From Customer Insightsto Customer Action: How to Capture Intent Through AI, ML & NLP
1. 2019 Zeta Global – Proprietary & Confidential www.zetaglobal.com |
2019 Zeta Global – Proprietary & Confidential www.zetaglobal.com
From Customer Insights
to Customer Action
How to Capture IntentThrough
AI, ML & NLP
2. 2019 Zeta Global – Proprietary & Confidential www.zetaglobal.com |
• Meet Ron Sadi
• Case Study Overview
• Identity, Signals & Connectivity
• What the Data Tells Us
• Q & A
Agenda
3. 2019 Zeta Global · Proprietary & Confidential
Meet Zeta Global
3
+ Vice President, Data Cloud
+ 10-year veteran of digital marketing and data
strategy
+ Leads business enablement and data strategy,
working with brands and retailers to maximize the
value of their 1st-3rd party data
+ Prior to that, he led Data Partnerships at Disqus
before the company was acquired by Zeta in 2017
+ Bay Area native
Ron Sadi
4. What Does Data + AI Strive to Do?
Better Audiences
Better Inputs
Better Insights
= Better Outcomes
7. 2019 Zeta Global · Proprietary & Confidential
Fundamentally DifferentApproachTo People-Based
Targeting & Effectiveness
Identity: “DataAt Rest”
200M & 360M
Permissioned Individuals
&
Emails (US)
4B
Digital
Profiles (US)
2,500
Data Elements per
Individual
1T
Content Consumption
Signals / Month
10B
Transactions /Annually
250B
LocationVisits / Month
Data Sources:
+ Proprietary Social
Commenting Platform
+ RTB Bid Request Data
+ Strategic Partnership with
PlaceIQ
+ 500+ real-time intender
segments generated by ML
& NLP
Data Sources:
+ Proprietary O&O Digital
Properties and Sources
+ Partnerships with leading:
Premium Digital Publishers
Offline Data Compilers
Credit Bureaus
Financial Institutions
Intent: “Data In Motion”
Zeta Data CloudIdentity Intent
Data Sources
Zeta’s people-based data enables scalable, integrated
omni-channel activation and deterministic measurement
8. 2019 Zeta Global · Proprietary & Confidential
Make the ubiquitous mileage rewards personal by applying
Zeta’s proprietary signals surrounding competitive browse,
airline transactions, dominate card in wallet, and interests to
stimulate brand usage amongst existing members and
frequent travelers
SOLUTION
CHALLENGE
METHOD
Loyalty programs no longer drive choice of an airline and
program participants have low mileage accumulation to realize
the program value
8
8
OWNED EMAIL
INGEST DATA
1
CUSTOMER
SIGNALS
SYNTHESIZE DATA
Apply Weighted Algorithm For Individual Scoring
ACTIVATE DATA
In Close Loop Deterministic Approach
2
3
OWNED SITE
RESEARCH
OTA BROWSE
MILES
EARN/BURN
FLIGHTS
TRAVELED
AFFINITY
LIFE STAGE
CREDIT
SCOREPARTNER SPEND
SITE OPTIMIZATION
PROGRAMMATIC
TRIGGERED EMAIL
PARTNER RECO EMAILS
= ZETA DIFFERENTIATOR
KPI
Incremental mileage purchases pre-holiday and member
net promotion
Leading U.S. Airline – Client Introduction
9. 2019 Zeta Global · Proprietary & Confidential
Leading U.S. Airline – Client Use Cases
Conquest of other
airline frequent travelers
Disrupt OTA channel
purchase during
preflight research
Reactivate lapsed
customers through
personalized experiences
aligned to interest
Innovate beyond
managing mileage
accruals to retain
loyalists with new real
time partner
redemption options
10. 2019 Zeta Global · Proprietary & Confidential
Our Approach: We capture a user’s identity, pair it with their behavioral,
transactional and location signals and identify their intent.
Vanessa’s Identity Graph
We know Vanessa is likely
recently engaged and can
activate on this life event
through:
VANESSA | Zeta ID: 4188210
Postal
Demographic
Web
Behavior
Location
Transaction
+ Members of household
increase from 1 to 2
+ Marital Status = Single
+ Occupation = Marketing
Technology
+ Higher spending at Etsy
and arts & crafts brands
+ Viewing articles on
wedding planning
+ Researching honeymoon
destinations, resorts and
amenities + Visiting country clubs, event
halls and wedding venues
+ Increased visits at high-end
department stores
Recently
Engaged
0.91
12. 2019 Zeta Global · Proprietary & Confidential
AI-derived Intent
Drives Audience Creation
Methodology for Determining Intent:
+ Define clusters tied to consumer life-stage (e.g.,
new parent) and client requirements (e.g., profile
targets)
+ Classify keywords and actions that provide signals at
varied confidence levels
+ Score individuals based on the semantic relationships
created by their content consumption
Bring to life via high-value audiences and 1:1 real-
time messaging deployed through Zeta platform
12
13. 2019 Zeta Global · Proprietary & Confidential
Data Cloud Understands Page Level Content
URL Title Domain Audience Segment Confidence
https://thepointsguy.com/reviews/hotel-
banke-autograph-collection/
Paris Right Banke: A Review of the
Hotel Banke, Autograph Collection
ThePointsGuy.com Resorts 0.92
Resorts
0.92
14. 2019 Zeta Global · Proprietary & Confidential
Domain Domain Domain Domain
Zeta Score Zeta Score Zeta Score Zeta Score
https://www.msn.com/en-
us/travel/adventuretravel/14-
dreamy-honeymoon-destinations-
that-will-excite-you-more-than-
the-wedding/ss-BBPqdwt
https://www.oyster.com/articl
es/57735-5-luxury-hotels-in-
las-vegas-and-their-cheaper-
but-similar-alternatives/
https://abcnews.go.com/GMA
/Travel/royal-wedding-
honeymoon-destinations-fit-
royalty/story?id=55207068
Domain
Zeta Score
https://www.nomadicmatt.com
/travel-blogs/best-walking-
tours-paris
https://thepointsguy.com/revi
ews/hotel-banke-autograph-
collection/
Visits in Last 30 days Visits in Last 30 days Visits in Last 30 days
3 21
Visits in Last 30 days
5
Visits in Last 30 days
3
Date of Last Visit
October 14, 2019
Date of Last Visit
September 28, 2019
Date of Last Visit
September 2, 2019
Date of Last Visit
October 20, 2019
Date of Last Visit
October 12, 2019
0.92 0.90 0.88 0.89 0.82
Scoring of Content and Individual
Resort
0.90
VANESSA | Zeta ID: 4188210
15. 2019 Zeta Global · Proprietary & Confidential
Scoring of Content and Individual
Membership in Audiences
Behaviors:
Recency,
Frequency
Traits
Lookalike Audiences
Audience Taxonomy
On-Site NLP & Scoring
Metadata Full Text
Machine Learning & Scoring
VANESSA | Zeta ID: 4188210
16. 2019 Zeta Global · Proprietary & Confidential
Scoring of Content and Individual
Membership in Audiences
Behaviors:
Recency,
Frequency
Traits
Lookalike Audiences
Audience Taxonomy
On-Site NLP & Scoring
Metadata Full Text
Machine Learning & Scoring
VANESSA | Zeta ID: 4188210
17. 2019 Zeta Global · Proprietary & Confidential
Data Cloud
Data
Dictionary
Club Warehouse
Home
Furnishings
Home
Improvement
Store
Grocery
Health Club
Event Tickets
Harness Transactional Data
Across Brands
Combine with Location Data to
Drive Business Outcomes
MOVEMENT
DATA
MAPPING
DATA
Understand foot traffic trends by day and market
Identify market opportunities and
competitive pressures
Reach consumers with targeted messaging across
devices based on where they go in the real world
TRANSACTIONAL DATA
LOCATION DATA
VANESSA | Zeta ID:
4188210
18. 2019 Zeta Global – Proprietary & Confidential www.zetaglobal.com |
Demo Credit
Profile/
Ownership
Spend
IRL
Visitation
InputWeighting
Automated ML-Driven
Re-Balancing Based on
Feedback Loop Data
Optimizes Audiences
18
Intender Audiences Are Multi-Dimensional
Behavioral
19. 2019 Zeta Global · Proprietary & Confidential
+ Demonstrating Data Cloud Value and ROI Upfront
+ Aligns Key Business Objectives with Audiences
+ Expands Identity Resolution with Zeta’ Data Cloud
+ Deeper Data Driven Insights Driving Key Opportunities
Storied Outcomes are
Illustrated through Zeta’s Insight
Reports
Data Cloud
Insight Report
20. 2019 Zeta Global · Proprietary & Confidential
Profile and Identity Analysis
25. 2019 Zeta Global – Proprietary & Confidential www.zetaglobal.com |2019 Zeta Global · Proprietary & Confidential
3 Thoughts We Want
You To Take Away from
Today1. Better data inputs allow us to
capture intent, build audiences and
assign membership
2. Glean insights through AI, ML and
NLP by overlaying a lens
of behavioral, transactional,
demographic and location data on
customers and prospects to predict
behavior
3. Trigger personalized 1:1 messaging
to drive better engagement,
increase brand affinity and customer
lifetime value.
Neej opening
Just placeholder slide to document the overall narrative. We lead/ask questions on some of the slides to make it interactive family feud. Winner gets a white claw delivered (and perhaps a non alcoholic alternative. Fortune cookies?)
1. Setup format, get client and use cases from audience. Focus on one where new movers is relevant
2. Discuss Identity, Signals, Connectivity as baseline
3. Talk about Identity and how it relates to our client/usecase. DEMO: identity explorer
4. Talk about signals, the scale, variety and differentiation for our client/usecase. DEMO: ZDI
5. Talk about connectivity/audiences/merged audiences. DEMO: living-demo
6. Assume we get data from client, get into solutioning with IR report. DEMO: IR
7. PDC pitch/applications. DEMO: screenshot
8. outcomes
9. Sales stages and tools and people to utilize at each. DC team.
10. Q&A
11. Appendix for reference/reading
Ron
Ron
Ron
Ron
Sample of publishers where customers are consuming content on a day to day basis
Large brands
Diverse verticals
SUV Interest
Children's Apparel
Pre-Mover
Define the taxonomy through our linguistic experts and domain knowledge
Classify content on the page level
Score individuals in and out of those audiences and at different confidence levels based on their behaviors
Winnie
Introduce you to Vanessa
Snapshot of Vanessa's online content consumption behavior
We look at the type of content Vanessa is consuming – is it high scoring or low scoring
How recently is she looking at that content and how frequently
Score Vanessa in and out of the Pre Mover audience and with a corresponding confidence score based on her online interest and intent
We use NLP & Scoring methodology to classify pages into audience taxonomies then look at our user’s behaviors, recency and traits to move them in and out of those audiences
Winnie
Online and offline spending across a wide array of brands
Point of interest data – what places are you visiting, when are you visiting them and how frequently
Winnie
Online and offline spending across a wide array of brands
Point of interest data – what places are you visiting, when are you visiting them and how frequently
Winnie
Sophia
Do you have a marketing flywheel?
By no means have we cracked the code to solving some of the foundational challenges facing marketers today. But we feel that our approach can help contribute to making consumer outreach more valuable for all parties and are committed to investing to help us all get back to growth.