SlideShare a Scribd company logo
1 of 27
Download to read offline
Proprietary, Confidential, & Thoughtful
Rohan Singh Rajput
Senior Data Scientist
Headspace
Building Personalized Meditation
Experiences
Proprietary, Confidential, & Thoughtful
Agenda
● Introduction
● History
● Phase I - Modeling
● Phase II - A/B Testing
● Learnings
● Opportunities
● Q/A
2
Proprietary, Confidential, & Thoughtful
Headspace is an online meditation and
mindfulness company
3
Proprietary, Confidential, & Thoughtful
4
5
Proprietary, Confidential, & Thoughtful
History
● Launched in 2020, the content customizer started to
countering the declining retention rate.
● Every headspace user was getting the same editorial content
that used to refresh everyday.
● Headspace had over 1000+ unique content and more than 2
million B2C subscribers
6
Proprietary, Confidential, & Thoughtful
Hypothesis
Personalized recommendations will help members find
more relevant content, thereby increasing their engagement
with Headspace
● Content Customizer (a.k.a CC) is our recommendation
system that serves personalized recommendations
● It uses historical user-item interaction data to generate
recommendation everyday
Proprietary, Confidential, Thoughtful
Proprietary, Confidential, & Thoughtful
Proprietary, Confidential, & Thoughtful
7
Phase I
Modeling
Proprietary, Confidential, & Thoughtful
Components of Modeling
● Data
○ B2C and B2B Subscribers
○ Feature Store
■ User tenure, demographics, language, time of the day etc.
■ Item type, playtime duration, content creation date etc.
○ IOS and Android
● Evaluation
○ Offline: NDGC@K , HITRate@K, Precision@K, etc.
○ Online: A/B Testing
● Inference and Post processing
○ Nightly batched training
○ Payload to prediction service
○ Fix sequence, remove popular and recently completed
Proprietary, Confidential, & Thoughtful
8
9
Proprietary, Confidential, & Thoughtful
Modeling Pipeline
● Deployed for both B2C and B2B members
● Deployed to both iOS and Android platform
Data Pull Model Training
Generate
Prediction
Post
Processing
Send Data to
the Prediction
Service
10
Proprietary, Confidential, & Thoughtful
Version Name Model Type
LightFM model using user and item
features
Feature Based Model
Matrix Factorization model using
historical content interaction
NCF Model using user and item
features
Feature Based Model
Deep learning model that helps to
predict the nonlinear latent factors
and users explicit pos/neg feedback
TiSASRec Model using historical
sequences
Sequence Based Model
Transformer model to capture
sequences of content completes
History of Models
Proprietary, Confidential, & Thoughtful
LightFM Model
● Reasons
○ Matrix Factorization
○ Implicit and Explicit Feedback
○ GPU Optimization
○ Highly scalable
● Result
○ Offline: Precision@k and Recall@k
○ Online: No statistically significant lift (content
start/complete) in A/B test
Kula, Maciej. "Metadata embeddings for user and item
cold-start recommendations." arXiv preprint
arXiv:1507.08439 (2015)
11
Image Credit: Google Developer
Proprietary, Confidential, & Thoughtful
Neural Collaborative Filtering Model
● Reasons
○ Non-linearity
○ User-item features
○ Sparsity
● Result
○ Offline: Precision@k and Recall@k
○ Online: Over 2% statistically significant lift (content
start/complete) in A/B test
He, Xiangnan, et al. "Neural collaborative filtering." Proceedings of the 26th international
conference on world wide web. 2017.
12
Proprietary, Confidential, & Thoughtful
Time Interval Aware Self-Attention for Sequential Recommendation
(TiSASRec)
● Reasons
○ Sequential Data
○ Temporal Dynamics
○ Transformer Architecture
● Result
○ Offline: NDCG@k and HitRate@k
○ Online: Over 4.68% statistically significant lift
(content play) in A/B test
Li, Jiacheng, Yujie Wang, and Julian McAuley. "Time interval aware self-attention for sequential recommendation." Proceedings of the
13th international conference on web search and data mining. 2020.
13
Proprietary, Confidential, & Thoughtful
System Architecture
14
Proprietary, Confidential, & Thoughtful
Proprietary, Confidential, & Thoughtful
15
Phase II
A/B Testing
16
Proprietary, Confidential, & Thoughtful
Introduction
● “If you canʼt measure it, you canʼt improve it.” — Peter Drucker.
● Online Controlled Experiments, a.k.a A/B testing, are the gold
standard for estimating causality with high probability.
● A data-driven decision-making culture helps estimate the
measurement uncertainty to refute the null hypothesis based on
experimental data.
● Furthermore, a random assignment of the users into a
control-treatment group allows us to safely ignore the unobserved
factors and model the parameters as random variables
Proprietary, Confidential, & Thoughtful
17
Proprietary, Confidential, & Thoughtful
Experiment Design
● Recommendation surface area
○ Todayʼs tab
○ Hero and Recommended tab on
Meditate/Sleep/Move/Focus
● Platform Selection
○ iOS
○ Android
● Audience Selection
○ User type: B2C, B2B
○ Subscription type: Paid, Free trials
○ Language: English, German, French
○ Region: North America, International
● Driver Metrics and Business KPI
○ Metrics: Content Start, Content Complete, Content Play,
Content start to complete ratio
○ KPIs: DAU, WAU, MAU, Paid conversion, Churn Rate
Proprietary, Confidential, & Thoughtful
Proprietary, Confidential, & Thoughtful
18
Experiment Design
19
Proprietary, Confidential, & Thoughtful
Experiment Design
● Sample Size and Experiment Duration:
○ Baseline conversion rate
○ Minimum detectable lift
○ Use standard setting with alpha = 0.05 and beta =
20% for 1 sided test
○ Classical test:
https://www.statsig.com/calculator
○ Sequence test:
https://docs.statsig.com/experiments-plus/pow
er-analysis
Proprietary, Confidential, & Thoughtful
20
Proprietary, Confidential, & Thoughtful
Experiment Design
● Important details
○ Prefer A/B test over Multivariate test
○ Use 1 or 2 driver metrics
○ Use guardrail metrics for rollback
○ Avoid early peeking
○ Use confidence intervals over p-values
Proprietary, Confidential, & Thoughtful
21
Proprietary, Confidential, & Thoughtful
Experiment Architecture
● Content Customizer Model: A Machine Learning model
that generates personalized recommendations for every
eligible user. It sends its prediction to the prediction
service using offline inference.
● Prediction Service: A microservice that sends
predictions to the layout service asynchronously. It is
required to update the layout service after
post-processing of the content.
● Layout Service: A bundle of services that creates a
layout for the client for respective tabs. It receives
requests from the client and serves appropriate content
to the layout.
● A/B Testing Tool: A/B testing platform that helps to
administer the online controlled experiment. It
randomly buckets the eligible users into control and
treatment.
Proprietary, Confidential, & Thoughtful
A/B Testing
Tool
22
Proprietary, Confidential, & Thoughtful
Monitoring and Evaluation
● Confidence Interval is much more reliable
indicator of effectiveness than sole p-value
○ p = 0.06 (this)
○ -9% to 20% (v/s this)
● Always use practical significance threshold
along with statistical significance threshold
○ Statistical Significance is a theoretical
indicator of the impact
○ Practical significance is a practical
indicator of the impact
Proprietary, Confidential, & Thoughtful
Proprietary, Confidential, & Thoughtful
Learnings
● Feature Space
○ User features are not rich enough to understand the user behavior
■ Subscription mismatch
■ Sparse interaction
■ Privacy and Compliance
○ Item features are fragmented
■ Many features are stored in multiple system
■ Overlapping content type
● UI Rigidity
○ Fluid design for multiple screens
○ High performing static content
○ Single model everywhere
○ Explicit Feedback
Proprietary, Confidential, & Thoughtful
23
High Performing Content
Static Content
24
Proprietary, Confidential, & Thoughtful
Opportunities
● Client Reform
○ The new shelves based design helps to surface context based
recommendations
○ Provide separate screen to present notified recommendations
○ Search based recommendations
● Modeling
○ Long term value (Align with the productʼs vision)
○ Multi-modal recommendation (text, image, audio)
○ Personalized shelves ranking
○ Causal AI to get explainability
○ Optimize novelty, diversity and relevance
Proprietary, Confidential, & Thoughtful
Personalized
Personalized
24
25
Proprietary, Confidential, & Thoughtful
Opportunities
● Robust E2E System
○ Impression data
○ Analyze data drift, model drift and concept drift
○ Smart caching
○ Real time inference with continuous learning
○ Interconnected recommendation (email, push, notification)
Proprietary, Confidential, & Thoughtful
Personalized
Personalized
25
26
Proprietary, Confidential, & Thoughtful
References
○ Engineering Blog :
■ https://medium.com/headspace-engineering/mindful-experiment
ation-evaluate-recommendation-system-performance-using-a-b-t
esting-at-headspace-3c8c05d0ae3b
○ CXL Guide
■ https://cxl.com/blog/ab-testing-guide/
Proprietary, Confidential, & Thoughtful
27
Proprietary, Confidential, & Thoughtful
Thank you
27
https://www.linkedin.com/in/rohan3/

More Related Content

Similar to A_B Testing Personalized Meditation Recommendations.pdf

Architecting for analytics
Architecting for analyticsArchitecting for analytics
Architecting for analyticsRob Winters
 
Agile enterprise analytics on aws
Agile enterprise analytics on awsAgile enterprise analytics on aws
Agile enterprise analytics on awsDon Gillis
 
Machine Learning At Tubi
Machine Learning At TubiMachine Learning At Tubi
Machine Learning At TubiJaya Kawale
 
Learnings from Developing a New B2B SaaS Product (Suryaveer Lodha (Sunny) Pro...
Learnings from Developing a New B2B SaaS Product (Suryaveer Lodha (Sunny) Pro...Learnings from Developing a New B2B SaaS Product (Suryaveer Lodha (Sunny) Pro...
Learnings from Developing a New B2B SaaS Product (Suryaveer Lodha (Sunny) Pro...IT Arena
 
Transition to a modern data platform
Transition to a modern data platform Transition to a modern data platform
Transition to a modern data platform Michael Ghen
 
Video Recommendation Engines as a Service
Video Recommendation Engines as a ServiceVideo Recommendation Engines as a Service
Video Recommendation Engines as a ServiceKamil Sindi
 
ING's Customer-Centric Data Journey from Community Idea to Private Cloud Depl...
ING's Customer-Centric Data Journey from Community Idea to Private Cloud Depl...ING's Customer-Centric Data Journey from Community Idea to Private Cloud Depl...
ING's Customer-Centric Data Journey from Community Idea to Private Cloud Depl...DataWorks Summit/Hadoop Summit
 
Experience Matters: Understanding the New ROI of UX/CX
Experience Matters: Understanding the New ROI of UX/CX Experience Matters: Understanding the New ROI of UX/CX
Experience Matters: Understanding the New ROI of UX/CX Mediacurrent
 
Introduction to business analysis
Introduction to business analysisIntroduction to business analysis
Introduction to business analysisMichael Kramarenko
 
Analytics is Taking over the World (Again) - UKOUG Tech'17
Analytics is Taking over the World (Again) - UKOUG Tech'17Analytics is Taking over the World (Again) - UKOUG Tech'17
Analytics is Taking over the World (Again) - UKOUG Tech'17Rittman Analytics
 
Planning For The Personalization Journey: From Empathy to Engagement
 Planning For The Personalization Journey: From Empathy to Engagement Planning For The Personalization Journey: From Empathy to Engagement
Planning For The Personalization Journey: From Empathy to EngagementRachel Wandishin
 
Using Data Science to Build an End-to-End Recommendation System
Using Data Science to Build an End-to-End Recommendation SystemUsing Data Science to Build an End-to-End Recommendation System
Using Data Science to Build an End-to-End Recommendation SystemVMware Tanzu
 
National Wildlife Federation- OMS- Dreamcore 2011
National Wildlife Federation- OMS- Dreamcore 2011National Wildlife Federation- OMS- Dreamcore 2011
National Wildlife Federation- OMS- Dreamcore 2011nonlinear creations
 
3 Challenges of Building Complex Dashboards with Open Source Components
3 Challenges of Building Complex Dashboards with Open Source Components3 Challenges of Building Complex Dashboards with Open Source Components
3 Challenges of Building Complex Dashboards with Open Source ComponentsRyan MacCarrigan
 
Selecting a Digital Collections Management System: Getting Large Projects Don...
Selecting a Digital Collections Management System: Getting Large Projects Don...Selecting a Digital Collections Management System: Getting Large Projects Don...
Selecting a Digital Collections Management System: Getting Large Projects Don...Cliff Landis
 
Customer Research For Product Managers - Dawn of The Data Age Lecture Series
Customer Research For Product Managers - Dawn of The Data Age Lecture SeriesCustomer Research For Product Managers - Dawn of The Data Age Lecture Series
Customer Research For Product Managers - Dawn of The Data Age Lecture SeriesLuciano Pesci, PhD
 
Educate 2017: Neverending Story: Exploring Learnosity's ongoing product evolu...
Educate 2017: Neverending Story: Exploring Learnosity's ongoing product evolu...Educate 2017: Neverending Story: Exploring Learnosity's ongoing product evolu...
Educate 2017: Neverending Story: Exploring Learnosity's ongoing product evolu...Learnosity
 
GraphTour London 2020 - Customer Journey
GraphTour London 2020  - Customer Journey GraphTour London 2020  - Customer Journey
GraphTour London 2020 - Customer Journey Neo4j
 
MVP (Minimum Viable Product) Readiness | Boost Labs
MVP (Minimum Viable Product) Readiness | Boost LabsMVP (Minimum Viable Product) Readiness | Boost Labs
MVP (Minimum Viable Product) Readiness | Boost LabsBoost Labs
 

Similar to A_B Testing Personalized Meditation Recommendations.pdf (20)

Architecting for analytics
Architecting for analyticsArchitecting for analytics
Architecting for analytics
 
Agile enterprise analytics on aws
Agile enterprise analytics on awsAgile enterprise analytics on aws
Agile enterprise analytics on aws
 
Machine Learning At Tubi
Machine Learning At TubiMachine Learning At Tubi
Machine Learning At Tubi
 
Knowledge Discovery
Knowledge DiscoveryKnowledge Discovery
Knowledge Discovery
 
Learnings from Developing a New B2B SaaS Product (Suryaveer Lodha (Sunny) Pro...
Learnings from Developing a New B2B SaaS Product (Suryaveer Lodha (Sunny) Pro...Learnings from Developing a New B2B SaaS Product (Suryaveer Lodha (Sunny) Pro...
Learnings from Developing a New B2B SaaS Product (Suryaveer Lodha (Sunny) Pro...
 
Transition to a modern data platform
Transition to a modern data platform Transition to a modern data platform
Transition to a modern data platform
 
Video Recommendation Engines as a Service
Video Recommendation Engines as a ServiceVideo Recommendation Engines as a Service
Video Recommendation Engines as a Service
 
ING's Customer-Centric Data Journey from Community Idea to Private Cloud Depl...
ING's Customer-Centric Data Journey from Community Idea to Private Cloud Depl...ING's Customer-Centric Data Journey from Community Idea to Private Cloud Depl...
ING's Customer-Centric Data Journey from Community Idea to Private Cloud Depl...
 
Experience Matters: Understanding the New ROI of UX/CX
Experience Matters: Understanding the New ROI of UX/CX Experience Matters: Understanding the New ROI of UX/CX
Experience Matters: Understanding the New ROI of UX/CX
 
Introduction to business analysis
Introduction to business analysisIntroduction to business analysis
Introduction to business analysis
 
Analytics is Taking over the World (Again) - UKOUG Tech'17
Analytics is Taking over the World (Again) - UKOUG Tech'17Analytics is Taking over the World (Again) - UKOUG Tech'17
Analytics is Taking over the World (Again) - UKOUG Tech'17
 
Planning For The Personalization Journey: From Empathy to Engagement
 Planning For The Personalization Journey: From Empathy to Engagement Planning For The Personalization Journey: From Empathy to Engagement
Planning For The Personalization Journey: From Empathy to Engagement
 
Using Data Science to Build an End-to-End Recommendation System
Using Data Science to Build an End-to-End Recommendation SystemUsing Data Science to Build an End-to-End Recommendation System
Using Data Science to Build an End-to-End Recommendation System
 
National Wildlife Federation- OMS- Dreamcore 2011
National Wildlife Federation- OMS- Dreamcore 2011National Wildlife Federation- OMS- Dreamcore 2011
National Wildlife Federation- OMS- Dreamcore 2011
 
3 Challenges of Building Complex Dashboards with Open Source Components
3 Challenges of Building Complex Dashboards with Open Source Components3 Challenges of Building Complex Dashboards with Open Source Components
3 Challenges of Building Complex Dashboards with Open Source Components
 
Selecting a Digital Collections Management System: Getting Large Projects Don...
Selecting a Digital Collections Management System: Getting Large Projects Don...Selecting a Digital Collections Management System: Getting Large Projects Don...
Selecting a Digital Collections Management System: Getting Large Projects Don...
 
Customer Research For Product Managers - Dawn of The Data Age Lecture Series
Customer Research For Product Managers - Dawn of The Data Age Lecture SeriesCustomer Research For Product Managers - Dawn of The Data Age Lecture Series
Customer Research For Product Managers - Dawn of The Data Age Lecture Series
 
Educate 2017: Neverending Story: Exploring Learnosity's ongoing product evolu...
Educate 2017: Neverending Story: Exploring Learnosity's ongoing product evolu...Educate 2017: Neverending Story: Exploring Learnosity's ongoing product evolu...
Educate 2017: Neverending Story: Exploring Learnosity's ongoing product evolu...
 
GraphTour London 2020 - Customer Journey
GraphTour London 2020  - Customer Journey GraphTour London 2020  - Customer Journey
GraphTour London 2020 - Customer Journey
 
MVP (Minimum Viable Product) Readiness | Boost Labs
MVP (Minimum Viable Product) Readiness | Boost LabsMVP (Minimum Viable Product) Readiness | Boost Labs
MVP (Minimum Viable Product) Readiness | Boost Labs
 

More from VWO

Fueling A_B experiments with behavioral insights (1).pdf
Fueling A_B experiments with behavioral insights (1).pdfFueling A_B experiments with behavioral insights (1).pdf
Fueling A_B experiments with behavioral insights (1).pdfVWO
 
Research and Discovery Tools for Experimentation - 17 Apr 2024 - v 2.3 (1).pdf
Research and Discovery Tools for Experimentation - 17 Apr 2024 - v 2.3 (1).pdfResearch and Discovery Tools for Experimentation - 17 Apr 2024 - v 2.3 (1).pdf
Research and Discovery Tools for Experimentation - 17 Apr 2024 - v 2.3 (1).pdfVWO
 
Life between variants by Tomas Voves.pdf
Life between variants by Tomas Voves.pdfLife between variants by Tomas Voves.pdf
Life between variants by Tomas Voves.pdfVWO
 
Message-Market Fit VWO Workshop by Daphne tideman
Message-Market Fit VWO Workshop by Daphne tidemanMessage-Market Fit VWO Workshop by Daphne tideman
Message-Market Fit VWO Workshop by Daphne tidemanVWO
 
Converting with Comedy: Research Parallels for CRO
Converting with Comedy: Research Parallels for CROConverting with Comedy: Research Parallels for CRO
Converting with Comedy: Research Parallels for CROVWO
 
Unifying feature management with experiments - Server Side Webinar (1).pdf
Unifying feature management with experiments - Server Side Webinar (1).pdfUnifying feature management with experiments - Server Side Webinar (1).pdf
Unifying feature management with experiments - Server Side Webinar (1).pdfVWO
 
Run more experiments with fewer resources
Run more experiments with fewer resourcesRun more experiments with fewer resources
Run more experiments with fewer resourcesVWO
 
Crafting High-Converting eCommerce Landing Pages
Crafting High-Converting eCommerce Landing PagesCrafting High-Converting eCommerce Landing Pages
Crafting High-Converting eCommerce Landing PagesVWO
 
Diseñando según la regulación actual de cookies para lograr más conversión
Diseñando según la regulación actual de cookies para lograr más conversiónDiseñando según la regulación actual de cookies para lograr más conversión
Diseñando según la regulación actual de cookies para lograr más conversiónVWO
 
VWO - Mark de Winter - Run more experiments with fewer resources.pdf
VWO - Mark de Winter - Run more experiments with fewer resources.pdfVWO - Mark de Winter - Run more experiments with fewer resources.pdf
VWO - Mark de Winter - Run more experiments with fewer resources.pdfVWO
 
Turn Website Visits into Conversions (2).pdf
Turn Website Visits into Conversions (2).pdfTurn Website Visits into Conversions (2).pdf
Turn Website Visits into Conversions (2).pdfVWO
 
Principales obstáculos de la experimentación y cómo abordarlos en StartUps...
Principales obstáculos de la experimentación y cómo abordarlos en StartUps...Principales obstáculos de la experimentación y cómo abordarlos en StartUps...
Principales obstáculos de la experimentación y cómo abordarlos en StartUps...VWO
 
Understand user struggles. Optimize mobile app UX.
Understand user struggles. Optimize mobile app UX.Understand user struggles. Optimize mobile app UX.
Understand user struggles. Optimize mobile app UX.VWO
 
Turn Website Visits into Conversions (1).pdf
Turn Website Visits into Conversions (1).pdfTurn Website Visits into Conversions (1).pdf
Turn Website Visits into Conversions (1).pdfVWO
 
Converting with Comedy: Research Parallels for CRO
Converting with Comedy: Research Parallels for CROConverting with Comedy: Research Parallels for CRO
Converting with Comedy: Research Parallels for CROVWO
 
Optimizing Mobile App UX for better engagement & conversions.pptx
Optimizing Mobile App UX for better engagement & conversions.pptxOptimizing Mobile App UX for better engagement & conversions.pptx
Optimizing Mobile App UX for better engagement & conversions.pptxVWO
 
Optimizing Mobile App UX for better engagement & conversions.pdf
Optimizing Mobile App UX for better engagement & conversions.pdfOptimizing Mobile App UX for better engagement & conversions.pdf
Optimizing Mobile App UX for better engagement & conversions.pdfVWO
 
7 puntos clave y prácticos en todo proyecto de CRO
7 puntos clave y prácticos en todo proyecto de CRO7 puntos clave y prácticos en todo proyecto de CRO
7 puntos clave y prácticos en todo proyecto de CROVWO
 
Tackling Conversion Problems with Personalization | VWO Webinars
Tackling Conversion Problems with Personalization | VWO WebinarsTackling Conversion Problems with Personalization | VWO Webinars
Tackling Conversion Problems with Personalization | VWO WebinarsVWO
 
Turning Business Challenges into Testable Ideas - 29 Nov '23.pdf
Turning Business Challenges into Testable Ideas - 29 Nov '23.pdfTurning Business Challenges into Testable Ideas - 29 Nov '23.pdf
Turning Business Challenges into Testable Ideas - 29 Nov '23.pdfVWO
 

More from VWO (20)

Fueling A_B experiments with behavioral insights (1).pdf
Fueling A_B experiments with behavioral insights (1).pdfFueling A_B experiments with behavioral insights (1).pdf
Fueling A_B experiments with behavioral insights (1).pdf
 
Research and Discovery Tools for Experimentation - 17 Apr 2024 - v 2.3 (1).pdf
Research and Discovery Tools for Experimentation - 17 Apr 2024 - v 2.3 (1).pdfResearch and Discovery Tools for Experimentation - 17 Apr 2024 - v 2.3 (1).pdf
Research and Discovery Tools for Experimentation - 17 Apr 2024 - v 2.3 (1).pdf
 
Life between variants by Tomas Voves.pdf
Life between variants by Tomas Voves.pdfLife between variants by Tomas Voves.pdf
Life between variants by Tomas Voves.pdf
 
Message-Market Fit VWO Workshop by Daphne tideman
Message-Market Fit VWO Workshop by Daphne tidemanMessage-Market Fit VWO Workshop by Daphne tideman
Message-Market Fit VWO Workshop by Daphne tideman
 
Converting with Comedy: Research Parallels for CRO
Converting with Comedy: Research Parallels for CROConverting with Comedy: Research Parallels for CRO
Converting with Comedy: Research Parallels for CRO
 
Unifying feature management with experiments - Server Side Webinar (1).pdf
Unifying feature management with experiments - Server Side Webinar (1).pdfUnifying feature management with experiments - Server Side Webinar (1).pdf
Unifying feature management with experiments - Server Side Webinar (1).pdf
 
Run more experiments with fewer resources
Run more experiments with fewer resourcesRun more experiments with fewer resources
Run more experiments with fewer resources
 
Crafting High-Converting eCommerce Landing Pages
Crafting High-Converting eCommerce Landing PagesCrafting High-Converting eCommerce Landing Pages
Crafting High-Converting eCommerce Landing Pages
 
Diseñando según la regulación actual de cookies para lograr más conversión
Diseñando según la regulación actual de cookies para lograr más conversiónDiseñando según la regulación actual de cookies para lograr más conversión
Diseñando según la regulación actual de cookies para lograr más conversión
 
VWO - Mark de Winter - Run more experiments with fewer resources.pdf
VWO - Mark de Winter - Run more experiments with fewer resources.pdfVWO - Mark de Winter - Run more experiments with fewer resources.pdf
VWO - Mark de Winter - Run more experiments with fewer resources.pdf
 
Turn Website Visits into Conversions (2).pdf
Turn Website Visits into Conversions (2).pdfTurn Website Visits into Conversions (2).pdf
Turn Website Visits into Conversions (2).pdf
 
Principales obstáculos de la experimentación y cómo abordarlos en StartUps...
Principales obstáculos de la experimentación y cómo abordarlos en StartUps...Principales obstáculos de la experimentación y cómo abordarlos en StartUps...
Principales obstáculos de la experimentación y cómo abordarlos en StartUps...
 
Understand user struggles. Optimize mobile app UX.
Understand user struggles. Optimize mobile app UX.Understand user struggles. Optimize mobile app UX.
Understand user struggles. Optimize mobile app UX.
 
Turn Website Visits into Conversions (1).pdf
Turn Website Visits into Conversions (1).pdfTurn Website Visits into Conversions (1).pdf
Turn Website Visits into Conversions (1).pdf
 
Converting with Comedy: Research Parallels for CRO
Converting with Comedy: Research Parallels for CROConverting with Comedy: Research Parallels for CRO
Converting with Comedy: Research Parallels for CRO
 
Optimizing Mobile App UX for better engagement & conversions.pptx
Optimizing Mobile App UX for better engagement & conversions.pptxOptimizing Mobile App UX for better engagement & conversions.pptx
Optimizing Mobile App UX for better engagement & conversions.pptx
 
Optimizing Mobile App UX for better engagement & conversions.pdf
Optimizing Mobile App UX for better engagement & conversions.pdfOptimizing Mobile App UX for better engagement & conversions.pdf
Optimizing Mobile App UX for better engagement & conversions.pdf
 
7 puntos clave y prácticos en todo proyecto de CRO
7 puntos clave y prácticos en todo proyecto de CRO7 puntos clave y prácticos en todo proyecto de CRO
7 puntos clave y prácticos en todo proyecto de CRO
 
Tackling Conversion Problems with Personalization | VWO Webinars
Tackling Conversion Problems with Personalization | VWO WebinarsTackling Conversion Problems with Personalization | VWO Webinars
Tackling Conversion Problems with Personalization | VWO Webinars
 
Turning Business Challenges into Testable Ideas - 29 Nov '23.pdf
Turning Business Challenges into Testable Ideas - 29 Nov '23.pdfTurning Business Challenges into Testable Ideas - 29 Nov '23.pdf
Turning Business Challenges into Testable Ideas - 29 Nov '23.pdf
 

Recently uploaded

Exploring The World Of Adult Ad Networks.pdf
Exploring The World Of Adult Ad Networks.pdfExploring The World Of Adult Ad Networks.pdf
Exploring The World Of Adult Ad Networks.pdfadult marketing
 
Call Girls in Lajpat Nagar Delhi 💯Call Us 🔝8264348440🔝
Call Girls in Lajpat Nagar Delhi 💯Call Us 🔝8264348440🔝Call Girls in Lajpat Nagar Delhi 💯Call Us 🔝8264348440🔝
Call Girls in Lajpat Nagar Delhi 💯Call Us 🔝8264348440🔝soniya singh
 
McDonald's: A Journey Through Time (PPT)
McDonald's: A Journey Through Time (PPT)McDonald's: A Journey Through Time (PPT)
McDonald's: A Journey Through Time (PPT)DEVARAJV16
 
Talent Management for mba 3rd sem useful
Talent Management for mba 3rd sem usefulTalent Management for mba 3rd sem useful
Talent Management for mba 3rd sem usefulAtifaArbar
 
Word Count for Writers: Examples of Word Counts for Sample Genres
Word Count for Writers: Examples of Word Counts for Sample GenresWord Count for Writers: Examples of Word Counts for Sample Genres
Word Count for Writers: Examples of Word Counts for Sample GenresLisa M. Masiello
 
Most Influential HR Leaders Leading the Corporate World, 2024 (Final file).pdf
Most Influential HR Leaders Leading the Corporate World, 2024 (Final file).pdfMost Influential HR Leaders Leading the Corporate World, 2024 (Final file).pdf
Most Influential HR Leaders Leading the Corporate World, 2024 (Final file).pdfCIO Business World
 
Most Impressive Construction Leaders in Tech, Making Waves in the Industry, 2...
Most Impressive Construction Leaders in Tech, Making Waves in the Industry, 2...Most Impressive Construction Leaders in Tech, Making Waves in the Industry, 2...
Most Impressive Construction Leaders in Tech, Making Waves in the Industry, 2...CIO Business World
 
Exploring Web 3.0 Growth marketing: Navigating the Future of the Internet
Exploring Web 3.0 Growth marketing: Navigating the Future of the InternetExploring Web 3.0 Growth marketing: Navigating the Future of the Internet
Exploring Web 3.0 Growth marketing: Navigating the Future of the Internetnehapardhi711
 
Red bull marketing presentation pptxxxxx
Red bull marketing presentation pptxxxxxRed bull marketing presentation pptxxxxx
Red bull marketing presentation pptxxxxx216310017
 
Storyboards for my Final Major Project Video
Storyboards for my Final Major Project VideoStoryboards for my Final Major Project Video
Storyboards for my Final Major Project VideoSineadBidwell
 
marketing strategy of tanishq word PPROJECT.pdf
marketing strategy of tanishq word PPROJECT.pdfmarketing strategy of tanishq word PPROJECT.pdf
marketing strategy of tanishq word PPROJECT.pdfarsathsahil
 
Mastering SEO in the Evolving AI-driven World
Mastering SEO in the Evolving AI-driven WorldMastering SEO in the Evolving AI-driven World
Mastering SEO in the Evolving AI-driven WorldScalenut
 
From Chance to Choice - Tactical Link Building for International SEO
From Chance to Choice - Tactical Link Building for International SEOFrom Chance to Choice - Tactical Link Building for International SEO
From Chance to Choice - Tactical Link Building for International SEOSzymon Słowik
 
The 10 Most Inspirational Leaders LEADING THE WAY TO SUCCESS, 2024
The 10 Most Inspirational Leaders LEADING THE WAY TO SUCCESS, 2024The 10 Most Inspirational Leaders LEADING THE WAY TO SUCCESS, 2024
The 10 Most Inspirational Leaders LEADING THE WAY TO SUCCESS, 2024CIO Business World
 
2024 SEO Trends for Business Success (WSA)
2024 SEO Trends for Business Success (WSA)2024 SEO Trends for Business Success (WSA)
2024 SEO Trends for Business Success (WSA)Jomer Gregorio
 
What I learned from auditing over 1,000,000 websites - SERP Conf 2024 Patrick...
What I learned from auditing over 1,000,000 websites - SERP Conf 2024 Patrick...What I learned from auditing over 1,000,000 websites - SERP Conf 2024 Patrick...
What I learned from auditing over 1,000,000 websites - SERP Conf 2024 Patrick...Ahrefs
 
DIGITAL MARKETING STRATEGY_INFOGRAPHIC IMAGE.pdf
DIGITAL MARKETING STRATEGY_INFOGRAPHIC IMAGE.pdfDIGITAL MARKETING STRATEGY_INFOGRAPHIC IMAGE.pdf
DIGITAL MARKETING STRATEGY_INFOGRAPHIC IMAGE.pdfmayanksharma0441
 
Fiverr's Product Marketing Interview Assignment
Fiverr's Product Marketing Interview AssignmentFiverr's Product Marketing Interview Assignment
Fiverr's Product Marketing Interview AssignmentFarrel Brest
 
Michael Kors marketing assignment swot analysis
Michael Kors marketing assignment swot analysisMichael Kors marketing assignment swot analysis
Michael Kors marketing assignment swot analysisjunaid794917
 
Master the Art of Digital Recruitment in Asia.pdf
Master the Art of Digital Recruitment in Asia.pdfMaster the Art of Digital Recruitment in Asia.pdf
Master the Art of Digital Recruitment in Asia.pdfHigher Education Marketing
 

Recently uploaded (20)

Exploring The World Of Adult Ad Networks.pdf
Exploring The World Of Adult Ad Networks.pdfExploring The World Of Adult Ad Networks.pdf
Exploring The World Of Adult Ad Networks.pdf
 
Call Girls in Lajpat Nagar Delhi 💯Call Us 🔝8264348440🔝
Call Girls in Lajpat Nagar Delhi 💯Call Us 🔝8264348440🔝Call Girls in Lajpat Nagar Delhi 💯Call Us 🔝8264348440🔝
Call Girls in Lajpat Nagar Delhi 💯Call Us 🔝8264348440🔝
 
McDonald's: A Journey Through Time (PPT)
McDonald's: A Journey Through Time (PPT)McDonald's: A Journey Through Time (PPT)
McDonald's: A Journey Through Time (PPT)
 
Talent Management for mba 3rd sem useful
Talent Management for mba 3rd sem usefulTalent Management for mba 3rd sem useful
Talent Management for mba 3rd sem useful
 
Word Count for Writers: Examples of Word Counts for Sample Genres
Word Count for Writers: Examples of Word Counts for Sample GenresWord Count for Writers: Examples of Word Counts for Sample Genres
Word Count for Writers: Examples of Word Counts for Sample Genres
 
Most Influential HR Leaders Leading the Corporate World, 2024 (Final file).pdf
Most Influential HR Leaders Leading the Corporate World, 2024 (Final file).pdfMost Influential HR Leaders Leading the Corporate World, 2024 (Final file).pdf
Most Influential HR Leaders Leading the Corporate World, 2024 (Final file).pdf
 
Most Impressive Construction Leaders in Tech, Making Waves in the Industry, 2...
Most Impressive Construction Leaders in Tech, Making Waves in the Industry, 2...Most Impressive Construction Leaders in Tech, Making Waves in the Industry, 2...
Most Impressive Construction Leaders in Tech, Making Waves in the Industry, 2...
 
Exploring Web 3.0 Growth marketing: Navigating the Future of the Internet
Exploring Web 3.0 Growth marketing: Navigating the Future of the InternetExploring Web 3.0 Growth marketing: Navigating the Future of the Internet
Exploring Web 3.0 Growth marketing: Navigating the Future of the Internet
 
Red bull marketing presentation pptxxxxx
Red bull marketing presentation pptxxxxxRed bull marketing presentation pptxxxxx
Red bull marketing presentation pptxxxxx
 
Storyboards for my Final Major Project Video
Storyboards for my Final Major Project VideoStoryboards for my Final Major Project Video
Storyboards for my Final Major Project Video
 
marketing strategy of tanishq word PPROJECT.pdf
marketing strategy of tanishq word PPROJECT.pdfmarketing strategy of tanishq word PPROJECT.pdf
marketing strategy of tanishq word PPROJECT.pdf
 
Mastering SEO in the Evolving AI-driven World
Mastering SEO in the Evolving AI-driven WorldMastering SEO in the Evolving AI-driven World
Mastering SEO in the Evolving AI-driven World
 
From Chance to Choice - Tactical Link Building for International SEO
From Chance to Choice - Tactical Link Building for International SEOFrom Chance to Choice - Tactical Link Building for International SEO
From Chance to Choice - Tactical Link Building for International SEO
 
The 10 Most Inspirational Leaders LEADING THE WAY TO SUCCESS, 2024
The 10 Most Inspirational Leaders LEADING THE WAY TO SUCCESS, 2024The 10 Most Inspirational Leaders LEADING THE WAY TO SUCCESS, 2024
The 10 Most Inspirational Leaders LEADING THE WAY TO SUCCESS, 2024
 
2024 SEO Trends for Business Success (WSA)
2024 SEO Trends for Business Success (WSA)2024 SEO Trends for Business Success (WSA)
2024 SEO Trends for Business Success (WSA)
 
What I learned from auditing over 1,000,000 websites - SERP Conf 2024 Patrick...
What I learned from auditing over 1,000,000 websites - SERP Conf 2024 Patrick...What I learned from auditing over 1,000,000 websites - SERP Conf 2024 Patrick...
What I learned from auditing over 1,000,000 websites - SERP Conf 2024 Patrick...
 
DIGITAL MARKETING STRATEGY_INFOGRAPHIC IMAGE.pdf
DIGITAL MARKETING STRATEGY_INFOGRAPHIC IMAGE.pdfDIGITAL MARKETING STRATEGY_INFOGRAPHIC IMAGE.pdf
DIGITAL MARKETING STRATEGY_INFOGRAPHIC IMAGE.pdf
 
Fiverr's Product Marketing Interview Assignment
Fiverr's Product Marketing Interview AssignmentFiverr's Product Marketing Interview Assignment
Fiverr's Product Marketing Interview Assignment
 
Michael Kors marketing assignment swot analysis
Michael Kors marketing assignment swot analysisMichael Kors marketing assignment swot analysis
Michael Kors marketing assignment swot analysis
 
Master the Art of Digital Recruitment in Asia.pdf
Master the Art of Digital Recruitment in Asia.pdfMaster the Art of Digital Recruitment in Asia.pdf
Master the Art of Digital Recruitment in Asia.pdf
 

A_B Testing Personalized Meditation Recommendations.pdf

  • 1. Proprietary, Confidential, & Thoughtful Rohan Singh Rajput Senior Data Scientist Headspace Building Personalized Meditation Experiences
  • 2. Proprietary, Confidential, & Thoughtful Agenda ● Introduction ● History ● Phase I - Modeling ● Phase II - A/B Testing ● Learnings ● Opportunities ● Q/A 2
  • 3. Proprietary, Confidential, & Thoughtful Headspace is an online meditation and mindfulness company 3
  • 5. 5 Proprietary, Confidential, & Thoughtful History ● Launched in 2020, the content customizer started to countering the declining retention rate. ● Every headspace user was getting the same editorial content that used to refresh everyday. ● Headspace had over 1000+ unique content and more than 2 million B2C subscribers
  • 6. 6 Proprietary, Confidential, & Thoughtful Hypothesis Personalized recommendations will help members find more relevant content, thereby increasing their engagement with Headspace ● Content Customizer (a.k.a CC) is our recommendation system that serves personalized recommendations ● It uses historical user-item interaction data to generate recommendation everyday Proprietary, Confidential, Thoughtful
  • 7. Proprietary, Confidential, & Thoughtful Proprietary, Confidential, & Thoughtful 7 Phase I Modeling
  • 8. Proprietary, Confidential, & Thoughtful Components of Modeling ● Data ○ B2C and B2B Subscribers ○ Feature Store ■ User tenure, demographics, language, time of the day etc. ■ Item type, playtime duration, content creation date etc. ○ IOS and Android ● Evaluation ○ Offline: NDGC@K , HITRate@K, Precision@K, etc. ○ Online: A/B Testing ● Inference and Post processing ○ Nightly batched training ○ Payload to prediction service ○ Fix sequence, remove popular and recently completed Proprietary, Confidential, & Thoughtful 8
  • 9. 9 Proprietary, Confidential, & Thoughtful Modeling Pipeline ● Deployed for both B2C and B2B members ● Deployed to both iOS and Android platform Data Pull Model Training Generate Prediction Post Processing Send Data to the Prediction Service
  • 10. 10 Proprietary, Confidential, & Thoughtful Version Name Model Type LightFM model using user and item features Feature Based Model Matrix Factorization model using historical content interaction NCF Model using user and item features Feature Based Model Deep learning model that helps to predict the nonlinear latent factors and users explicit pos/neg feedback TiSASRec Model using historical sequences Sequence Based Model Transformer model to capture sequences of content completes History of Models
  • 11. Proprietary, Confidential, & Thoughtful LightFM Model ● Reasons ○ Matrix Factorization ○ Implicit and Explicit Feedback ○ GPU Optimization ○ Highly scalable ● Result ○ Offline: Precision@k and Recall@k ○ Online: No statistically significant lift (content start/complete) in A/B test Kula, Maciej. "Metadata embeddings for user and item cold-start recommendations." arXiv preprint arXiv:1507.08439 (2015) 11 Image Credit: Google Developer
  • 12. Proprietary, Confidential, & Thoughtful Neural Collaborative Filtering Model ● Reasons ○ Non-linearity ○ User-item features ○ Sparsity ● Result ○ Offline: Precision@k and Recall@k ○ Online: Over 2% statistically significant lift (content start/complete) in A/B test He, Xiangnan, et al. "Neural collaborative filtering." Proceedings of the 26th international conference on world wide web. 2017. 12
  • 13. Proprietary, Confidential, & Thoughtful Time Interval Aware Self-Attention for Sequential Recommendation (TiSASRec) ● Reasons ○ Sequential Data ○ Temporal Dynamics ○ Transformer Architecture ● Result ○ Offline: NDCG@k and HitRate@k ○ Online: Over 4.68% statistically significant lift (content play) in A/B test Li, Jiacheng, Yujie Wang, and Julian McAuley. "Time interval aware self-attention for sequential recommendation." Proceedings of the 13th international conference on web search and data mining. 2020. 13
  • 14. Proprietary, Confidential, & Thoughtful System Architecture 14
  • 15. Proprietary, Confidential, & Thoughtful Proprietary, Confidential, & Thoughtful 15 Phase II A/B Testing
  • 16. 16 Proprietary, Confidential, & Thoughtful Introduction ● “If you canʼt measure it, you canʼt improve it.” — Peter Drucker. ● Online Controlled Experiments, a.k.a A/B testing, are the gold standard for estimating causality with high probability. ● A data-driven decision-making culture helps estimate the measurement uncertainty to refute the null hypothesis based on experimental data. ● Furthermore, a random assignment of the users into a control-treatment group allows us to safely ignore the unobserved factors and model the parameters as random variables Proprietary, Confidential, & Thoughtful
  • 17. 17 Proprietary, Confidential, & Thoughtful Experiment Design ● Recommendation surface area ○ Todayʼs tab ○ Hero and Recommended tab on Meditate/Sleep/Move/Focus ● Platform Selection ○ iOS ○ Android ● Audience Selection ○ User type: B2C, B2B ○ Subscription type: Paid, Free trials ○ Language: English, German, French ○ Region: North America, International ● Driver Metrics and Business KPI ○ Metrics: Content Start, Content Complete, Content Play, Content start to complete ratio ○ KPIs: DAU, WAU, MAU, Paid conversion, Churn Rate Proprietary, Confidential, & Thoughtful
  • 18. Proprietary, Confidential, & Thoughtful 18 Experiment Design
  • 19. 19 Proprietary, Confidential, & Thoughtful Experiment Design ● Sample Size and Experiment Duration: ○ Baseline conversion rate ○ Minimum detectable lift ○ Use standard setting with alpha = 0.05 and beta = 20% for 1 sided test ○ Classical test: https://www.statsig.com/calculator ○ Sequence test: https://docs.statsig.com/experiments-plus/pow er-analysis Proprietary, Confidential, & Thoughtful
  • 20. 20 Proprietary, Confidential, & Thoughtful Experiment Design ● Important details ○ Prefer A/B test over Multivariate test ○ Use 1 or 2 driver metrics ○ Use guardrail metrics for rollback ○ Avoid early peeking ○ Use confidence intervals over p-values Proprietary, Confidential, & Thoughtful
  • 21. 21 Proprietary, Confidential, & Thoughtful Experiment Architecture ● Content Customizer Model: A Machine Learning model that generates personalized recommendations for every eligible user. It sends its prediction to the prediction service using offline inference. ● Prediction Service: A microservice that sends predictions to the layout service asynchronously. It is required to update the layout service after post-processing of the content. ● Layout Service: A bundle of services that creates a layout for the client for respective tabs. It receives requests from the client and serves appropriate content to the layout. ● A/B Testing Tool: A/B testing platform that helps to administer the online controlled experiment. It randomly buckets the eligible users into control and treatment. Proprietary, Confidential, & Thoughtful A/B Testing Tool
  • 22. 22 Proprietary, Confidential, & Thoughtful Monitoring and Evaluation ● Confidence Interval is much more reliable indicator of effectiveness than sole p-value ○ p = 0.06 (this) ○ -9% to 20% (v/s this) ● Always use practical significance threshold along with statistical significance threshold ○ Statistical Significance is a theoretical indicator of the impact ○ Practical significance is a practical indicator of the impact Proprietary, Confidential, & Thoughtful
  • 23. Proprietary, Confidential, & Thoughtful Learnings ● Feature Space ○ User features are not rich enough to understand the user behavior ■ Subscription mismatch ■ Sparse interaction ■ Privacy and Compliance ○ Item features are fragmented ■ Many features are stored in multiple system ■ Overlapping content type ● UI Rigidity ○ Fluid design for multiple screens ○ High performing static content ○ Single model everywhere ○ Explicit Feedback Proprietary, Confidential, & Thoughtful 23 High Performing Content Static Content
  • 24. 24 Proprietary, Confidential, & Thoughtful Opportunities ● Client Reform ○ The new shelves based design helps to surface context based recommendations ○ Provide separate screen to present notified recommendations ○ Search based recommendations ● Modeling ○ Long term value (Align with the productʼs vision) ○ Multi-modal recommendation (text, image, audio) ○ Personalized shelves ranking ○ Causal AI to get explainability ○ Optimize novelty, diversity and relevance Proprietary, Confidential, & Thoughtful Personalized Personalized 24
  • 25. 25 Proprietary, Confidential, & Thoughtful Opportunities ● Robust E2E System ○ Impression data ○ Analyze data drift, model drift and concept drift ○ Smart caching ○ Real time inference with continuous learning ○ Interconnected recommendation (email, push, notification) Proprietary, Confidential, & Thoughtful Personalized Personalized 25
  • 26. 26 Proprietary, Confidential, & Thoughtful References ○ Engineering Blog : ■ https://medium.com/headspace-engineering/mindful-experiment ation-evaluate-recommendation-system-performance-using-a-b-t esting-at-headspace-3c8c05d0ae3b ○ CXL Guide ■ https://cxl.com/blog/ab-testing-guide/ Proprietary, Confidential, & Thoughtful
  • 27. 27 Proprietary, Confidential, & Thoughtful Thank you 27 https://www.linkedin.com/in/rohan3/