Ticketmaster datascience

1.
Fans just wannahave fun! @petricek

2.
Every year: hundreds ofmillions tickets sold worth Billions of Dollars

9.
! Recommendations!

10.
! Recommendations! ! Who are myfans? !

11.
! Recommendations! ! Who are myfans? !

12.
! Tickets to theFans! (not scalpers)! ! Recommendations! ! Who are my fans? !

13.

14.
Quality evaluation Data collection RecommendationStrategies

15.

19.
1px 1px“pixel”

20.
Javascript request URL:http://pixelserver:8080/pixel? evt=page_load imp=e4beb325-310e-493a-88cd-3da709ef509a impp=2d60a5ee-ac02-46df-8d0e-fbbe976361ef bid=bj6LvMJg-NnIOXp7tScUAHp-vcLsATCZoDvp08idDaIZo-4Iok1ppkDXZvA3RUWwoSV9IiRB pid=lILEaTVwqW8iwByWQzh0Iuk3w4kxXAhBzoFx2-PBdBriuqOKeRousDZgIgrVZQ mid=lILEaTVwqW8iwByWQzh0Iuk3w4kxXAhBzoFx2-PBdBriuqOKeRousDZgIgrVZQ sid=lS9oSQ72XqeVxPX0axzEjwFMJQ5RfAbYOdA5G_IssvgmOl4ﬁQHQ5gO0Lg_FDQNZ5hirrvm pgt=Artist%3A+On+Sale%3A+List dmn=TM_US aid=1687536 mkt=35 tsu=1433631937802 ref=http%3A%2F%2Fwww.ticketmaster.com%2F pgn=1 tml=tm_homeA_b_10001_1 fbt=not_authorized 1px 1px“pixel”

21.
Javascript request URL:http://pixelserver:8080/pixel? evt=page_load imp=e4beb325-310e-493a-88cd-3da709ef509a impp=2d60a5ee-ac02-46df-8d0e-fbbe976361ef bid=bj6LvMJg-NnIOXp7tScUAHp-vcLsATCZoDvp08idDaIZo-4Iok1ppkDXZvA3RUWwoSV9IiRB pid=lILEaTVwqW8iwByWQzh0Iuk3w4kxXAhBzoFx2-PBdBriuqOKeRousDZgIgrVZQ mid=lILEaTVwqW8iwByWQzh0Iuk3w4kxXAhBzoFx2-PBdBriuqOKeRousDZgIgrVZQ sid=lS9oSQ72XqeVxPX0axzEjwFMJQ5RfAbYOdA5G_IssvgmOl4ﬁQHQ5gO0Lg_FDQNZ5hirrvm pgt=Artist%3A+On+Sale%3A+List dmn=TM_US aid=1687536 mkt=35 tsu=1433631937802 ref=http%3A%2F%2Fwww.ticketmaster.com%2F pgn=1 tml=tm_homeA_b_10001_1 fbt=not_authorized OfMonsters and Men 1px 1px“pixel”

24.
http://pixelserver/api/pixel?…

25.

26.

27.

28.
Number of requestsover time

29.
Number of requestsover time something big went on sale

30.
Number of requestsover time something big went on sale ~10 x

31.

32.

33.
Ticket Alert

40.
~100m users tens of thousands artists

41.
~100m users tens of thousands artists

42.
~100m users thousands of artists

43.
~100m users thousands of artists Availability Recommendability

45.
Online Recommendations

46.
Top Sellers Ofﬂine aggregateover many days

47.
Hot right now 30min slidingwindow

48.
Hot right now 30min slidingwindow count rank “hot right now”ﬁlter

49.
After viewing fans eventuallybought

50.
Personalized recommendations

51.
100m users tens of thousands artists

52.

53.

54.

55.

56.
100m users Recommender! Service

57.
Service endpoints Public endpoints

58.
Ticket! Availability ! Service Event Feed Recommender! Service Pixel! Service 100msSLA async Storm

61.
$ docker-compose up! :-)

62.

63.

64.

65.
Quality

66.
Expert feedback

71.
Availability

72.
A/B testing

73.
Stealth testing

74.
hdfs bolt

75.
hdfs bolttee bolt Recommender! service replay! bolt hdfsbolt

76.
hdfs bolttee bolt Recommender! service replay! bolt hdfsbolt

77.
recA (live) clickA impA

78.
recA (live) clickA impA CTR recA 0 0.25 0.5 0.75 1

79.
recB recA (live) clickA impA CTR recA 0 0.25 0.5 0.75 1

80.
recB recA (live) clickA impA CTR recA 0 0.25 0.5 0.75 1

81.
recB recA (live) clickA impA CTR recA 0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 1 recB

82.

83.
Contextual bandits Next steps

84.
Contextual bandits Exploration vExploitation Next steps

85.
Contextual bandits Exploration vExploitation More context (user, artist, venue, event metadata) Next steps

86.

87.

93.
$$$ $$$ $ $

94.
$$$ $$$ $ $

95.
$$$ $$$ $ $

96.
$$$ $$$ $ $

97.
Counting unique users

98.
Counting unique users O(N*log(N))

99.
Counting unique users O(N*log(N)) O(N)

100.
Counting unique users O(N*log(N)) O(N) HyperLogLog

101.
Counting unique users O(N*log(N)) O(N) HyperLogLog O(N)

102.
HLL $$$HLL HLL

103.
HLL $$$HLL HLL

104.
HLL $$$HLL HLL HLL HLL HLL HLL HLL HLL HLL HLL HLL HLL HLL map HLL

105.
HLL $$$HLL + + + reduce HLL HLL HLL HLL HLL HLL HLL HLL HLL HLL HLL HLL map HLL

106.
Unique female users —————————— Uniqueusers total Unique users with family —————————— Unique users total Unique high income users ——————————— Unique users total $$$HLL = HLL HLL HLL HLL = = HLL

107.
SF Philharmonic Phish Meshuggagh BackstreetBoys

108.
Chvrchs 15% Asian

109.
Chvrchs 15% AsianK. Michelle 57% African American

110.
Chvrchs 15% AsianK. Michelle 57% African American Ramon Ayala 92% Hispanic

111.
Chvrchs 15% AsianK. Michelle 57% African American Ramon Ayala 92% Hispanic Tom Petty 96% Caucasian

112.
Millennials Grits andBiscuits Gen X Depeche Mode Baby Boomers Chicago Musical

121.
Tens of thousands models Next

122.

123.

134.
0 500 1000 Price

135.
0 500 1000 Price

137.
|+

138.
|+

141.
vowpal wabbit JNI Storm

142.

143.

144.

145.

146.

147.

148.

149.
vw vw vw vw ?

150.
vw vw vw vw ?

151.
vw vw vw vw ?

152.
vw vw vw vw ?

153.
vw vw vw vw ?

154.
In-cart conversion 4 x

155.
Arms Race Next steps

156.

157.

158.

159.
Now what?

160.
Decades of granular and longitudinaldata Now what?

161.
Recommend the best Seat foryou

162.
linkedin.com/in/petricek

Ticketmaster datascience

More Related Content

Similar to Ticketmaster datascience

Recently uploaded

Ticketmaster datascience