SlideShare a Scribd company logo
Where to Go on Your Next Trip?
Optimizing Travel Destination Based on User
Preferences
Julia Kiseleva, Melanie J.I. Mueller, Lucas Bernardi,
Chad Davis, Ivan Kovacek, Mats Stafseng Einarsen, Jaap
Kamps, Alexander Tuzhilin, Djoerd Hiemstra
SIGIR 2015,
Chile
Booking.com
Booking.com
• World largest online travel agent
• > 220 countries
• > 81.000 destinations
• > 710.000 bookable hotels worldwide
• > 30.000.000 unique users
• >> 100.000 unique visitors per day
Destination Finder
Destination Finder
Destination Finder
How Do We Get Endorsements?
• Only users who stayed at a hotel in a destination can
endorse it
• Free text endorsements since 2013
• Since 2014 free text endorsements standardized to 256
canonical tags
• Used NLP techniques to extract the canonical base
More Numbers:
• > 13.000.000 total unique endorsements
• > 60.000 destinations
Endorsement Histogram
Endorsement Histogram
Endorsement-Destination Histogram
Endorsement-Destination Histogram
Endorsements for Santiago
Destination Finder
Problem:
How to Optimize ranking of recommended
destinations?
Problem Setup
• Challenge:
o Recommender or Information Retrieval System?
• Information Retrieval:
o Users have information need which is expressed as a query
o System has to satisfy this information needs
• Recommender:
o System predicts what users might be interested
We have both:
1) Users search for activities
2) Users don’t know where they want to go
Why is It Hard?
Problem Characteristics:
• S - Sparsity: new or rare users/destinations
• V - Volatility: users’ interests/endorsements of destinations change over
time
• I - Identity: a failure to match data from the same users
• P - Personas: users have different interests at different, possibly closely
points in time
Continuous Cold Start Problem!
Ranking Destinations for ‘Beach’
• Keep it simple!!! We care about performance!
• Naïve Bayes:
P(Miami, Beach) = P(Miami) * P(Miami | Beach)
P(Miami | Beach) = # ‘Beach’ endorsements for Miami
# ‘Beach’ endorsements
What and How to Compare?
• Booking.com Baseline
• Random
• Most Popular Destination
• Naïve Bayes
Objection:
Increase User Engagement
(Clickers per SERP)
A/B Testing Setup
• 50/50 traffic split
• Experiments run for N full weeks according to desired
power and significance levels
• Hypothesis tests are performed according to targeted
metrics (G-test in our case)
A/B Testing
Version A Version B Version C
A/B Testing
Ranker # Users Engagement
Baseline 9928
Random 10079
Popularity 9838
Naïve Bayes 9895
Amount of
clickers
A/B Testing
Ranker # Users Engagement
Baseline 9928 25.61% +/- 0.72%
Random 10079 24.46% +/- 0.71%
Popularity 9838 25.50% +/- 0.73%
Naïve Bayes 9895 26.73% +/- 0.73%
A/B Testing
Ranker # Users Engagement
Baseline 9928 25.61% +/- 0.72%
Random 10079 24.46% +/- 0.71%
Popularity 9838 25.50% +/- 0.73%
Naïve Bayes 9895 26.73% +/- 0.73%
Conclusion and Future Work
• Interesting application
• Surprise 1: Keep simple baseline in production system
• Surprise 2: Random performed not bad => effect
serendipity
For the Future:
• Improve the ranking by taking contexts into account
Conclusion and Future Work
• Interesting application
• Surprise 1: Keep simple baseline in production system
• Surprise 2: Random performed not bad => effect
serendipity
For the Future:
• Improve the ranking by taking contexts into account

More Related Content

Similar to Where to Go on Your Next Trip? Optimizing Travel Destinations Based on User Preferences

Introduction to Paid Customer Acquisition
Introduction to Paid Customer AcquisitionIntroduction to Paid Customer Acquisition
Introduction to Paid Customer AcquisitionIntelligent_ly
 
Meta searching in hospitality
Meta searching in hospitalityMeta searching in hospitality
Meta searching in hospitalityhafemank5152
 
Reputation management (session 3) ppt
Reputation management (session 3) pptReputation management (session 3) ppt
Reputation management (session 3) pptAlan Stevens
 
2013 Local Search Rankings Las Vegas Michael Dorausch
2013 Local Search Rankings Las Vegas Michael Dorausch2013 Local Search Rankings Las Vegas Michael Dorausch
2013 Local Search Rankings Las Vegas Michael DorauschMichael Dorausch
 
My Travel Club - A Business Plan
My Travel Club - A Business PlanMy Travel Club - A Business Plan
My Travel Club - A Business PlanJon Hussey
 
Value Chain Analysis of Hotels.com and Booking.com
Value Chain Analysis of Hotels.com and Booking.comValue Chain Analysis of Hotels.com and Booking.com
Value Chain Analysis of Hotels.com and Booking.comcagankoc
 
CPTW - Trainer HQ - Google Marketing for Fitness Professionals
CPTW - Trainer HQ - Google Marketing for Fitness ProfessionalsCPTW - Trainer HQ - Google Marketing for Fitness Professionals
CPTW - Trainer HQ - Google Marketing for Fitness ProfessionalsAlexei Kouleshov
 
CRO Masterclass 21st August 2014
CRO Masterclass 21st August 2014 CRO Masterclass 21st August 2014
CRO Masterclass 21st August 2014 GlowMetrics
 
My Travel Club Business Plan
My Travel Club Business PlanMy Travel Club Business Plan
My Travel Club Business PlanJon Hussey
 
Travel club presentation_final
Travel club presentation_finalTravel club presentation_final
Travel club presentation_finalJon Hussey
 
How digital marketing will help you rebound from covid 19
How digital marketing will help you rebound from covid 19How digital marketing will help you rebound from covid 19
How digital marketing will help you rebound from covid 19RezStream
 
Making it Count: Paid Search
Making it Count: Paid SearchMaking it Count: Paid Search
Making it Count: Paid SearchE-Tourism Africa
 
Ross Plotkin - Take a Walk on the Paid Side - KahenaCon Spring 2012
Ross Plotkin - Take a Walk on the Paid Side - KahenaCon Spring 2012Ross Plotkin - Take a Walk on the Paid Side - KahenaCon Spring 2012
Ross Plotkin - Take a Walk on the Paid Side - KahenaCon Spring 2012Kahena Digital Marketing
 
Review Pro - Reputation Management for Secret Retreats
Review Pro - Reputation Management for Secret Retreats Review Pro - Reputation Management for Secret Retreats
Review Pro - Reputation Management for Secret Retreats www.bookgreener.com
 
WEBINAR VIDEO: Cornell/Revinate study of how to improve your TripAdvisor rank...
WEBINAR VIDEO: Cornell/Revinate study of how to improve your TripAdvisor rank...WEBINAR VIDEO: Cornell/Revinate study of how to improve your TripAdvisor rank...
WEBINAR VIDEO: Cornell/Revinate study of how to improve your TripAdvisor rank...tnooz
 
Optimizely Workshop: Personalization - Strategy Essentials
Optimizely Workshop: Personalization - Strategy EssentialsOptimizely Workshop: Personalization - Strategy Essentials
Optimizely Workshop: Personalization - Strategy EssentialsOptimizely
 
Online Marketing presentation
Online Marketing presentation Online Marketing presentation
Online Marketing presentation Sam shetty
 

Similar to Where to Go on Your Next Trip? Optimizing Travel Destinations Based on User Preferences (20)

Introduction to Paid Customer Acquisition
Introduction to Paid Customer AcquisitionIntroduction to Paid Customer Acquisition
Introduction to Paid Customer Acquisition
 
Meta searching in hospitality
Meta searching in hospitalityMeta searching in hospitality
Meta searching in hospitality
 
Reputation management (session 3) ppt
Reputation management (session 3) pptReputation management (session 3) ppt
Reputation management (session 3) ppt
 
2013 Local Search Rankings Las Vegas Michael Dorausch
2013 Local Search Rankings Las Vegas Michael Dorausch2013 Local Search Rankings Las Vegas Michael Dorausch
2013 Local Search Rankings Las Vegas Michael Dorausch
 
My Travel Club - A Business Plan
My Travel Club - A Business PlanMy Travel Club - A Business Plan
My Travel Club - A Business Plan
 
Value Chain Analysis of Hotels.com and Booking.com
Value Chain Analysis of Hotels.com and Booking.comValue Chain Analysis of Hotels.com and Booking.com
Value Chain Analysis of Hotels.com and Booking.com
 
CPTW - Trainer HQ - Google Marketing for Fitness Professionals
CPTW - Trainer HQ - Google Marketing for Fitness ProfessionalsCPTW - Trainer HQ - Google Marketing for Fitness Professionals
CPTW - Trainer HQ - Google Marketing for Fitness Professionals
 
CRO Masterclass 21st August 2014
CRO Masterclass 21st August 2014 CRO Masterclass 21st August 2014
CRO Masterclass 21st August 2014
 
My Travel Club Business Plan
My Travel Club Business PlanMy Travel Club Business Plan
My Travel Club Business Plan
 
Travel club presentation_final
Travel club presentation_finalTravel club presentation_final
Travel club presentation_final
 
How digital marketing will help you rebound from covid 19
How digital marketing will help you rebound from covid 19How digital marketing will help you rebound from covid 19
How digital marketing will help you rebound from covid 19
 
Making it Count: Paid Search
Making it Count: Paid SearchMaking it Count: Paid Search
Making it Count: Paid Search
 
Clicks2 Customers
Clicks2 CustomersClicks2 Customers
Clicks2 Customers
 
How destinations are attracting visitors with better guest intelligence | Jos...
How destinations are attracting visitors with better guest intelligence | Jos...How destinations are attracting visitors with better guest intelligence | Jos...
How destinations are attracting visitors with better guest intelligence | Jos...
 
Ross Plotkin - Take a Walk on the Paid Side - KahenaCon Spring 2012
Ross Plotkin - Take a Walk on the Paid Side - KahenaCon Spring 2012Ross Plotkin - Take a Walk on the Paid Side - KahenaCon Spring 2012
Ross Plotkin - Take a Walk on the Paid Side - KahenaCon Spring 2012
 
Review Pro - Reputation Management for Secret Retreats
Review Pro - Reputation Management for Secret Retreats Review Pro - Reputation Management for Secret Retreats
Review Pro - Reputation Management for Secret Retreats
 
WEBINAR VIDEO: Cornell/Revinate study of how to improve your TripAdvisor rank...
WEBINAR VIDEO: Cornell/Revinate study of how to improve your TripAdvisor rank...WEBINAR VIDEO: Cornell/Revinate study of how to improve your TripAdvisor rank...
WEBINAR VIDEO: Cornell/Revinate study of how to improve your TripAdvisor rank...
 
Seer_PPC 101
Seer_PPC 101Seer_PPC 101
Seer_PPC 101
 
Optimizely Workshop: Personalization - Strategy Essentials
Optimizely Workshop: Personalization - Strategy EssentialsOptimizely Workshop: Personalization - Strategy Essentials
Optimizely Workshop: Personalization - Strategy Essentials
 
Online Marketing presentation
Online Marketing presentation Online Marketing presentation
Online Marketing presentation
 

More from Julia Kiseleva

Evaluation Personal Assistants
Evaluation Personal AssistantsEvaluation Personal Assistants
Evaluation Personal AssistantsJulia Kiseleva
 
From queries to dialogues
From queries to dialoguesFrom queries to dialogues
From queries to dialoguesJulia Kiseleva
 
Predicting User Satisfaction with Intelligent Assistants
Predicting User Satisfaction with Intelligent AssistantsPredicting User Satisfaction with Intelligent Assistants
Predicting User Satisfaction with Intelligent AssistantsJulia Kiseleva
 
Using Contextual Information to Understand Searching and Browsing Behavior
Using Contextual Information to Understand Searching and Browsing BehaviorUsing Contextual Information to Understand Searching and Browsing Behavior
Using Contextual Information to Understand Searching and Browsing BehaviorJulia Kiseleva
 
Understanding and Predicting User Satisfaction with Intelligent Assistants
Understanding and Predicting User Satisfaction with Intelligent AssistantsUnderstanding and Predicting User Satisfaction with Intelligent Assistants
Understanding and Predicting User Satisfaction with Intelligent AssistantsJulia Kiseleva
 
Understanding User Satisfaction with Intelligent Assistants
Understanding User Satisfaction with Intelligent AssistantsUnderstanding User Satisfaction with Intelligent Assistants
Understanding User Satisfaction with Intelligent AssistantsJulia Kiseleva
 
Behavioral Dynamics from the SERP’s Perspective: What are Failed SERPs and Ho...
Behavioral Dynamics from the SERP’s Perspective: What are Failed SERPs and Ho...Behavioral Dynamics from the SERP’s Perspective: What are Failed SERPs and Ho...
Behavioral Dynamics from the SERP’s Perspective: What are Failed SERPs and Ho...Julia Kiseleva
 
Modelling and Detecting Changes in User Satisfaction
Modelling and Detecting Changes in User SatisfactionModelling and Detecting Changes in User Satisfaction
Modelling and Detecting Changes in User SatisfactionJulia Kiseleva
 
Predicting Current User Intent with Contextual Markov Models
Predicting Current User Intent with Contextual Markov ModelsPredicting Current User Intent with Contextual Markov Models
Predicting Current User Intent with Contextual Markov ModelsJulia Kiseleva
 
The talk at Twente University on 28 July 2014
The talk at Twente University on 28 July 2014 The talk at Twente University on 28 July 2014
The talk at Twente University on 28 July 2014 Julia Kiseleva
 
Discovering Temporal Hidden Contexts in Web Sessions for User Trail Prediction
Discovering Temporal Hidden Contexts in Web Sessions for User Trail PredictionDiscovering Temporal Hidden Contexts in Web Sessions for User Trail Prediction
Discovering Temporal Hidden Contexts in Web Sessions for User Trail PredictionJulia Kiseleva
 
Context Mining and Integration in Web Predictive Analytics
Context Mining and Integration in Web Predictive AnalyticsContext Mining and Integration in Web Predictive Analytics
Context Mining and Integration in Web Predictive AnalyticsJulia Kiseleva
 

More from Julia Kiseleva (12)

Evaluation Personal Assistants
Evaluation Personal AssistantsEvaluation Personal Assistants
Evaluation Personal Assistants
 
From queries to dialogues
From queries to dialoguesFrom queries to dialogues
From queries to dialogues
 
Predicting User Satisfaction with Intelligent Assistants
Predicting User Satisfaction with Intelligent AssistantsPredicting User Satisfaction with Intelligent Assistants
Predicting User Satisfaction with Intelligent Assistants
 
Using Contextual Information to Understand Searching and Browsing Behavior
Using Contextual Information to Understand Searching and Browsing BehaviorUsing Contextual Information to Understand Searching and Browsing Behavior
Using Contextual Information to Understand Searching and Browsing Behavior
 
Understanding and Predicting User Satisfaction with Intelligent Assistants
Understanding and Predicting User Satisfaction with Intelligent AssistantsUnderstanding and Predicting User Satisfaction with Intelligent Assistants
Understanding and Predicting User Satisfaction with Intelligent Assistants
 
Understanding User Satisfaction with Intelligent Assistants
Understanding User Satisfaction with Intelligent AssistantsUnderstanding User Satisfaction with Intelligent Assistants
Understanding User Satisfaction with Intelligent Assistants
 
Behavioral Dynamics from the SERP’s Perspective: What are Failed SERPs and Ho...
Behavioral Dynamics from the SERP’s Perspective: What are Failed SERPs and Ho...Behavioral Dynamics from the SERP’s Perspective: What are Failed SERPs and Ho...
Behavioral Dynamics from the SERP’s Perspective: What are Failed SERPs and Ho...
 
Modelling and Detecting Changes in User Satisfaction
Modelling and Detecting Changes in User SatisfactionModelling and Detecting Changes in User Satisfaction
Modelling and Detecting Changes in User Satisfaction
 
Predicting Current User Intent with Contextual Markov Models
Predicting Current User Intent with Contextual Markov ModelsPredicting Current User Intent with Contextual Markov Models
Predicting Current User Intent with Contextual Markov Models
 
The talk at Twente University on 28 July 2014
The talk at Twente University on 28 July 2014 The talk at Twente University on 28 July 2014
The talk at Twente University on 28 July 2014
 
Discovering Temporal Hidden Contexts in Web Sessions for User Trail Prediction
Discovering Temporal Hidden Contexts in Web Sessions for User Trail PredictionDiscovering Temporal Hidden Contexts in Web Sessions for User Trail Prediction
Discovering Temporal Hidden Contexts in Web Sessions for User Trail Prediction
 
Context Mining and Integration in Web Predictive Analytics
Context Mining and Integration in Web Predictive AnalyticsContext Mining and Integration in Web Predictive Analytics
Context Mining and Integration in Web Predictive Analytics
 

Recently uploaded

Bug Bounty Blueprint : A Beginner's Guide
Bug Bounty Blueprint : A Beginner's GuideBug Bounty Blueprint : A Beginner's Guide
Bug Bounty Blueprint : A Beginner's GuideVarun Mithran
 
Article writing on excessive use of internet.pptx
Article writing on excessive use of internet.pptxArticle writing on excessive use of internet.pptx
Article writing on excessive use of internet.pptxabhinandnam9997
 
Multi-cluster Kubernetes Networking- Patterns, Projects and Guidelines
Multi-cluster Kubernetes Networking- Patterns, Projects and GuidelinesMulti-cluster Kubernetes Networking- Patterns, Projects and Guidelines
Multi-cluster Kubernetes Networking- Patterns, Projects and GuidelinesSanjeev Rampal
 
Case study on merger of Vodafone and Idea (VI).pptx
Case study on merger of Vodafone and Idea (VI).pptxCase study on merger of Vodafone and Idea (VI).pptx
Case study on merger of Vodafone and Idea (VI).pptxAnkitscribd
 
ER(Entity Relationship) Diagram for online shopping - TAE
ER(Entity Relationship) Diagram for online shopping - TAEER(Entity Relationship) Diagram for online shopping - TAE
ER(Entity Relationship) Diagram for online shopping - TAEHimani415946
 
Pvtaan Social media marketing proposal.pdf
Pvtaan Social media marketing proposal.pdfPvtaan Social media marketing proposal.pdf
Pvtaan Social media marketing proposal.pdfPvtaan
 
How Do I Begin the Linksys Velop Setup Process?
How Do I Begin the Linksys Velop Setup Process?How Do I Begin the Linksys Velop Setup Process?
How Do I Begin the Linksys Velop Setup Process?Linksys Velop Login
 
audience research (emma) 1.pptxkkkkkkkkkkkkkkkkk
audience research (emma) 1.pptxkkkkkkkkkkkkkkkkkaudience research (emma) 1.pptxkkkkkkkkkkkkkkkkk
audience research (emma) 1.pptxkkkkkkkkkkkkkkkkklolsDocherty
 
The AI Powered Organization-Intro to AI-LAN.pdf
The AI Powered Organization-Intro to AI-LAN.pdfThe AI Powered Organization-Intro to AI-LAN.pdf
The AI Powered Organization-Intro to AI-LAN.pdfSiskaFitrianingrum
 
The+Prospects+of+E-Commerce+in+China.pptx
The+Prospects+of+E-Commerce+in+China.pptxThe+Prospects+of+E-Commerce+in+China.pptx
The+Prospects+of+E-Commerce+in+China.pptxlaozhuseo02
 
一比一原版UTS毕业证悉尼科技大学毕业证成绩单如何办理
一比一原版UTS毕业证悉尼科技大学毕业证成绩单如何办理一比一原版UTS毕业证悉尼科技大学毕业证成绩单如何办理
一比一原版UTS毕业证悉尼科技大学毕业证成绩单如何办理aagad
 
The Use of AI in Indonesia Election 2024: A Case Study
The Use of AI in Indonesia Election 2024: A Case StudyThe Use of AI in Indonesia Election 2024: A Case Study
The Use of AI in Indonesia Election 2024: A Case StudyDamar Juniarto
 

Recently uploaded (13)

Bug Bounty Blueprint : A Beginner's Guide
Bug Bounty Blueprint : A Beginner's GuideBug Bounty Blueprint : A Beginner's Guide
Bug Bounty Blueprint : A Beginner's Guide
 
Article writing on excessive use of internet.pptx
Article writing on excessive use of internet.pptxArticle writing on excessive use of internet.pptx
Article writing on excessive use of internet.pptx
 
Multi-cluster Kubernetes Networking- Patterns, Projects and Guidelines
Multi-cluster Kubernetes Networking- Patterns, Projects and GuidelinesMulti-cluster Kubernetes Networking- Patterns, Projects and Guidelines
Multi-cluster Kubernetes Networking- Patterns, Projects and Guidelines
 
Case study on merger of Vodafone and Idea (VI).pptx
Case study on merger of Vodafone and Idea (VI).pptxCase study on merger of Vodafone and Idea (VI).pptx
Case study on merger of Vodafone and Idea (VI).pptx
 
ER(Entity Relationship) Diagram for online shopping - TAE
ER(Entity Relationship) Diagram for online shopping - TAEER(Entity Relationship) Diagram for online shopping - TAE
ER(Entity Relationship) Diagram for online shopping - TAE
 
Pvtaan Social media marketing proposal.pdf
Pvtaan Social media marketing proposal.pdfPvtaan Social media marketing proposal.pdf
Pvtaan Social media marketing proposal.pdf
 
How Do I Begin the Linksys Velop Setup Process?
How Do I Begin the Linksys Velop Setup Process?How Do I Begin the Linksys Velop Setup Process?
How Do I Begin the Linksys Velop Setup Process?
 
audience research (emma) 1.pptxkkkkkkkkkkkkkkkkk
audience research (emma) 1.pptxkkkkkkkkkkkkkkkkkaudience research (emma) 1.pptxkkkkkkkkkkkkkkkkk
audience research (emma) 1.pptxkkkkkkkkkkkkkkkkk
 
The AI Powered Organization-Intro to AI-LAN.pdf
The AI Powered Organization-Intro to AI-LAN.pdfThe AI Powered Organization-Intro to AI-LAN.pdf
The AI Powered Organization-Intro to AI-LAN.pdf
 
The+Prospects+of+E-Commerce+in+China.pptx
The+Prospects+of+E-Commerce+in+China.pptxThe+Prospects+of+E-Commerce+in+China.pptx
The+Prospects+of+E-Commerce+in+China.pptx
 
一比一原版UTS毕业证悉尼科技大学毕业证成绩单如何办理
一比一原版UTS毕业证悉尼科技大学毕业证成绩单如何办理一比一原版UTS毕业证悉尼科技大学毕业证成绩单如何办理
一比一原版UTS毕业证悉尼科技大学毕业证成绩单如何办理
 
The Use of AI in Indonesia Election 2024: A Case Study
The Use of AI in Indonesia Election 2024: A Case StudyThe Use of AI in Indonesia Election 2024: A Case Study
The Use of AI in Indonesia Election 2024: A Case Study
 
The Best AI Powered Software - Intellivid AI Studio
The Best AI Powered Software - Intellivid AI StudioThe Best AI Powered Software - Intellivid AI Studio
The Best AI Powered Software - Intellivid AI Studio
 

Where to Go on Your Next Trip? Optimizing Travel Destinations Based on User Preferences

  • 1. Where to Go on Your Next Trip? Optimizing Travel Destination Based on User Preferences Julia Kiseleva, Melanie J.I. Mueller, Lucas Bernardi, Chad Davis, Ivan Kovacek, Mats Stafseng Einarsen, Jaap Kamps, Alexander Tuzhilin, Djoerd Hiemstra SIGIR 2015, Chile
  • 3. Booking.com • World largest online travel agent • > 220 countries • > 81.000 destinations • > 710.000 bookable hotels worldwide • > 30.000.000 unique users • >> 100.000 unique visitors per day
  • 7. How Do We Get Endorsements? • Only users who stayed at a hotel in a destination can endorse it • Free text endorsements since 2013 • Since 2014 free text endorsements standardized to 256 canonical tags • Used NLP techniques to extract the canonical base More Numbers: • > 13.000.000 total unique endorsements • > 60.000 destinations
  • 13. Destination Finder Problem: How to Optimize ranking of recommended destinations?
  • 14. Problem Setup • Challenge: o Recommender or Information Retrieval System? • Information Retrieval: o Users have information need which is expressed as a query o System has to satisfy this information needs • Recommender: o System predicts what users might be interested We have both: 1) Users search for activities 2) Users don’t know where they want to go
  • 15. Why is It Hard? Problem Characteristics: • S - Sparsity: new or rare users/destinations • V - Volatility: users’ interests/endorsements of destinations change over time • I - Identity: a failure to match data from the same users • P - Personas: users have different interests at different, possibly closely points in time Continuous Cold Start Problem!
  • 16. Ranking Destinations for ‘Beach’ • Keep it simple!!! We care about performance! • Naïve Bayes: P(Miami, Beach) = P(Miami) * P(Miami | Beach) P(Miami | Beach) = # ‘Beach’ endorsements for Miami # ‘Beach’ endorsements
  • 17. What and How to Compare? • Booking.com Baseline • Random • Most Popular Destination • Naïve Bayes Objection: Increase User Engagement (Clickers per SERP)
  • 18. A/B Testing Setup • 50/50 traffic split • Experiments run for N full weeks according to desired power and significance levels • Hypothesis tests are performed according to targeted metrics (G-test in our case)
  • 19. A/B Testing Version A Version B Version C
  • 20. A/B Testing Ranker # Users Engagement Baseline 9928 Random 10079 Popularity 9838 Naïve Bayes 9895 Amount of clickers
  • 21. A/B Testing Ranker # Users Engagement Baseline 9928 25.61% +/- 0.72% Random 10079 24.46% +/- 0.71% Popularity 9838 25.50% +/- 0.73% Naïve Bayes 9895 26.73% +/- 0.73%
  • 22. A/B Testing Ranker # Users Engagement Baseline 9928 25.61% +/- 0.72% Random 10079 24.46% +/- 0.71% Popularity 9838 25.50% +/- 0.73% Naïve Bayes 9895 26.73% +/- 0.73%
  • 23. Conclusion and Future Work • Interesting application • Surprise 1: Keep simple baseline in production system • Surprise 2: Random performed not bad => effect serendipity For the Future: • Improve the ranking by taking contexts into account
  • 24.
  • 25. Conclusion and Future Work • Interesting application • Surprise 1: Keep simple baseline in production system • Surprise 2: Random performed not bad => effect serendipity For the Future: • Improve the ranking by taking contexts into account

Editor's Notes

  1. Put numbers
  2. Start from IR problem = nothing new Users don’t know what they want => serendipity is very important +> formulate it as Recommendation Problem Think!!!!