2. 2
Agenda
Retail and consumer
organizations have started
to develop more
personalized interaction
with customers, based on
rapid analysis of a broad
range of customer
attributes and propensities,
known metaphorically as
“genes”.
These may be used to
target campaigns more
accurately, or to generate
the next best action in real-
time for a specific
customer.
Business
Opportunities Technology
Challenges
4. 4
Omnichannel Evolution for Retail and Consumer
Systems of Record
• Omnichannel eCommerce
• “Click and Collect”
• Some Personalization
Systems of Engagement
• Omnichannel Marketing
• “Click and Connect”
• Full Personalization
Glue Reply has helped a number
of leading retailers to implement
Omnichannel eCommerce
Our retail and consumer clients
are now looking seriously at
Omnichannel Marketing
Other possible applications of Omnichannel
Engagement include education (pastoral care
for students) and citizen-led journalism.
5. 5
From Conversion to Persuasion
Conversion is not just about this week's revenue. We need
to develop the ability to detect slow-acting and cumulative
effects as well as instant one-off effects. Obviously this is
more difficult, but it is not impossible.
The future for Internet marketing lies in developing non-
linear systems that deliver exactly what prospects need,
when they need it, so they can accomplish their goals in
the manner most comfortable to them.
Conversion is a linear process.
Persuasion is a non-linear process.
Source: Digital Intelligence Today
6. 6
Personalization involves four capabilities
Personalization
Targeting
• Starting with what we
want to promote.
• Selecting consumers for
a given campaign
Customization
• Starts with what the
consumer asks for.
• Take consumer
demands at face value
Contextualization
• Engaging with the
consumer’s world.
• Infers consumer desires
from context.
Co-Creation
• Providing a platform for
active consumer
engagement.
7. 7
Targeting and Personalization
• Produces a list of consumers for a given
message
Targeting Algorithm
• Produces a list of messages and other
actions for a given consumer
Personalization Algorithm
• Both personalization and targeting
require some kind of matching
algorithm.
• The desired “match” is the same in both
contexts. So the two algorithms should
probably have a common core.
Similar or Different?
Targeting
• From Content to
Individual
• Here’s a campaign
message – who are
the best people to
receive it?
Personalization
• From Individual to
Content
• Here’s a consumer –
what message do
we want to give
them?
10. 10
Engagement Framework
Learning
& Development
Data Science
Knowledge & Memory
Consumer
Genome
Information Gathering
Consumer
Monitoring
Decision & Policy
Next Best Action
Consumer Behaviour
Communication & Collaboration
Omnichannel Marketing
Sense-Making
Consumer Analytics
11. 11
Why Real-Time?
Next Action
Real-Time
Context
Product
Genome
Consumer
Genome
• “RTD and online
recommendation engines
are great for helping you
understand that
customers who bought X
also bought Y, but that
doesn't capture the intent
that the customer is
expressing in that current
session. It won't tell you
that somebody is
shopping for a gift, not
buying what they normally
buy. And it won't tell you
that the customer just
purchased a TV, so stop
showing them other TVs
and start showing them
HDMI cables and speaker
systems.” Christophe
Bisciglia, CEO Wibidata
12. 12
Demographics
and Life Events
Socioeconomic category
Life events – work, marriage, children
Product
Experience and
Affinity
Which products do they already have?
Which products are they likely to be interested in?
Responsiveness Price-Sensitivity
Feature Sensitivity
Response to Merchandising and Marketing
Response to Direct Offers
Preferences Communication Style
Channel
Privacy and Consent
Consumer Genome
13. 13
Product Genome
Product
Features
Strong features of product – browsing data suggests this is what attracted
customers to buy
Neutral features of product – little link to purchasing decision
Weak features of product – this is what persuaded customers to buy
something else
Pricing and
Promotion
History of deals for this product
Current / planned deals for this product
Product
Connections
People who viewed X bought Y instead
People who bought X also bought Y at the same time
People who bought X also bought Y at a later time.
14. 14
Real-Time ContextCustomer
Focus
Where is the customer right now?
What can we infer about the customer’s purpose? For example, buying
Christmas presents, or something to wear for the Christmas party?
Which product is the customer looking at now? Which products has the
customer viewed recently?
Customer
Network
Alone, shopping with friends, social media?
Product
Inventory
What is the stock / supply situation?
Product
Transactions
What has the customer previously bought? (Including earlier today)
What is the customer buying now? (Already in basket)
What was the customer buying earlier today? (Left in basket)
15. 15
Customer Journey
Zero
Moment of
Truth
Google Search
(early consideration)
First
Moment of
Truth
Shop / Website
(pre-purchase evaluation)
Second
Moment of
Truth
Getting the product home
(post-purchase evaluation)
Third
Moment of
Truth
Social media
(sharing with network)
Traditional
Extended
16. 16
Inferences from Incomplete Data
Visible Data
data we collect from our own
systems and processes
Processed Data
i.e. transformed by processes
under our control
Dark Data
e.g. interactions with
competitors
Transformed Data
i.e. transformed by processes
outside our control
(e.g. Social Media)
Conventional BI
converts
operational data
into useful analytics
Welcome to the
world of “Big Data”
17. 17
Knowledge to Inference to Decision
Recent activity Product Holding Profile
What we know
Product affinity
Recent activity Product Holding Profile
Potential
inferences
What we know
Product
A
Profile
Product affinity
What we infer
What we know
Left in
basket
Product
A
Churn
Propensity
C
Credit
ScoreA
Demographic
segment D
Profile
Product affinity
What we decide
What we infer
What we know
Left in
basket
Product
A
Churn
Propensity
C
Credit
ScoreA
Demographic
segment D
18. 22
Capability Reference Model
“Know the Customer
Base” is a plural
capability, understanding
the mass of consumers to
detect common patterns
and trends.
“Know the Customer” is a
singular capability,
applying (common)
patterns and trends to an
individual consumer.
19. 23
A simple model
Business Intelligence.
Transformations.
Descriptive models.
Predictive models.
Marketing
Propositions
Interaction strategies
Decisioning
Choose
Personalised
interaction
Customer
Available propositions
Matching logic
Strategy
Channel Context
Customer
Profile
Historic Data
Behaviour
BI Repository
20. 24
A slightly less simple model
Interaction
Channel
Decision Support
Candidate
Activities
Best Activity
Monitor
Outcome
Derive Characteristics
Transform,
Aggregate,
Descriptive
Models
Predictive Models
Current Customer
Characteristics
Historical Data
Propositions and
Strategies
Strategy
Management
Master Data
Management
OperationalSystems
Propositions and
Decisioning
Rules
Propositions
Customer LevelData
Raw DataOperational
History
Zero Latency
Characteristics
Chosen Activity
Subsequent
Behaviour
Master Data
Marketing
Analytics
Customer
Context
Trigger
Response
21. 25
Principles of Consumer Engagement
Holistic Understanding how multiple factors interact to produce particular
behaviours and preferences at a given point in time.
Consumer
Context
Understand consumer pathways – including changes and repeating
patterns over time.
Understand the consumer’s network – friends and influences.
Consumer
Perspective
Don’t just see things from the company’s perspective. Understand
what these events mean to the consumers themselves.
Closed
Loop
Feedback
The outcome of each action helps to calibrate the next action.
Rapid feedback supports broader experimentation and promotes
effective learning.
Ethical Respecting consumer preferences and values.
22. 26
Consumer Characteristics
• Such things as name, age, length of tenure etc., subject to simple transformations.
Simple attributes
• What products or what types of product does the consumer hold or have they held.
• Subject to transformations informed by master data management.
Product holdings
• Simple mathematical derivations such as “total average monthly spend over last 6 months”, “spend
this month to date”, “average number of calls to call centre per month”.
Aggregated values
• A descriptive model classifies consumers, but without reference to predicting any specific future
behaviour.
• Examples would be segmentations, which classify consumers into various groupings based upon
their demographics and behaviour.
Descriptive model outputs
• A predictive model uses consumers’ past behaviour and demographics to predict future behaviour.
• An example would be a churn propensity model, which predicts the likelihood of a consumer to leave
the organisation for a competitor. These may be derived by data mining techniques to determine
predictive attributes.
• A particular subset of predictive model is the scorecard, which assigns a score to various attributes,
giving a total score that is used as a predictor. This derivation is particularly open and may be used
in cases where transparency is required for regulatory reasons, for example credit scoring.
• Predictive models tend to be informed by proposition information, whereas the preceding types tend
to be simply descriptive.
Predictive Model outputs
23. 27
Next Best Activity
The next best activity for a given consumer
is selected based on consumer data …
• Current consumer characteristics.
– May include real-time data from
operational systems, and pre-
calculated data based on history.
• Interaction channel context.
– Provides the consumer identity, the
channel identity and any other
information available about the
triggering interaction.
• Propositions and strategies.
– Provide the logic by which a decision
is made, and define the possible
next actions.
… as well as relevant decision rules and
strategies relating to …
• The aims of the organisation
• The needs of the consumer
• The channel by which the consumer is
interacting
• The eligibility of the consumer for the
various available propositions
• The suitability of the consumer for the
various available propositions
• The costs to the organisation of the
available propositions, and the potential
margin to be made
• Preferences expressed by the consumer
(including privacy and consent)
24. 28
Plugging Personalization into the TouchPoint Process (Email)
Plan Email
Campaign
Create
Consumer
List
Compose
Email
Deliver
Email
Consumer Data
Personalization
Control
Customer
Selection
Control Email
Content
Control
Delivery
Timing
Inhibit
Unwanted
Emails
In this model, we take an
existing marketing process
(eCRM) and plug in some
intelligent personalization
based on the consumer
characteristics.
The model shows four different
points in the eCRM process
where intelligence could be
plugged in. These do not
necessarily have to be
implemented at the same time.
25. 29
Plugging Personalization into the TouchPoint Process (Online Interaction)
Identify
Consumer
Customize
Display
Customize
Navigation
Customize
Offer
Consumer Data
Personalization
Select
Banners and
Images
Control
Search
Sequence
Select
Relevant
Offers
In this model, we take an
existing online interaction and
plug in some intelligent
personalization based on the
consumer characteristics.
The model shows four different points in
the online interaction where intelligence
could be plugged in. These do not
necessarily have to be implemented at
the same time.
Build “Just
For You”
Panels
26. 30
Key Questions - Summary
Why?
• Cross sell? Upsell?
• Retention?
• Acquisition? Cost
• Savings?
• Drive margin?
Who?
• Business units:
• Marketing?
• Analytics?
• Product planning?
When?
• Product lifecycle?
• Latency constraints?
• Strategy reaction time?
Where?
• Which channels?
• Centralised or distributed
decision making?
What?
• Level of decision making
(person, account, device,
organisation)?
• What products?
• What facts?
How?
• System landscape
• Means of customer
• identification?
• Means of strategy
• control?
29. 35
What is the Value of Personalization?
• Message across all channels are more relevant to consumers increasing
their affinity with the channels and brand
• Consumer-led – consumers should feel that we are directly responding to
their actions and preferences.
Engagement
• Improved conversion rate on campaigns.
• Reduced churn.
• Reduced price sensitivity – offers can be based on consumer desire
rather than discounts
• Lifetime value of consumer. Align consumer incentive to consumer value.
Economics
• More effective use of digital campaigns as more targetted,
more coordinated , more timely.
• Growing accuracy of consumer profile, thanks to continuous feedback.
• Support for innovation (e.g. trial offers or campaigns), because faster and
more comprehensive feedback takes away some of the risk
Efficiency