Collaborative filtering in CRM_INMA_alt
1
OBJECTIVES AND METHOD
ONE EMAIL DOES NOT FIT ALL Our objective is to increase the value of email newsletters by making sure each recipient receives personalized recommendations of unread articles directly to the inbox.
TO MEET OUR OBJECTIVE WE HAVE DEVELOPED AN ADVANCED RECOMMENDATION FEATURE
The recommendation feature is based on collaborative filtering as method. ”Collaborative filtering is a method of making automatic predictions (filtering) about the interests of a user by collecting preferences or taste information from many users (collaborating)” For Schibsted this means that many subscribers read the same articles online (similar reading habit). There are however always articles that one individual have read that others with similar reading habits have not read. With the use of advanced algorithms such unread articles are pulled by the recommendation feature and further pushed to the right recipients. 4
THE RECOMMENDATION FEATURE PUBLISHES INDIVIDUAL RSS FEEDS ON THE WEB WITH UNREAD ARTICLES THAT SUBSCRIBERS WITH SIMILAR READING HABITS HAVE READ
This allows us to individually merge the recommendations into marketing emails based on an unique ID (SPID ID) on each user.
5
SO HOW EXACTLY DOES IT WORK?
6
WHEN SENDING A NEWSLETTER IN OUR ESP, ALL RECIPIENTS RECEIVES A NEWSLETTER WITH UNIQUE PERSONALIZED CONTENT
APP
A custom built app are put in the newsletter template in the email builder in our ESP. The content and design that should look the same for everyone (ie logo header) is put outside the app.
App settings define rss url. All urls are unique based on unique id. The ids are pulled from data object on customer level in ESP (from CRM system)
The send generate unique emails with personalized content for all recipients. The newsletters is nicely designed with CSS and html in the app settings.
7
RESULTS
8
COLLABORATIVE FILTERING DRIVES UP TO 121% BETTER ENGAGEMENT (CLICK RATE) COMPARED TO MANUALLY CURATED EDITORIAL NEWSLETTER Weekly A/B tests of manually curated (regular) and personalized (rss) newsletter: WEEK
REGULAR
RSS
REGULAR
RSS
REGULAR
SEND
SEND
OPEN RATE
OPEN RATE
CLICK RATE
RSS CLICK RATE
39 40 42 43 44 45 46 47 48
Preliminary results also show great numbers (54%) on article completion (read) after click (29% on regular).
9
COLLABORATIVE FILTERING IN EMAIL GROW DIGITAL READERSHIP AND ENGAGEMENT
With collaborative filtering in email our editorial newsletters has become much more engaging as our subscribers are encouraged to read as well as discover more of our great content. 10
WITH COLLABORATIVE FILTERING IN EMAIL SCHIBSTED IS TAKING PERSONALIZATION ONE STEP FURTHER COMPARED TO MOST OTHER PLAYERS IN THE MEDIA INDUSTRY
”By this development, (…) Schibsted get one step ahead of the competition in sending individual and relevant content to their subscribers.” Arve Warholm, Deloitte Digital
11
NEXT STEPS
12
WE WILL CONTINUE TO IMPROVE AND DEVELOP AS WELL AS LAUNCH THE RECOMMENDATION FEATURE ON ALL BRANDS IN 2017 1. First step was testing and proof of concept on Aftenposten
2. Next step is full launch on all Schibsted subscription media brands 13
INNOVATION APPEAR WHEN SMART PEOPLE ACROSS THE ORGANIZATION WORK TOGETHER This product shows what might come out of breaking down silos and working together across different units within the organization. THE TEAM (from left): Eirik Winsnes (Development Editor, Aftenposten), Espen Tandberg (Product Manager Aftenposten, Schibsted Product & Tech), Hans Martin Cramer (Product Manager, SMP Curate), Ellinor Sande (CRM Designer, Schibsted Norge Abonnementsmedia), Adam Marklund (baby on the job with his father), Arnbjørn J. S. Marklund (CRM Manager, Schibsted Norge Abonnementsmedia) 14
THANK YOU Category 10: Best idea to grow digital readership or engagement
Collaborative filtering in email Eirik Winsnes – Development Editor, Aftenposten, [email protected] Arnbjørn Marklund - CRM Manager, Schibsted, [email protected] 15
OBJECTIVES AND METHOD
ONE EMAIL DOES NOT FIT ALL Our objective is to increase the value of email newsletters by making sure each recipient receives personalized recommendations of unread articles directly to the inbox.
TO MEET OUR OBJECTIVE WE HAVE DEVELOPED AN ADVANCED RECOMMENDATION FEATURE
The recommendation feature is based on collaborative filtering as method. ”Collaborative filtering is a method of making automatic predictions (filtering) about the interests of a user by collecting preferences or taste information from many users (collaborating)” For Schibsted this means that many subscribers read the same articles online (similar reading habit). There are however always articles that one individual have read that others with similar reading habits have not read. With the use of advanced algorithms such unread articles are pulled by the recommendation feature and further pushed to the right recipients. 4
THE RECOMMENDATION FEATURE PUBLISHES INDIVIDUAL RSS FEEDS ON THE WEB WITH UNREAD ARTICLES THAT SUBSCRIBERS WITH SIMILAR READING HABITS HAVE READ
This allows us to individually merge the recommendations into marketing emails based on an unique ID (SPID ID) on each user.
5
SO HOW EXACTLY DOES IT WORK?
6
WHEN SENDING A NEWSLETTER IN OUR ESP, ALL RECIPIENTS RECEIVES A NEWSLETTER WITH UNIQUE PERSONALIZED CONTENT
APP
A custom built app are put in the newsletter template in the email builder in our ESP. The content and design that should look the same for everyone (ie logo header) is put outside the app.
App settings define rss url. All urls are unique based on unique id. The ids are pulled from data object on customer level in ESP (from CRM system)
The send generate unique emails with personalized content for all recipients. The newsletters is nicely designed with CSS and html in the app settings.
7
RESULTS
8
COLLABORATIVE FILTERING DRIVES UP TO 121% BETTER ENGAGEMENT (CLICK RATE) COMPARED TO MANUALLY CURATED EDITORIAL NEWSLETTER Weekly A/B tests of manually curated (regular) and personalized (rss) newsletter: WEEK
REGULAR
RSS
REGULAR
RSS
REGULAR
SEND
SEND
OPEN RATE
OPEN RATE
CLICK RATE
RSS CLICK RATE
39 40 42 43 44 45 46 47 48
Preliminary results also show great numbers (54%) on article completion (read) after click (29% on regular).
9
COLLABORATIVE FILTERING IN EMAIL GROW DIGITAL READERSHIP AND ENGAGEMENT
With collaborative filtering in email our editorial newsletters has become much more engaging as our subscribers are encouraged to read as well as discover more of our great content. 10
WITH COLLABORATIVE FILTERING IN EMAIL SCHIBSTED IS TAKING PERSONALIZATION ONE STEP FURTHER COMPARED TO MOST OTHER PLAYERS IN THE MEDIA INDUSTRY
”By this development, (…) Schibsted get one step ahead of the competition in sending individual and relevant content to their subscribers.” Arve Warholm, Deloitte Digital
11
NEXT STEPS
12
WE WILL CONTINUE TO IMPROVE AND DEVELOP AS WELL AS LAUNCH THE RECOMMENDATION FEATURE ON ALL BRANDS IN 2017 1. First step was testing and proof of concept on Aftenposten
2. Next step is full launch on all Schibsted subscription media brands 13
INNOVATION APPEAR WHEN SMART PEOPLE ACROSS THE ORGANIZATION WORK TOGETHER This product shows what might come out of breaking down silos and working together across different units within the organization. THE TEAM (from left): Eirik Winsnes (Development Editor, Aftenposten), Espen Tandberg (Product Manager Aftenposten, Schibsted Product & Tech), Hans Martin Cramer (Product Manager, SMP Curate), Ellinor Sande (CRM Designer, Schibsted Norge Abonnementsmedia), Adam Marklund (baby on the job with his father), Arnbjørn J. S. Marklund (CRM Manager, Schibsted Norge Abonnementsmedia) 14
THANK YOU Category 10: Best idea to grow digital readership or engagement
Collaborative filtering in email Eirik Winsnes – Development Editor, Aftenposten, [email protected] Arnbjørn Marklund - CRM Manager, Schibsted, [email protected] 15
