This document discusses how data science can catalyze digital transformation. It begins by defining data science and digital transformation, noting that digital transformation provides the foundation for data science enablement. Several challenges of digital transformation are outlined, including increasing competition, changing consumer behavior, and legacy technical and cultural issues. The presentation argues that data science can help address these challenges by leveraging customer data to personalize engagement across channels. Specific data science techniques are described, such as propensity modeling, content personalization, social media interaction analysis, text analysis, and image/video analysis. Current limitations of these approaches are acknowledged, and the document concludes by emphasizing the relationship between data science and digital transformation.
2. Presenters
ManChon (Kevin) U, PhD
Head of Marketing Analytics & Data Science, Carnival Cruise Line
KU@Carnival.com
Marc Fridson
Principal Data Scientist, Carnival Cruise Line
MFridson@carnival.com
3. Agenda
● Definitions
● Challenges
○ Business
○ Technical
○ Cultural
● Why Data Science & Why Digital Transformation?
● Leverage Data Science during Digital Transformation
● How?
○ Propensity Modeling
○ Content Personalization
○ Social Ecosystem Interaction
○ Text Analysis
○ Image and Video Analysis
● Current Limitations: Reality vs Hype
● Conclusion
4. Relationship between DS and DT
● Data Science
○ Technical Skills
○ Mathematics + Statistics
Expertise
○ Business Knowledge
● Digital Transformation
○ Application of digital
technologies in all aspects to
enable new types of innovation
and creativity in a particular
domain.
Digital Transformation is the
foundation, while Data Science
is the media for enablement!
6. Challenges - Business
● Increasing competitions
● Too much data/ signal/ information
● Consumer Behavior Is Changing
● Question: How can we earn (and re-
earn) each customer’s consideration
in every micro-moment throughout the
customer journey?
○ Acquisition: Increased Efficiency Through
Better Targeting
■ Be in the right place, at the right
time, and be useful.
○ Conversion: Drive Revenue Through
Relevance
■ Deliver the right message, to the
right person, at the right time.
7. Challenges - Technical
● Legacy Data Processing Framework
○ No data governance process
○ No single source of truth, data marts from each department solely serve their own
purposes
○ No aggregated view of data from various data sources
○ No aggregated view of customers from multiple systems (i.e. Hotel, Gaming, Web, Social,
E-Mail, etc.)
○ RDBMS only data storage heavily rely on ETL, nearly no flexibility to business users
● Legacy Data Engineering & Analytics Environment
○ No centralized environment for data engineering, let alone in large scale (i.e.
Ingestion, ETL/ELT, etc.)
○ No state-of-the-art platform for advanced analytics or for modeling (i.e. distributed
computing framework)
○ No centralized platform to develop and deploy models
● Lack of Computational Power
○ On-premises VS. Cloud
○ No capacity to process data in a timely manner and/or to perform deep analytics on data
at a granular level, let alone for building predictive models on such scale
8. Challenges - Cultural
● Legacy Skill Sets
○ Excel/SAS only
● Afraid of New Technology/Change
○ “My SAS program is just fine… it takes only 24 hours to get a number”
● Experience-Driven VS Data-Driven
○ “I’ve been doing this for years…”
○ “Although the data says so, but…”
10. Leveraging Data Science during Digital Transformation
● Capture customer preference
○ From multiple channels and synchronize the updates
● Leverage insights to form an executable contact strategy
○ Leverage the insights derived from preferences data and past web,
email, and social campaign activities into an executable contact
strategy.
● Build aggregated view of customer profiles (Customer DNA)
○ Aggregate customer profiles from multi-sources
● Perform deep analytics
○ Ability to perform profiling/customer analytics with Customer DNA
● Build advanced predictive models
○ Alert detection (i.e. loyalty, churn, cross-sell opportunities)
● Make decision on next-best-action (Both inbound and
outbound)
○ Whom to offer? What to offer? When to offer? How to offer?
12. Multichannel Engagement
● Engagement across multiple channels is key, and
understanding how to leverage emerging ones is essential
○ Baby Boomers are mostly engaged through traditional methods like mail
■ Since this is the most expensive marketing channel, propensity
modeling is important to identify most likely customers to
maximize conversion rate
○ Generation X and Millennials primary forms of engagement are email
and social media
■ In email, customization of an email body and message frequency
important are critical factors for continued engagement and
potential conversion
■ Facebook knows user demographics and interests, this allows
marketers to target individuals with very specific likes and
provides a platform for them to be directly engaged (e.g.,
Chatbots) and referred directly to a revenue generating
transaction
13. Multichannel Engagement Cont’d
○ Generation Z is primarily engaged through a wide range of social
media channels
■ Snapchat (Created Experiences)
■ Instagram (Captured Moments)
■ Facebook (Acquaintance Updates and Communication)
■ Twitter (Information from Interests and Influencers)
14. Propensity Modeling
● New Customers
○ Calculating the Projecting Lifetime Value of a Customer
■ Given the characteristics of an individual what is the probability of an offer
being successful at acquiring the customer (Bayesian Inference)
■ Based on buying behavior exhibited by our existing customer features, what is the
estimated lifetime cruise spend
● Lifetime Spend =expected $ spent per booking X # of lifetime bookings
■ Score and sort top prospective customers
● Score potential customers by acquisition probability X $ Lifetime Spend
● Sort from greatest to least
● Include from prospect #1 to the number of customers your budget will allow
○ Existing Customers
■ Based on the customer’s previous buying history what is the likelihood that a
current offer will persuade them to make a purchase they otherwise would not have
made
■ Calculate the additional revenue gained by targeting the customer with the offer
versus if they had not (i.e., would they have booked anyway had they not received
the offer?)
15. Content Personalization
● Communication Frequency
● Demographic Analysis
● Previous booking patterns
○ # of cabins
○ Room type: interior, balcony, suite
○ Average cabin price
● Content Focus/Emphasis
● Activity recommendations
○ Shore Excursions
○ Drink Packages
○ Upgrades
17. Social Ecosystem Interaction
● Average social media usage is in excess of 2 hours per
day
● Active and background usage of social media platforms
provides new mechanisms to attract new customers and to
better engage existing ones
● Targeting social influencers, allows companies to
leverage the network effect as a catalyst for marketing
campaigns
● All of these platforms offer some sort of messaging
component
○ Facebook Messenger
○ Twitter (Tweets, Direct Messaging)
18. Social Ecosystem Interaction Cont’d
● This provides companies with an opportunity for stronger
engagement with the customer and the potential to
automate manual research and customer service
● There have been major strides in AI and Natural Language
Processing (NLP) due to advances in computational power
○ Best in class Voice User Interfaces (VUI) from the leaders in the
space such as Amazon, Google and Facebook are far more limited than
people realize
○ Existing VUIs are very systematic when it comes to how to interact
with them
■ One of the reasons you need to specify a skill to use with Alexa,
is because it needs context for your intent
■ Customers want a smooth interaction and have human level
expectations when dealing with AI bots, which can lead to
dissatisfaction and attrition
19. Image and Video Analysis
● Object Detection
● Scene Detection
● Activity Detection
● Facial Analysis
Work in Progress
21. Thank You, We Are Hiring!
ManChon (Kevin) U, PhD
Head of Marketing Analytics & Data Science, Carnival Cruise Line
KU@Carnival.com
Marc Fridson
Principal Data Scientist, Carnival Cruise Line
MFridson@carnival.com