How to Build AI based

Personalization at Scale

Gautam Gupta

Technology Leader

AI / ML / Cloud Focused

Intuit

Work on Data @Scale

in/guptagautam/

What is Personalization?

• Personalization is the process of tailoring product/service to individual users' characteristics or preferences.

• A personalized experience establishes goodwill, increases loyalty, and brings a smile to your face.

Why Personalization?

The goal of personalization is not just cross-selling, up-selling or more sales. It is not about following customer to put right thing at right place at the right time.

The goal of personalization is to make it easier for customers

to find & consume what they want, how they want and when they want.

This Photo by Unknown Author is licensed under CC BY-SA

What is Hyper-Personalization?

• Hyper-personalization takes personalized marketing a step further by leveraging artificial intelligence (AI) and real-time data to deliver more relevant content, product, and service information to each user.

Image Credit: https://www.capgemini.com/consulting-fr/wp-content/uploads/sites/31/2017/08/hyperpersonnalisation_vs_segmentation_english_05-01-2017.pdf

How does Personalization work?

• Product elements are tailored to specifically target each individual consumer.

This Photo by Unknown Author is licensed under CC BY-SA

This Photo by Unknown Author is licensed under CC BY-NC-ND

Personalization Cycle

Product

Personalize

Consume Delight

Repeat

Benefits of Personalization

• Meet the customer's needs more effectively and efficiently

• Make interactions faster and easier

• Increase customer satisfaction

• More likelihood of repeat visits

Personalization: Traditional Approach

Demographic Content

Transactional data

Intuition based

Personalization: New Approach

1-to-1 Interaction

1-to-1 Interaction : Incentives

Image source: Sigma Beauty

1-to-1 Interaction: Birthdays &

anniversaries

Image source: Sunglass Hut

1-to-1 Marketing : Real-time availability

Image source: Carnival

1-to-1 Marketing :

Abandonment emails

Image source: Emarsys via Replacements, Ltd.

Personalization: Universal Profile

Universal Profile

Loves Cold Weather

Can not Walk!

Removable Nose

Avoids Hot places!

Personalization & AI in Marketing

Reach

• AI Generated Content

• Voice Search

Act

• Recommendation

• Predictive analytics

Convert

• Dynamic Pricing

• Personalized Offers

Engage

• Predictive Customer Service

• Chat Bots

Artificial Intelligence

Source: https://rapidminer.com/blog/artificial-intelligence-machine-learning-deep-learning/

Challenges of Real time Personalization

Reacting to User Interactions in Real time

Handling Cold Start (Insufficient data for new

users/behavior)

Avoid showing most popular conclusions

Scale

Real time Personalization: Use cases

Recommendations

Search Ranking

Related Items

Notifications

Personalization through Social Listening

Go through messages on Social Media and

apply AI technologies to perform sentiment

analysis

Then personalize the messaging based on current social trend and user’s personal

attributes

This Photo by Unknown Author is licensed under CC BY-NC-ND

Risk of Personalization

Personalized messages/targeting may appear creepy

E.g. Borrowed my friend’s phone and I am thinking of buying the same phone.

This Photo by Unknown Author is licensed under CC BY

Realtime AI Architecture

Amazon Athena

Amazon Managed Streaming for Kafka

Amazon S3

Amazon EC2

AWS Lambda

Data Scientists

End Customers Amazon DynamoDB

Applications Kafka

AI Model

Amazon Personalize for Personalization

End Customers

Browser

Web Server

Amazon Personalize

Real time events

Recommendation Request

Inference Query

AI Model for Personalization

Classification

This Photo by Unknown Author is licensed under CC BY-NC-ND

AI Model for Personalization Recommendation Engine

This Photo by Unknown Author is licensed under CC BY-NC-ND

AI Model for Personalization

Intuitive Insights Deduce based on customer behavior not based on Surveys

AI Feedback Loop

USE LEARNINGS TO IMPROVE THE AI MODEL

REAL TIME FEEDBACK ANALYTICS FOR OFFLINE MODEL IMPROVEMENT

AI Personalization Pitfalls

Performance Issues

Data Availability

Data Quality

Feedback loop

Questions