Lecture 2 Flashcards
Shoppers once relied on a familiar salesperson to find what they wanted. Today…
Today’s distracted consumers, bombarded with information and options, often struggle to find products
Salespeople would draw on their knowledge or quickly deduce about customer, locate the perfect product and suggest additional items. Today…
- Shorthanded retailer floor staff can’t replicate the personal touch that shoppers once depended on
- Consumers are still largely on their own when they shop online
IDC predicts that the “digital universe” (the data created and copied every year) will reach __ ___ (180 followed by 21 zeros) in 2025
180 zettabytes
Poor __ __ is enemy #1 to the profitable use of
machine learning
data quality
To properly train a predictive model, historical data must meet these exceptionally broad and high quality standards
- Data must be right (i.e. correct, properly labeled, de-duped)
- But you must also have the right data (i.e. unbiased over the entire range of inputs leveraged to develop the predictive model)
In a study involving 75 executives, only __ found that their departments fell within the minimum acceptable range of 97 or more correct data records out of 100
3%
How to Improve Data Quality
- Clarify your objectives and assess whether you have
the right data to support these objectives - Build plenty of time to execute data quality
fundamentals into your overall project plan - Maintain an audit trail as you prepare the training data
- Charge a specific individual (or team) with responsibility for data quality
- Obtain independent, rigorous quality assurance
Customer Data
1
Name, Date of Birth, Gender, Postal Address, Telephone, Email, Social Network
Profiles, Account Info, Job Info, Income
2
Transaction (offline & online), Communication, Online Activity, Social
Network Activity, Customer Service
3
Attitudinal Info, Opinion, Motivational, Interests
4
Family Details, Lifestyle Details, Career Info
How to Collect Identity Data
POS System & Online Transaction Database
Clienteling
Social Network Profile & Other Customer Profile Features
3rd party data sources
How to Collect Quantitative Data
Transaction Database
Web Analytics Tool
3rd Party Pixels
In Store Tracking
Google Analytics & Adobe Analytics are platforms that
collect online data and compile it into useful reports
To start collecting data you need to create an account and add __ __ to your site
javascript code.
Every time a user visits a page, the code will collect interaction data and other information from the browser (i.e. language, type of browser, device, etc…)
Web Analytics tools allow us to track every action the
customer is performing on a website:
Browsing Behavior
Search Data
Purchase History
The Challenge of Web Data
Unique visitors does not necessarily mean unique customers
One customer may be tied to multiple cookies (see slide 29 on how to solve it)
T/F
Online retailers know so much more about their customers than their offline counterparts
True
__ __ have created direct connection to their customers, which in turn allows them to collect massive amounts of data about them
Online retailers
Through AI, online retailers are able to create more-personalized…
…customer experiences, fostering levels of satisfaction, connection, and customer loyalty
Amazon has created a new digital marketing model based on a
1-to-1 relationship with the customer, informed by data collection, optimized with machine learning, and nurtured with other forms of AI
3rd Party Pixel Tracking is a tool used to
track and analyze website traffic, individual user behavior, and ad impressions
Hidden in the background of a web page or email so that they aren’t part of the user’s experience
They are complementary to your web analytics tool
New technology can track live foot traffic in a store and
break down shoppers into a variety of data segments
In Store Tracking–Data is pulled from
IoT sensors, beacons and branded app
In Store Tracking
Additionally, depending on the retailer, data is also taken and
joined to POS systems and online data
On an average, consumers in the US use
4 devices each day
predicted to increase with IoT
companies use multiple tools to store different customer attributes (i.e. CRM, Email, Ecommerce, POS, Social Media)
CRM, Email, Ecommerce, POS, Social Media)
Businesses are left with isolated data sets and are paralyzed when it comes to connecting
data silos
Most marketing teams don’t operate with systems that
talk to each other
Most marketing teams manually analyze disparate data points to uncover insights. This data fragmentation is
costing marketers real dollars as they lose the ability to effectively optimize campaigns and fold learnings into future plans
Customer Data Silos
Customer data collected from different sources but not connected to each other bc the systems don’t “talk” to each other.
To eliminate silos, collect all data points into
a centralized system able to analyze them and surface actionable insights in real-time. Marketers can then start to roll out personalized content, translate strategies across all channels, and efficiently improve customers’ experiences.
What is an Identity Graph
• •
What is an Identity Graph
A database that stores all identifiers that correlate with individual
customers, creating a unified customer view and breaking silos These identifiers could be anything from usernames to email,
phone, cookies and even offline identifiers like loyalty card number
Across a consumer’s journey, multiple identifiers may be associated
with an individual
The ID graph collects these identifiers and
connects them to the customer’s profile and any related data points, including behavioral data like browsing activity or purchase history
There are 2 types of customer profiles:
Authenticated
Non-Authenticated
Authenticated:
uses durable IDs such as an email address or log-in ID
Non-Authenticated:
uses shorter-lived identifiers that don’t translate across devices
Authenticated Profiles
IDs include:
email addresses or customer IDs that
require a log-in
more durable than cookie-based data that
expires and is restricted to the Web
Upon authentication (ex: Site log-ins), the ID graph links this 1st party data to the various bits of data used to uniquely identify this person across channels or devices
With each customer interaction, the persistent profile accumulates data, increasing clarity and value over time
Non-Authenticated Profiles are built from identifiers like
cookies or device IDs, which
are shorter-lived or don’t translate across devices
Work fine within a single channel or for a single campaign
Unable to continually collect and connect customer data
Results in just a partial view of a customer; essentially, a snapshot in time
ID graphs use 2 different data matching methodologies:
Deterministic
Probabilistic
Deterministic
uses known customer information (i.e. log-in data, hashed email addresses) to match and recognize individuals across devices with 100% certainty
Probabilistic
uses anonymized data signals (ex: IP address, device, browser, location, OS) to create likely statistical connections across devices, achieving greater scale but lower accuracy
What can marketers do with an ID graph?
Power people based targeting
Enhance customer engagement
Attribute and optimize performance
What can marketers do with an ID graph?
Power people based targeting
ID graph collects/ connects all personal identifiers to one person, so a customer can be recognized and targeted with the right content and in the right context across screens
What can marketers do with an ID graph?
Enhance customer engagement
Uploading and matching offline data with digital identifiers and behaviors within the ID graph delivers a 360-degree view of a customer, which can be leveraged to anticipate what consumers may need and strategize future interactions
What can marketers do with an ID graph?
Attribute and optimize performance
Having all customer device, channel and behavior data in one place allows advertisers to accurately measure reach and frequency of their campaigns.
