SlideShare a Scribd company logo
1 of 19
Northeastern University, 4 April 2018
Jen Wang
Wayfair Data Science Team, Projects, and Case Study --
Uplift Modeling for Driving Incremental Revenue in Display Remarketing
2
Wayfair: e-Commerce Tech Company
Our typical customer:
35 to 65 year old
woman with annual
household income of
$50k to $250k;
comScore median
household income of
$82k
3
A Clear Online Leader in Home Goods
OtherDirect Retail
4
Main goals of seminar
1. Data Science Team in Wayfair – How Data Science work is
organized at Wayfair, and the different types of projects we work on
2. Marketing Data Science – How Data Science projects are aligned
against different points of the marketing funnel
3. Case Studies in MKT DS – Uplift Modeling for Driving Incremental
Revenue in Display Remarketing
5
Data Science Team in Wayfair
6
Venn Diagram for Wayfair Data Science
Commonalities
across
companies
Trade-offs
Research
Application Engineering
• Develop & apply machine learning algorithms to find
answers to business problems
• Great range of algorithm complexity (from linear / logistic
regression to deep learning), but always need sufficient,
“big” data to get good results
• Standard set of technical tools, from R / Python for
scripting to Spark for big data processing
• Innovate by creating new algorithms /
approaches to solve problems
• Typically “bet big”, but doesn’t always pay
off
• Innovate by efficiently adapting existing
approaches to solve problems
• Can sometimes lead to more incremental
progress, tricky to build for long-term
• Typically use business rules & simpler
algorithms
• Focus on robustness & scalability first, then
modeling
Business
Problem
Solving
Engineering
Research /
Modeling /
ML
Warning:
No Unicorns!!!
Modeling
• Build “right” model first, then whittle away to get
in form ready for production
• Need to be mindful of 80/20
7
Data Science Groups at Wayfair
DS Infrastructure
DS Operation
Catalog
Optimization
NLP / CNN
Competitive
Intelligence
NLP
DS Marketing
Customer
Scoring &
Bidding
B2B
Uplift Model Text Mining
E.g. E.g.
Business
Problem
Solving
Engineering
Research /
Modeling /
ML
Business
Problem
Solving
Engineering
Research /
Modeling /
ML
DS Product
Recommenda-
-tion System
Visual Search
Reinforcement
learning / CF
E.g.
CNN
Business
Problem
Solving
Engineering
Research /
Modeling /
ML
8
Marketing Data Science: Business Problems
9
The key objective of Marketing Data Science is aligned to maximize return on marketing
investment by optimizing budget allocation, channel strategy and customer journey touchpoint
MKT Channel A
E.g. TV
MKT Channel B
E.g. Search
MKT Channel X
E.g. Retargeting
MKT
Budget Maximize MROI =
𝑅𝑒𝑣𝑒𝑛𝑢𝑒
𝐴𝑑 𝐶𝑜𝑠𝑡
By…
1. Guide the right budget allocation across
channels in our marketing portfolio
2. Provide channel level tactical guidance
towards delivering the right message to the
right customer through the right channel
3. Lineup the marketing treatments in the right
sequence and right cadence along the
customer life journey
Customer Life
Time Value
$
T1 T2 TN
Systematic View of Marketing Marketing DS Objective
10
(1/5) Example of Marketing Data Science Project
MKT Channel A
E.g. TV
MKT Channel B
E.g. Search
MKT Channel X
E.g. Retargeting
MKT
Budget
1. Guide the right budget allocation across
channels in our marketing portfolio
Q for DS:
How would you measure the revenue
contributed by different channels?
Customer Life
Time Value
$
T1 T2 TN
Order Attribution base on
Incremental Value
11
(2/5) Example of Marketing Data Science Project
MKT Channel A
E.g. TV
MKT Channel B
E.g. Search
MKT Channel X
E.g. Retargeting
MKT
Budget
2. Provide channel level tactical guidance
towards delivering the right message to the
right customer through the right channel
Q for DS:
• How to decide which TV channel
should be invested with more ads?
Customer Life
Time Value
$
T1
TV Targeting
T2 TN
12
(3/5) Example of Marketing Data Science Project
MKT Channel A
E.g. TV
MKT Channel B
E.g. Search
MKT Channel X
E.g. Retargeting
MKT
Budget
2. Provide channel level tactical guidance
towards delivering the right message to the
right customer through the right channel
Q for DS:
• How much we should bid on each
google keyword?
Customer Life
Time Value
$
T1
Keyword Bidding
T2 TN
13
(4/5) Example of Marketing Data Science Project
MKT Channel A
E.g. TV
MKT Channel B
E.g. Search
MKT Channel X
E.g. Retargeting
MKT
Budget
2. Provide channel level tactical guidance
towards delivering the right message to the
right customer through the right channel
Q for DS:
• How much we should bid the ads on
each customer?
Customer Life
Time Value
$
T1
Display Ads
T2 TN
14
(5/5) Example of Marketing Data Science Project
MKT Channel A
E.g. TV
MKT Channel B
E.g. Search
MKT Channel X
E.g. Retargeting
MKT
Budget
3. Schedule the marketing treatments in the
right sequence and right cadence along the
customer life journey
Q for DS:
• How often we should send the
marketing emails to customers?
Customer Life
Time Value
$
T1 T2 TN
15
Case Study:
Uplift Modeling to Drive Incremental Revenue in Display Remarketing
16
Customer Scoring
Uplift modeling for prediction of incremental revenue as base bid for each customer
𝒚=
Ad-inventory Scoring
Click-through rate prediction as bid modifier across Internet
Data Science Solutions
𝑦1 = P(buy | Ad) - P(buy | no Ad)
y2 = $Rev(buy | Ad)
Expected incr. Rev
Uplift
Base Bid
Case Study – Uplift Modeling in Display Remarketing
Display Remarketing
Why is it challenging?
• Billions of bidding opportunities across
Internet per day
• Real-time bidding
• Customer-level prediction
• Causal effects of Ad targeting?
*numbers are Illustrative only
Final Bid!
X
𝐶𝑇𝑅 =
𝐶𝑖𝑐𝑘𝑠
𝐼𝑚𝑝𝑟𝑒𝑠𝑠𝑖𝑜𝑛𝑠
Bid Modifier
17
Modeling and Evaluation
• Random Targeting to Collect Data
• Uplift Modeling
• Score (Predicted Uplift) = P(buy | Ad) - P(buy | no Ad)
• Uplift = Test CVR - Control CVR
Control Group:
PSA
Test Group:
Ads Seen
Case Study – Uplift Modeling in Display Remarketing
Background
• A method for modeling and predicting
causal effects
• Target most incremental (or persuadable)
customers
• Obama Camp persuaded millions of
voters with Uplift Modeling in 2012
Persuadables Sure Things
Lost Causes Sleeping Dogs
Will Convert if Not Treated
WillConvertifTreated
No Yes
NoYes
NumberofIncrementalCustomers
10 20 30 40 50 60 70 80 90 100
20 30
1040
01020
Number of Customers Targeted
&
Random Targeting
Perfect Uplift Model
Good Uplift Model
Perfect Conversion Model
Good Conversion Model
18
Jen Wang’s Journey to Data Science
Ph.D. in
(Biophysical)
Chemistry
Postdoc in
Drug Design
Health Data
Science
Fellow
Marketing
Data
Scientist
19
Wayfair: We Are Hiring!
• Wayfair DS Career: https://www.wayfaircareers.com/
• Wayfair DS Blog: http://tech.wayfair.com/category/data-science/
If you are interested in Wayfair data science...
Happy to answer any question you have…
• LinkedIn: https://www.linkedin.com/in/jenzhenwang/

More Related Content

What's hot

Uplift Modeling Workshop
Uplift Modeling WorkshopUplift Modeling Workshop
Uplift Modeling Workshopodsc
 
Misha Bilenko, Principal Researcher, Microsoft at MLconf SEA - 5/01/15
Misha Bilenko, Principal Researcher, Microsoft at MLconf SEA - 5/01/15Misha Bilenko, Principal Researcher, Microsoft at MLconf SEA - 5/01/15
Misha Bilenko, Principal Researcher, Microsoft at MLconf SEA - 5/01/15MLconf
 
Tips and tricks to win kaggle data science competitions
Tips and tricks to win kaggle data science competitionsTips and tricks to win kaggle data science competitions
Tips and tricks to win kaggle data science competitionsDarius Barušauskas
 
Kaggle presentation
Kaggle presentationKaggle presentation
Kaggle presentationHJ van Veen
 
Feature Engineering
Feature EngineeringFeature Engineering
Feature EngineeringHJ van Veen
 
Machine Learning for Q&A Sites: The Quora Example
Machine Learning for Q&A Sites: The Quora ExampleMachine Learning for Q&A Sites: The Quora Example
Machine Learning for Q&A Sites: The Quora ExampleXavier Amatriain
 
Counterfactual Learning for Recommendation
Counterfactual Learning for RecommendationCounterfactual Learning for Recommendation
Counterfactual Learning for RecommendationOlivier Jeunen
 
Why start using uplift models for more efficient marketing campaigns
Why start using uplift models for more efficient marketing campaignsWhy start using uplift models for more efficient marketing campaigns
Why start using uplift models for more efficient marketing campaignsData Con LA
 
Deeper Things: How Netflix Leverages Deep Learning in Recommendations and Se...
 Deeper Things: How Netflix Leverages Deep Learning in Recommendations and Se... Deeper Things: How Netflix Leverages Deep Learning in Recommendations and Se...
Deeper Things: How Netflix Leverages Deep Learning in Recommendations and Se...Sudeep Das, Ph.D.
 
Past, Present & Future of Recommender Systems: An Industry Perspective
Past, Present & Future of Recommender Systems: An Industry PerspectivePast, Present & Future of Recommender Systems: An Industry Perspective
Past, Present & Future of Recommender Systems: An Industry PerspectiveJustin Basilico
 
Personalizing "The Netflix Experience" with Deep Learning
Personalizing "The Netflix Experience" with Deep LearningPersonalizing "The Netflix Experience" with Deep Learning
Personalizing "The Netflix Experience" with Deep LearningAnoop Deoras
 
Machine Learning Strategies for Time Series Prediction
Machine Learning Strategies for Time Series PredictionMachine Learning Strategies for Time Series Prediction
Machine Learning Strategies for Time Series PredictionGianluca Bontempi
 
Rishabh Mehrotra - Recommendations in a Marketplace: Personalizing Explainabl...
Rishabh Mehrotra - Recommendations in a Marketplace: Personalizing Explainabl...Rishabh Mehrotra - Recommendations in a Marketplace: Personalizing Explainabl...
Rishabh Mehrotra - Recommendations in a Marketplace: Personalizing Explainabl...MLconf
 
Sequential Decision Making in Recommendations
Sequential Decision Making in RecommendationsSequential Decision Making in Recommendations
Sequential Decision Making in RecommendationsJaya Kawale
 
Churn prediction data modeling
Churn prediction data modelingChurn prediction data modeling
Churn prediction data modelingPierre Gutierrez
 
Time, Context and Causality in Recommender Systems
Time, Context and Causality in Recommender SystemsTime, Context and Causality in Recommender Systems
Time, Context and Causality in Recommender SystemsYves Raimond
 
Missing values in recommender models
Missing values in recommender modelsMissing values in recommender models
Missing values in recommender modelsParmeshwar Khurd
 
Recommender system introduction
Recommender system   introductionRecommender system   introduction
Recommender system introductionLiang Xiang
 
The How and Why of Feature Engineering
The How and Why of Feature EngineeringThe How and Why of Feature Engineering
The How and Why of Feature EngineeringAlice Zheng
 

What's hot (20)

Uplift Modeling Workshop
Uplift Modeling WorkshopUplift Modeling Workshop
Uplift Modeling Workshop
 
Customer Analytics Best Practice
Customer Analytics Best PracticeCustomer Analytics Best Practice
Customer Analytics Best Practice
 
Misha Bilenko, Principal Researcher, Microsoft at MLconf SEA - 5/01/15
Misha Bilenko, Principal Researcher, Microsoft at MLconf SEA - 5/01/15Misha Bilenko, Principal Researcher, Microsoft at MLconf SEA - 5/01/15
Misha Bilenko, Principal Researcher, Microsoft at MLconf SEA - 5/01/15
 
Tips and tricks to win kaggle data science competitions
Tips and tricks to win kaggle data science competitionsTips and tricks to win kaggle data science competitions
Tips and tricks to win kaggle data science competitions
 
Kaggle presentation
Kaggle presentationKaggle presentation
Kaggle presentation
 
Feature Engineering
Feature EngineeringFeature Engineering
Feature Engineering
 
Machine Learning for Q&A Sites: The Quora Example
Machine Learning for Q&A Sites: The Quora ExampleMachine Learning for Q&A Sites: The Quora Example
Machine Learning for Q&A Sites: The Quora Example
 
Counterfactual Learning for Recommendation
Counterfactual Learning for RecommendationCounterfactual Learning for Recommendation
Counterfactual Learning for Recommendation
 
Why start using uplift models for more efficient marketing campaigns
Why start using uplift models for more efficient marketing campaignsWhy start using uplift models for more efficient marketing campaigns
Why start using uplift models for more efficient marketing campaigns
 
Deeper Things: How Netflix Leverages Deep Learning in Recommendations and Se...
 Deeper Things: How Netflix Leverages Deep Learning in Recommendations and Se... Deeper Things: How Netflix Leverages Deep Learning in Recommendations and Se...
Deeper Things: How Netflix Leverages Deep Learning in Recommendations and Se...
 
Past, Present & Future of Recommender Systems: An Industry Perspective
Past, Present & Future of Recommender Systems: An Industry PerspectivePast, Present & Future of Recommender Systems: An Industry Perspective
Past, Present & Future of Recommender Systems: An Industry Perspective
 
Personalizing "The Netflix Experience" with Deep Learning
Personalizing "The Netflix Experience" with Deep LearningPersonalizing "The Netflix Experience" with Deep Learning
Personalizing "The Netflix Experience" with Deep Learning
 
Machine Learning Strategies for Time Series Prediction
Machine Learning Strategies for Time Series PredictionMachine Learning Strategies for Time Series Prediction
Machine Learning Strategies for Time Series Prediction
 
Rishabh Mehrotra - Recommendations in a Marketplace: Personalizing Explainabl...
Rishabh Mehrotra - Recommendations in a Marketplace: Personalizing Explainabl...Rishabh Mehrotra - Recommendations in a Marketplace: Personalizing Explainabl...
Rishabh Mehrotra - Recommendations in a Marketplace: Personalizing Explainabl...
 
Sequential Decision Making in Recommendations
Sequential Decision Making in RecommendationsSequential Decision Making in Recommendations
Sequential Decision Making in Recommendations
 
Churn prediction data modeling
Churn prediction data modelingChurn prediction data modeling
Churn prediction data modeling
 
Time, Context and Causality in Recommender Systems
Time, Context and Causality in Recommender SystemsTime, Context and Causality in Recommender Systems
Time, Context and Causality in Recommender Systems
 
Missing values in recommender models
Missing values in recommender modelsMissing values in recommender models
Missing values in recommender models
 
Recommender system introduction
Recommender system   introductionRecommender system   introduction
Recommender system introduction
 
The How and Why of Feature Engineering
The How and Why of Feature EngineeringThe How and Why of Feature Engineering
The How and Why of Feature Engineering
 

Similar to Wayfair's Data Science Team and Case Study: Uplift Modeling

Big Data LDN 2017: Advanced Analytics Applied to Marketing Attribution
Big Data LDN 2017: Advanced Analytics Applied to Marketing AttributionBig Data LDN 2017: Advanced Analytics Applied to Marketing Attribution
Big Data LDN 2017: Advanced Analytics Applied to Marketing AttributionMatt Stubbs
 
Proposal For Analyzing Organizational Process Bottlenecks PowerPoint Presenta...
Proposal For Analyzing Organizational Process Bottlenecks PowerPoint Presenta...Proposal For Analyzing Organizational Process Bottlenecks PowerPoint Presenta...
Proposal For Analyzing Organizational Process Bottlenecks PowerPoint Presenta...SlideTeam
 
Proposal For Measuring Business Performance PowerPoint Presentation Slides
Proposal For Measuring Business Performance PowerPoint Presentation SlidesProposal For Measuring Business Performance PowerPoint Presentation Slides
Proposal For Measuring Business Performance PowerPoint Presentation SlidesSlideTeam
 
Rethinking Marketing: New Roles, Responsibilities and Reports
Rethinking Marketing: New Roles, Responsibilities and ReportsRethinking Marketing: New Roles, Responsibilities and Reports
Rethinking Marketing: New Roles, Responsibilities and ReportsG3 Communications
 
The Age Of New Reality Marketing V5.1 Final
The Age Of New Reality Marketing V5.1 FinalThe Age Of New Reality Marketing V5.1 Final
The Age Of New Reality Marketing V5.1 FinalTony Mooney
 
Embedded analytics and digital transformation
Embedded analytics and digital transformationEmbedded analytics and digital transformation
Embedded analytics and digital transformationGuha Athreya
 
WE and Belgium ICT buying power
WE and Belgium ICT buying powerWE and Belgium ICT buying power
WE and Belgium ICT buying powerDidier Andrieu
 
Dataiku tatvic webinar presentation
Dataiku tatvic webinar presentationDataiku tatvic webinar presentation
Dataiku tatvic webinar presentationTatvic Analytics
 
Business analytics -Abhay Mahalley
Business analytics -Abhay MahalleyBusiness analytics -Abhay Mahalley
Business analytics -Abhay MahalleyAbhay Mahalley
 
Creating a Demand Generation Budget From Scratch
Creating a Demand Generation Budget From ScratchCreating a Demand Generation Budget From Scratch
Creating a Demand Generation Budget From ScratchSaasMQL
 
E-commerce Berlin Expo 2018 - How to boost your online sales using machine le...
E-commerce Berlin Expo 2018 - How to boost your online sales using machine le...E-commerce Berlin Expo 2018 - How to boost your online sales using machine le...
E-commerce Berlin Expo 2018 - How to boost your online sales using machine le...E-Commerce Berlin EXPO
 
Digital Marketing Playbook - How to create scalable, predictable revenue
Digital Marketing Playbook - How to create scalable, predictable revenueDigital Marketing Playbook - How to create scalable, predictable revenue
Digital Marketing Playbook - How to create scalable, predictable revenueSeth Hauben
 
The Always-On Approach: How to Continually Improve Your Streaming Advertising...
The Always-On Approach: How to Continually Improve Your Streaming Advertising...The Always-On Approach: How to Continually Improve Your Streaming Advertising...
The Always-On Approach: How to Continually Improve Your Streaming Advertising...Tinuiti
 
nextNY Online Marketing School - SEM Presentation
nextNY Online Marketing School - SEM PresentationnextNY Online Marketing School - SEM Presentation
nextNY Online Marketing School - SEM PresentationnextNY
 
Marketing Plan Of Energypac Engineering Limited
Marketing Plan Of Energypac Engineering LimitedMarketing Plan Of Energypac Engineering Limited
Marketing Plan Of Energypac Engineering Limitedsample_m2000
 
New Madison Ave: Data & Marketing Technology Solutions – April 2015
New Madison Ave: Data & Marketing Technology Solutions – April 2015New Madison Ave: Data & Marketing Technology Solutions – April 2015
New Madison Ave: Data & Marketing Technology Solutions – April 2015New Madison Ave
 
Database Marketing, part two: data enhancement, analytics, and attribution
Database Marketing, part two: data enhancement, analytics, and attribution Database Marketing, part two: data enhancement, analytics, and attribution
Database Marketing, part two: data enhancement, analytics, and attribution Relevate
 

Similar to Wayfair's Data Science Team and Case Study: Uplift Modeling (20)

Big Data LDN 2017: Advanced Analytics Applied to Marketing Attribution
Big Data LDN 2017: Advanced Analytics Applied to Marketing AttributionBig Data LDN 2017: Advanced Analytics Applied to Marketing Attribution
Big Data LDN 2017: Advanced Analytics Applied to Marketing Attribution
 
Proposal For Analyzing Organizational Process Bottlenecks PowerPoint Presenta...
Proposal For Analyzing Organizational Process Bottlenecks PowerPoint Presenta...Proposal For Analyzing Organizational Process Bottlenecks PowerPoint Presenta...
Proposal For Analyzing Organizational Process Bottlenecks PowerPoint Presenta...
 
Proposal For Measuring Business Performance PowerPoint Presentation Slides
Proposal For Measuring Business Performance PowerPoint Presentation SlidesProposal For Measuring Business Performance PowerPoint Presentation Slides
Proposal For Measuring Business Performance PowerPoint Presentation Slides
 
Rethinking Marketing: New Roles, Responsibilities and Reports
Rethinking Marketing: New Roles, Responsibilities and ReportsRethinking Marketing: New Roles, Responsibilities and Reports
Rethinking Marketing: New Roles, Responsibilities and Reports
 
The Age Of New Reality Marketing V5.1 Final
The Age Of New Reality Marketing V5.1 FinalThe Age Of New Reality Marketing V5.1 Final
The Age Of New Reality Marketing V5.1 Final
 
Building a Data Driven Business
Building a Data Driven BusinessBuilding a Data Driven Business
Building a Data Driven Business
 
Embedded analytics and digital transformation
Embedded analytics and digital transformationEmbedded analytics and digital transformation
Embedded analytics and digital transformation
 
WE and Belgium ICT buying power
WE and Belgium ICT buying powerWE and Belgium ICT buying power
WE and Belgium ICT buying power
 
Dataiku tatvic webinar presentation
Dataiku tatvic webinar presentationDataiku tatvic webinar presentation
Dataiku tatvic webinar presentation
 
Business analytics -Abhay Mahalley
Business analytics -Abhay MahalleyBusiness analytics -Abhay Mahalley
Business analytics -Abhay Mahalley
 
Business analytics !!
Business analytics !!Business analytics !!
Business analytics !!
 
Business analytics
Business analytics Business analytics
Business analytics
 
Creating a Demand Generation Budget From Scratch
Creating a Demand Generation Budget From ScratchCreating a Demand Generation Budget From Scratch
Creating a Demand Generation Budget From Scratch
 
E-commerce Berlin Expo 2018 - How to boost your online sales using machine le...
E-commerce Berlin Expo 2018 - How to boost your online sales using machine le...E-commerce Berlin Expo 2018 - How to boost your online sales using machine le...
E-commerce Berlin Expo 2018 - How to boost your online sales using machine le...
 
Digital Marketing Playbook - How to create scalable, predictable revenue
Digital Marketing Playbook - How to create scalable, predictable revenueDigital Marketing Playbook - How to create scalable, predictable revenue
Digital Marketing Playbook - How to create scalable, predictable revenue
 
The Always-On Approach: How to Continually Improve Your Streaming Advertising...
The Always-On Approach: How to Continually Improve Your Streaming Advertising...The Always-On Approach: How to Continually Improve Your Streaming Advertising...
The Always-On Approach: How to Continually Improve Your Streaming Advertising...
 
nextNY Online Marketing School - SEM Presentation
nextNY Online Marketing School - SEM PresentationnextNY Online Marketing School - SEM Presentation
nextNY Online Marketing School - SEM Presentation
 
Marketing Plan Of Energypac Engineering Limited
Marketing Plan Of Energypac Engineering LimitedMarketing Plan Of Energypac Engineering Limited
Marketing Plan Of Energypac Engineering Limited
 
New Madison Ave: Data & Marketing Technology Solutions – April 2015
New Madison Ave: Data & Marketing Technology Solutions – April 2015New Madison Ave: Data & Marketing Technology Solutions – April 2015
New Madison Ave: Data & Marketing Technology Solutions – April 2015
 
Database Marketing, part two: data enhancement, analytics, and attribution
Database Marketing, part two: data enhancement, analytics, and attribution Database Marketing, part two: data enhancement, analytics, and attribution
Database Marketing, part two: data enhancement, analytics, and attribution
 

Recently uploaded

Vector Search -An Introduction in Oracle Database 23ai.pptx
Vector Search -An Introduction in Oracle Database 23ai.pptxVector Search -An Introduction in Oracle Database 23ai.pptx
Vector Search -An Introduction in Oracle Database 23ai.pptxRemote DBA Services
 
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost SavingRepurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost SavingEdi Saputra
 
How to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected WorkerHow to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected WorkerThousandEyes
 
MINDCTI Revenue Release Quarter One 2024
MINDCTI Revenue Release Quarter One 2024MINDCTI Revenue Release Quarter One 2024
MINDCTI Revenue Release Quarter One 2024MIND CTI
 
EMPOWERMENT TECHNOLOGY GRADE 11 QUARTER 2 REVIEWER
EMPOWERMENT TECHNOLOGY GRADE 11 QUARTER 2 REVIEWEREMPOWERMENT TECHNOLOGY GRADE 11 QUARTER 2 REVIEWER
EMPOWERMENT TECHNOLOGY GRADE 11 QUARTER 2 REVIEWERMadyBayot
 
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers:  A Deep Dive into Serverless Spatial Data and FMECloud Frontiers:  A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FMESafe Software
 
Artificial Intelligence Chap.5 : Uncertainty
Artificial Intelligence Chap.5 : UncertaintyArtificial Intelligence Chap.5 : Uncertainty
Artificial Intelligence Chap.5 : UncertaintyKhushali Kathiriya
 
ICT role in 21st century education and its challenges
ICT role in 21st century education and its challengesICT role in 21st century education and its challenges
ICT role in 21st century education and its challengesrafiqahmad00786416
 
WSO2's API Vision: Unifying Control, Empowering Developers
WSO2's API Vision: Unifying Control, Empowering DevelopersWSO2's API Vision: Unifying Control, Empowering Developers
WSO2's API Vision: Unifying Control, Empowering DevelopersWSO2
 
Exploring Multimodal Embeddings with Milvus
Exploring Multimodal Embeddings with MilvusExploring Multimodal Embeddings with Milvus
Exploring Multimodal Embeddings with MilvusZilliz
 
Finding Java's Hidden Performance Traps @ DevoxxUK 2024
Finding Java's Hidden Performance Traps @ DevoxxUK 2024Finding Java's Hidden Performance Traps @ DevoxxUK 2024
Finding Java's Hidden Performance Traps @ DevoxxUK 2024Victor Rentea
 
Architecting Cloud Native Applications
Architecting Cloud Native ApplicationsArchitecting Cloud Native Applications
Architecting Cloud Native ApplicationsWSO2
 
Platformless Horizons for Digital Adaptability
Platformless Horizons for Digital AdaptabilityPlatformless Horizons for Digital Adaptability
Platformless Horizons for Digital AdaptabilityWSO2
 
TrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
TrustArc Webinar - Unlock the Power of AI-Driven Data DiscoveryTrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
TrustArc Webinar - Unlock the Power of AI-Driven Data DiscoveryTrustArc
 
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, AdobeApidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobeapidays
 
presentation ICT roal in 21st century education
presentation ICT roal in 21st century educationpresentation ICT roal in 21st century education
presentation ICT roal in 21st century educationjfdjdjcjdnsjd
 
"I see eyes in my soup": How Delivery Hero implemented the safety system for ...
"I see eyes in my soup": How Delivery Hero implemented the safety system for ..."I see eyes in my soup": How Delivery Hero implemented the safety system for ...
"I see eyes in my soup": How Delivery Hero implemented the safety system for ...Zilliz
 
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers:  A Deep Dive into Serverless Spatial Data and FMECloud Frontiers:  A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FMESafe Software
 
FWD Group - Insurer Innovation Award 2024
FWD Group - Insurer Innovation Award 2024FWD Group - Insurer Innovation Award 2024
FWD Group - Insurer Innovation Award 2024The Digital Insurer
 

Recently uploaded (20)

Vector Search -An Introduction in Oracle Database 23ai.pptx
Vector Search -An Introduction in Oracle Database 23ai.pptxVector Search -An Introduction in Oracle Database 23ai.pptx
Vector Search -An Introduction in Oracle Database 23ai.pptx
 
Understanding the FAA Part 107 License ..
Understanding the FAA Part 107 License ..Understanding the FAA Part 107 License ..
Understanding the FAA Part 107 License ..
 
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost SavingRepurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
 
How to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected WorkerHow to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected Worker
 
MINDCTI Revenue Release Quarter One 2024
MINDCTI Revenue Release Quarter One 2024MINDCTI Revenue Release Quarter One 2024
MINDCTI Revenue Release Quarter One 2024
 
EMPOWERMENT TECHNOLOGY GRADE 11 QUARTER 2 REVIEWER
EMPOWERMENT TECHNOLOGY GRADE 11 QUARTER 2 REVIEWEREMPOWERMENT TECHNOLOGY GRADE 11 QUARTER 2 REVIEWER
EMPOWERMENT TECHNOLOGY GRADE 11 QUARTER 2 REVIEWER
 
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers:  A Deep Dive into Serverless Spatial Data and FMECloud Frontiers:  A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
 
Artificial Intelligence Chap.5 : Uncertainty
Artificial Intelligence Chap.5 : UncertaintyArtificial Intelligence Chap.5 : Uncertainty
Artificial Intelligence Chap.5 : Uncertainty
 
ICT role in 21st century education and its challenges
ICT role in 21st century education and its challengesICT role in 21st century education and its challenges
ICT role in 21st century education and its challenges
 
WSO2's API Vision: Unifying Control, Empowering Developers
WSO2's API Vision: Unifying Control, Empowering DevelopersWSO2's API Vision: Unifying Control, Empowering Developers
WSO2's API Vision: Unifying Control, Empowering Developers
 
Exploring Multimodal Embeddings with Milvus
Exploring Multimodal Embeddings with MilvusExploring Multimodal Embeddings with Milvus
Exploring Multimodal Embeddings with Milvus
 
Finding Java's Hidden Performance Traps @ DevoxxUK 2024
Finding Java's Hidden Performance Traps @ DevoxxUK 2024Finding Java's Hidden Performance Traps @ DevoxxUK 2024
Finding Java's Hidden Performance Traps @ DevoxxUK 2024
 
Architecting Cloud Native Applications
Architecting Cloud Native ApplicationsArchitecting Cloud Native Applications
Architecting Cloud Native Applications
 
Platformless Horizons for Digital Adaptability
Platformless Horizons for Digital AdaptabilityPlatformless Horizons for Digital Adaptability
Platformless Horizons for Digital Adaptability
 
TrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
TrustArc Webinar - Unlock the Power of AI-Driven Data DiscoveryTrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
TrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
 
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, AdobeApidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
 
presentation ICT roal in 21st century education
presentation ICT roal in 21st century educationpresentation ICT roal in 21st century education
presentation ICT roal in 21st century education
 
"I see eyes in my soup": How Delivery Hero implemented the safety system for ...
"I see eyes in my soup": How Delivery Hero implemented the safety system for ..."I see eyes in my soup": How Delivery Hero implemented the safety system for ...
"I see eyes in my soup": How Delivery Hero implemented the safety system for ...
 
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers:  A Deep Dive into Serverless Spatial Data and FMECloud Frontiers:  A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
 
FWD Group - Insurer Innovation Award 2024
FWD Group - Insurer Innovation Award 2024FWD Group - Insurer Innovation Award 2024
FWD Group - Insurer Innovation Award 2024
 

Wayfair's Data Science Team and Case Study: Uplift Modeling

  • 1. Northeastern University, 4 April 2018 Jen Wang Wayfair Data Science Team, Projects, and Case Study -- Uplift Modeling for Driving Incremental Revenue in Display Remarketing
  • 2. 2 Wayfair: e-Commerce Tech Company Our typical customer: 35 to 65 year old woman with annual household income of $50k to $250k; comScore median household income of $82k
  • 3. 3 A Clear Online Leader in Home Goods OtherDirect Retail
  • 4. 4 Main goals of seminar 1. Data Science Team in Wayfair – How Data Science work is organized at Wayfair, and the different types of projects we work on 2. Marketing Data Science – How Data Science projects are aligned against different points of the marketing funnel 3. Case Studies in MKT DS – Uplift Modeling for Driving Incremental Revenue in Display Remarketing
  • 5. 5 Data Science Team in Wayfair
  • 6. 6 Venn Diagram for Wayfair Data Science Commonalities across companies Trade-offs Research Application Engineering • Develop & apply machine learning algorithms to find answers to business problems • Great range of algorithm complexity (from linear / logistic regression to deep learning), but always need sufficient, “big” data to get good results • Standard set of technical tools, from R / Python for scripting to Spark for big data processing • Innovate by creating new algorithms / approaches to solve problems • Typically “bet big”, but doesn’t always pay off • Innovate by efficiently adapting existing approaches to solve problems • Can sometimes lead to more incremental progress, tricky to build for long-term • Typically use business rules & simpler algorithms • Focus on robustness & scalability first, then modeling Business Problem Solving Engineering Research / Modeling / ML Warning: No Unicorns!!! Modeling • Build “right” model first, then whittle away to get in form ready for production • Need to be mindful of 80/20
  • 7. 7 Data Science Groups at Wayfair DS Infrastructure DS Operation Catalog Optimization NLP / CNN Competitive Intelligence NLP DS Marketing Customer Scoring & Bidding B2B Uplift Model Text Mining E.g. E.g. Business Problem Solving Engineering Research / Modeling / ML Business Problem Solving Engineering Research / Modeling / ML DS Product Recommenda- -tion System Visual Search Reinforcement learning / CF E.g. CNN Business Problem Solving Engineering Research / Modeling / ML
  • 8. 8 Marketing Data Science: Business Problems
  • 9. 9 The key objective of Marketing Data Science is aligned to maximize return on marketing investment by optimizing budget allocation, channel strategy and customer journey touchpoint MKT Channel A E.g. TV MKT Channel B E.g. Search MKT Channel X E.g. Retargeting MKT Budget Maximize MROI = 𝑅𝑒𝑣𝑒𝑛𝑢𝑒 𝐴𝑑 𝐶𝑜𝑠𝑡 By… 1. Guide the right budget allocation across channels in our marketing portfolio 2. Provide channel level tactical guidance towards delivering the right message to the right customer through the right channel 3. Lineup the marketing treatments in the right sequence and right cadence along the customer life journey Customer Life Time Value $ T1 T2 TN Systematic View of Marketing Marketing DS Objective
  • 10. 10 (1/5) Example of Marketing Data Science Project MKT Channel A E.g. TV MKT Channel B E.g. Search MKT Channel X E.g. Retargeting MKT Budget 1. Guide the right budget allocation across channels in our marketing portfolio Q for DS: How would you measure the revenue contributed by different channels? Customer Life Time Value $ T1 T2 TN Order Attribution base on Incremental Value
  • 11. 11 (2/5) Example of Marketing Data Science Project MKT Channel A E.g. TV MKT Channel B E.g. Search MKT Channel X E.g. Retargeting MKT Budget 2. Provide channel level tactical guidance towards delivering the right message to the right customer through the right channel Q for DS: • How to decide which TV channel should be invested with more ads? Customer Life Time Value $ T1 TV Targeting T2 TN
  • 12. 12 (3/5) Example of Marketing Data Science Project MKT Channel A E.g. TV MKT Channel B E.g. Search MKT Channel X E.g. Retargeting MKT Budget 2. Provide channel level tactical guidance towards delivering the right message to the right customer through the right channel Q for DS: • How much we should bid on each google keyword? Customer Life Time Value $ T1 Keyword Bidding T2 TN
  • 13. 13 (4/5) Example of Marketing Data Science Project MKT Channel A E.g. TV MKT Channel B E.g. Search MKT Channel X E.g. Retargeting MKT Budget 2. Provide channel level tactical guidance towards delivering the right message to the right customer through the right channel Q for DS: • How much we should bid the ads on each customer? Customer Life Time Value $ T1 Display Ads T2 TN
  • 14. 14 (5/5) Example of Marketing Data Science Project MKT Channel A E.g. TV MKT Channel B E.g. Search MKT Channel X E.g. Retargeting MKT Budget 3. Schedule the marketing treatments in the right sequence and right cadence along the customer life journey Q for DS: • How often we should send the marketing emails to customers? Customer Life Time Value $ T1 T2 TN
  • 15. 15 Case Study: Uplift Modeling to Drive Incremental Revenue in Display Remarketing
  • 16. 16 Customer Scoring Uplift modeling for prediction of incremental revenue as base bid for each customer 𝒚= Ad-inventory Scoring Click-through rate prediction as bid modifier across Internet Data Science Solutions 𝑦1 = P(buy | Ad) - P(buy | no Ad) y2 = $Rev(buy | Ad) Expected incr. Rev Uplift Base Bid Case Study – Uplift Modeling in Display Remarketing Display Remarketing Why is it challenging? • Billions of bidding opportunities across Internet per day • Real-time bidding • Customer-level prediction • Causal effects of Ad targeting? *numbers are Illustrative only Final Bid! X 𝐶𝑇𝑅 = 𝐶𝑖𝑐𝑘𝑠 𝐼𝑚𝑝𝑟𝑒𝑠𝑠𝑖𝑜𝑛𝑠 Bid Modifier
  • 17. 17 Modeling and Evaluation • Random Targeting to Collect Data • Uplift Modeling • Score (Predicted Uplift) = P(buy | Ad) - P(buy | no Ad) • Uplift = Test CVR - Control CVR Control Group: PSA Test Group: Ads Seen Case Study – Uplift Modeling in Display Remarketing Background • A method for modeling and predicting causal effects • Target most incremental (or persuadable) customers • Obama Camp persuaded millions of voters with Uplift Modeling in 2012 Persuadables Sure Things Lost Causes Sleeping Dogs Will Convert if Not Treated WillConvertifTreated No Yes NoYes NumberofIncrementalCustomers 10 20 30 40 50 60 70 80 90 100 20 30 1040 01020 Number of Customers Targeted & Random Targeting Perfect Uplift Model Good Uplift Model Perfect Conversion Model Good Conversion Model
  • 18. 18 Jen Wang’s Journey to Data Science Ph.D. in (Biophysical) Chemistry Postdoc in Drug Design Health Data Science Fellow Marketing Data Scientist
  • 19. 19 Wayfair: We Are Hiring! • Wayfair DS Career: https://www.wayfaircareers.com/ • Wayfair DS Blog: http://tech.wayfair.com/category/data-science/ If you are interested in Wayfair data science... Happy to answer any question you have… • LinkedIn: https://www.linkedin.com/in/jenzhenwang/