personalizing broadcast radio with musicovery radio 2.0 webinar 2016-03-03

Post on 19-Feb-2017

2.239 Views

Category:

Entertainment & Humor

0 Downloads

Preview:

Click to see full reader

TRANSCRIPT

musicovery.com

09/10/2015

Personalizing broadcast radio

03/03/2016

Presented by Vincent Castaignet, Musicovery CEO

musicovery.com

2

Speaker

Personalizing broadcast radio Realizing the full value potential of

behavioural data of your listeners by

providing personalized experiences

Vincent Castaignet

Musicovery, CEO

musicovery.com

3

• How to create smart radios ? • How to built personalized YouTube channels ? • How to substitute a song that a listener skips for one that he will like ? • How to generate notifications of personalized and geolocalized concerts ? • How to identify for an emerging artist most receptive listeners ? • How to send personalized notifications of new releases/hot news ? • How to built prize competiton/loyalty programmes, with ultra-targeted gifts

(concerts, collector editions, goodies) ?

Personalizing broadcast radio

musicovery.com

4

Appetite for

discovery Appetite for

popular songs

Appetite for

new releases Needs for

repetition

Appetite for

genre diversity

Preferences

for songs,

artists, genres

Saturated

songs Current interest for

artists, genres

Mood

diversity

Smart radios adapted to each user profile

musicovery.com

5

Karl Böhm Claudio Abbado Eugen Jochum Riccardo Muti

• Recommendation of an artist: Karajan

• Selection of a video from the artist

• Proposition of alternative videos/tags/playlist

• Channel adapts according to interactions

Personalized YouTube channels

musicovery.com

6

Song substitution

musicovery.com

7

Personalized and geolocalized concerts

musicovery.com

8

likely to like

very likely to

like

fans

Finding most receptive listeners for an

emerging artist

musicovery.com

9

Notification of new releases

musicovery.com

10

Prize competition, loyalty programmes

musicovery.com

11

How does it work ?

User dynamic music profile

musicovery.com

12

I don’t

like I know

My

tribe

Music I

don’t know

I like

Influencers

I trust

My

friends

mood,

activities

my habits :

when, where

Expanding listeners music universe

musicovery.com

13

Not familiar

with this

song/artist

I can say if it

fits my music

universe

I like it

Optimal number of

plays

I like but I’m

saturated now

The right repetition rate

musicovery.com

14

listener

filters by genres, years, popularity

artist song

mood

genre

search

interacts with propositions

generates behavioural data

Behavioural data collection opportunities

musicovery.com

15

Analytics tool :

• Measuring quality objectively :

playlists duration, skip rates, …

• Running A/Z tests for various

scenarios to generate playlists

• Selection of the best scenario to

each listener once minimum

confidence is reached

Optimizing recommendation generation

musicovery.com

16

Behavioural

data

Exploration of

the music

universe

User dynamic

music profile

Personalisation: Musicovery robust model

musicovery.com

17

Personalized

services

Behavioural

data

collection

The virtuous circle of personalization

musicovery.com

18

Introduction to Musicovery API

• generates all kinds of playlists: Mood, activities, artist, genre, song, theme…

• generates all types of recommendations: Similar songs and artists, for a specific listener

• various parameters to finetune

playlists/recommendations

API documentation

musicovery.com

19

Recommendations for a specific listener

• Parameters to call the API : &userid=xxx

• Response : a list of recommended songs

artists

genres

musicovery.com

20

A very fast personalization process

focused on big

beat genre

mainstream

Listener 2 favorite artists:

Daft Punk, David Guetta

Listener 1 favorite artists:

Daft Punk, Propellerheads

musicovery.com

24

Iterative process

Test Musicovery

API

Design a project

Develop a prototype

Develop a pilote

Roll out

musicovery.com

25

Case study :

Broadcast : leader in the US, 238 M audience radios/TVs

Smart radio IHeartRadio • Launched 2008

• True challenger to Pandora in the US

• Designed to complement their 750 local FM stations

• provides similar artist data

• Similarity type optimized : genre/era/social

musicovery.com

26

Thanks for your attention

Test the API on musicovery.com: access is open

and free !

vincent.castaignet@musicovery.com - +33 6 13 17 35 61

musicovery.com

27

Source of data

Data generated

automatically (from

audio and web)

Data produced and

structured by experts

SmartPlayTM

Users behaviour

and preferences

musicovery.com

28

Technology

Acoustic

descriptors

Semantic

descriptors

Personalized recommendations

Social

activity

Charts &

trends

orchestration,

rhythmic…

situations,

mood, genres,

themes

Individual

personalization

profiles, tribes,

friends… by country, real

time

preferences,

behavior (time,

location), history

Algorithm

musicovery.com

29

A global and a local tool

• SmartPlay is now available in : US, Canada, UK,

France, Italy, Spain, Germany, Portugal, Norway.

=> Additional territories on request.

• SmartPlay is built to adapt to local specifics :

o It includes local artists for each country

o Genres, themes, mood are catered to local

language(s) and the local catalogue

musicovery.com

30

Music catalogue

• Comprehensive and exhaustive for

Europe, North America and South

America

o all genres and all times

o 290,000 artists

o 1,5M songs

o Charts from 1960: US, Canada,

UK, France, Spain, Italy, Portugal,

Germany and Norway

• Continually growing

musicovery.com

31

Musicovery engine efficiency

= total listening duration : +35%

Numerous and

diverse ways to launch

playlists

Recommendations more consistent

with listener profile and listening

history

More playlists are launched playlists are played longer

Personalized playlists

playlists are more

relevant and consistent

+

musicovery.com

32

Professional SLAs

• SmartPlay runs on professional infrastructure :

o Front-end on Co-host on Bearstech

o Algorithm, webservices and API on Bearstech

o Back-end on Amazon Cloud Services

• SLAs:

o Guaranteed 99.9 % uptime by infrastructure suppliers

o Bandwidth: peak bandwidth at least 70% higher than past top

bandwidth

o Computing capacity: peak capacity at 70% past top

o Network provider: Availability 99% on a rolling 30 days

• Support:

o Customers (B2B partners and clients): 9-5 Monday to Friday

o Consumer (B2C users): 24 hour email support, answer in next

working day

musicovery.com

33

Playlist by genre

Get the results from the API

Tuning parameters to call the API :

&tag=techno

&tracksnumber=30

&listenercountry=fr

&popularitymin=25

&popularitymax=100

&yearmin=1992

&yearmax=2005

musicovery.com

34

Artists similar to

Astrud Gilberto

Get the results from the API

&artistmbid=bc710bcf-8815-42cf-bad2-3f1d12246aeb

&focusgenre=false

&focusera=true

&obscureartists=false

&popularitymin=25

&popularitymax=100

musicovery.com

35

Songs similar to

Rock the casbah

Get the results from the API

&id=66380

&similaritytype=100

&limitgenre=false

&listenercountry=fr

&popularitymin=50

&popularitymax=100

top related