Innovation:It All Begins With Data
David RaabRaab Associates Inc
Sailthru LiftOctober 20, 2015
99% of companies believe a single
customer view is important
… but only 24% have a single
customer view today
Source: Experian, Data Quality Benchmark Report, 2015
Why does bad data happen to
good marketers?
Everything happens for a
reason.
Sometimes the reason is you’re stupid and make
bad decisions.
Or maybe they just don’t know
how?
Well, yeah, there’s that.
Let’s Build Great Data Quality!• Requirements– Based on use cases– Data elements, quality standards, financial
value of meeting standards• Remediate– Process: audit existing situation, find root
causes, plan changes– Tools: parsing, standardization,
cleaning, validation, governance, etc.– Be SMART: specific, measurable, assignable,
realistic, time-related• Report & Repeat
Tell me more about data
quality.
Tell me more about data
quality.
Something for ME!
Common DataWhat It Is What You Can Do With It Comments• name/address• web behaviors• purchases• service interactions• demographics• lifestyle• interests• lifetime value
• reach people directly• target messages by segment• personalize messages based on
past purchases, demographics, interests
• analyze your customer base to understand them
• find lookalikes for display ads and mailing lists
• profile your customer base and find segments
• mostly first party• lots of value• few firms have it
fully integrated• foundation for
everything else
Intent DataWhat It Is What You Can Do With It Comments• Web content
consumption• ecommerce
hovers, carts purchases
• ad & email clicks• social posts &
shares• public data (moves,
job posts, press releases, etc.)
• (highly processed)
• target display ads to people in-market for your product
• get lists of people in-market for your product
• get lists of customers who may want new products
• gets lists of customers who may be ready to leave
• select best content based on person’s previous choices
• find market-wide trends in product interest and content topics
• mix of first and third party
• highly processed• sources vary in
quality• coverage often
limited
Location DataWhat It Is What You Can Do With It Comments• current lat/long of
mobile devices (phone, tablet, wearable, auto, etc.), based on IP address, GPS, or cell tower
• social check-in• travel transactions
(hotel, airline, etc.)• near-store
geofences• in-store beacons
• identify the location of specific customers in real time and send them related, personalized messages
• send messages to anonymous people at specific locations
• assign customers to segments based on their location patterns over time
• find trends and patterns in aggregate behaviors related to locations
• mostly first party• coverage limited by
need for opt-in and installed app
• privacy concerns may inhibit consumer participation
Internet of Things Data What It Is What You Can Do With It Comments• operating data• location data• environment data• user data• from wearables,
autos, appliances, infrastructure, etc.
• many devices can also communicate with users
• send personalized messages at right moment based on detailed time-series data about devices and customers
• anticipate and fulfill future needs (repairs, supplies, upgrades, etc.)
• send contextual ads through devices without sharing customer data
• assign customers to segments based on their data over time
• find trends and patterns in aggregate behaviors
• mostly first party• higher coverage
because opt-in often part of basic service
• very high data volumes
• anticipate interactions among IoT devices
• anticipate privacy / regulatory issues in the future
Semantic DataWhat It Is What You Can Do With It Comments• topics• concepts• sentiment• entities &
relationshipsfrom• commercial
content (Web pages, TV, video, books, movies, music, etc.)
• user -created content (social posts, reviews, phone calls, etc.)
• automatically scan and classify public and private comments
• automate simple conversations with sales and service
• identify situations requiring human intervention
• accurately recommend products, books, films, etc. based on deep understanding of topics, themes, styles, etc.
• extract concrete information (names, etc.) from unstructured sources
• uncover unexpected trends and content topics
• mix of first and third party
• requires advanced technology
• often creates structured data to load into conventional systems
• never fully accurate; systems must allow for errors
Visual / Audio DataWhat It Is What You Can Do With It Comments• topics• concepts• sentiment• entitiesfrom• user-generated
sites (YouTube, Instagram, Pinterest, etc.)
• commercial sites (NetFlix, GettyImages, Pandora, etc.)
• Web and ecommerce search
• identify customer interests based on non-text content created or consumed
• scan masses of user generated content to find items that matter
• target personalized messages based on content attributes
• target anonymous advertising based on content attributes
• identify social media influencers and understand their interests
• uncover trends and future subjects
• mostly third party• relies primarily on
content tags• some advanced
automated tagging may be available
• new but fast-growing source
All that data looks so great…
Identity Data Unlocks the Single ViewCommon Intent Location Internet
of ThingsSemantic Visual/
AudioPII
• email x x x x
• account ID x x x x x x
• postal x
• phone/SMS x x x
• credit card x
• loyalty ID x
• social ID x x x x
Non-PII
• cookie x x x x
• device ID x x x x x x
= different for each system= consistent across systems
Let’s Build Great Identities!
• Data Prep– Standardize, Clean, Validate
• Fuzzy Match– Postal name/address
• Deterministic Match– Stitch across channels
• Probabilistic Match– Build device graph
• Reference Based– Use external data
Are we there yet?
Are we there yet?
Are we there yet?
Identity is Just the Start
Other data types
Data storage
Personalization
Optimization
Identity data
Real time access
Identity is Just the Start
Other data types
Data storage
Personalization
Optimization
Identity data
Real time access Yay! Let’s build stuff!
It’s worth the trip!Are we there yet?
Are we there yet?We love data!