Individualized targeting based on attention history New attention metrics for reporting
106. Slide 116: Attention data Monetizing attention Powerplay is sponsored airtime Individualized targeting based on attention history New attention metrics for reporting No more CPMs - the ‘scrobble’ is our attention unit
107. Slide 117: Attention data Case study: tag cloud moderation
108. Slide 118: Attention data Case study: tag cloud moderation Paris Hilton tag cloud, October 2006
109. Slide 119: Attention data Case study: tag cloud moderation Paris Hilton tag cloud, October 2006 Paris Hilton tag cloud, February 2007
110. Slide 120: Attention data Case study: tag cloud moderation
111. Slide 121: Attention data Case study: tag cloud moderation Censorship is not acceptable in a folksonomy
112. Slide 122: Attention data Case study: tag cloud moderation Censorship is not acceptable in a folksonomy Attention data does not lie
113. Slide 123: Attention data Case study: tag cloud moderation Censorship is not acceptable in a folksonomy Attention data does not lie Weigh user tags by volume of attention user gives to content
114. Slide 124: Attention data Case study: tag cloud moderation Censorship is not acceptable in a folksonomy Attention data does not lie Weigh user tags by volume of attention user gives to content Attention earns trust
115. Slide 125: Future of Last.fm
116. Slide 126: Future of Last.fm more
117. Slide 127: Future of Last.fm more less
118. Slide 128: Future of Last.fm more less
119. Slide 129: Future of Last.fm • more more growth less
120. Slide 130: Future of Last.fm • more more growth • more streamable music less
121. Slide 131: Future of Last.fm • more more growth • more streamable music less • more ambient findability
122. Slide 132: Future of Last.fm • more more growth • more streamable music less • more ambient findability • more personalization / things you can do with your data
123. Slide 133: Future of Last.fm more less
124. Slide 134: Future of Last.fm • more fewer interfaces less
125. Slide 135: Future of Last.fm • more fewer interfaces • fewer barriers to entry less
126. Slide 136: Future of Last.fm • more fewer interfaces • fewer barriers to entry less • fewer gradients?
127. Slide 137: Thank-you. Questions? matt@last.fm anil@last.fm