segmentation for maximum output
TRANSCRIPT
‘Frozen segmentation’ does not work
‘Multi-dimensional, dynamic segmentation’ is much more effective – it may be challenging to create, but it will be less painful in the long run
Segmentation needs to be actionable. In the old days, you segmented customers on the basis of recency, frequency, and monetary value (RFM). Today, you segment customers based on what is meaningful to your business (value!).
What we’ve learned
You need to know something about people before you start trying to personalize
How many segments is too few, just right, or too many… is a matter of judgement
Law of diminishing returns: You’ll get 80% of the benefit from 20 segments, 90% from 200 segments, and 100% from 2,000 segments
What ve’ve learned
Managing segments can be overwhelming
Segment when you have ability to take action. Otherwise it’s just an analytical exercise
What I’ve learned
One size does not fit all
Customers expect you tailor your marketing approach to perfectly match their lifestyle—in addition to their product wants and needs
Boost relevance (make the right offer / show the right content at the right time)
Reach higher conversion rates
Avoid churn
Improve customer experience
Why?
Before you do anything else, ask yourself what you’re hoping to achieve from segmentation:
Identifying customer needs to make propositions more suitable for them?
Improving customer profitability by driving up average pricing?
Identifying new target customers?
Improving customer retention?
Identifying opportunities to grow or gain market share?
Be clear on business objectives and strategies
Segmentation is part of a Value Creation Chain
Segmentation sits at the beginning of the value creation chain.
It is used for probing and assessing a situation.
It is used to generate ideas on what to do.
The other analytical powerhouse, Predictive Models, sit at the other end.
It serves a different purpose from Segmentation
Segmentation vs predictive models
SegmentationIdea
GenerationMarket Launch
Measure Outcome
Predictive Model
Toronto Data Mining Forum Oct, 2011
Create persona’s
Set up taxonomy for your website & emails
Learn about a visitors preferences and interests without asking
Link behaviour / interactions / engagement toscores
Act, learn / fail, analyse, adjust (repeat)
Recipe for success
Persona
definition
Online
Value
Propositions
Outside-in
view
Inside-out
view
Functional
requirements
Content
requirements
Customer
journeysOff-site
(acquisition)
On-site
(conversion)
The approach
Persona
definitionOutside-in
view
Inside-out
view
Persona definition
How does our company perceive
its Persona’s
How do the Persona’s perceive our company and
products
Who are we targetting?
Their goalsTheir needs
Their value (CLV)The opportunities
How is our company perceiving its target groups?
Who do we want to target?
Who is benefiting from our products / services?
(Include both internal and external stakeholders)
How are they related?
Who are they influenced by?
How important are they?
What do we want them to do?
How can they bring benefit to our business?
What value can we bring to them?
Persona definition: Inside-out view
Inside-out
view
How?
Workshop with
key persons in the
company.
Creating persona sheets
Persona sheets
Picture
Name
Identity – age occupation
Abilities/skills
Context
Recurrent issues
Goals
Needs
Fears / OpportunitiesInfluencers
How do target groups perceive us?
Who is looking for us/our products?
What questions are they asking?
Where are they looking for us?
What other companies / products are they also looking at?
What are their needs?
What are their tasks?
What are they telling about us?
Persona definition: outside-in view
Outside-in
view
How?
Workshop with key
persons.
External survey
Social listening (optional)
Wikipedia: “the practice and science of classification.” A taxonomy can help you organize an unstructured collection of information
Taxonomy is used to identify the user’s preferences through the use of specific criteria called ‘tags’. These tags can be linked to sensors in order to trap the user’s click and qualify it. A taxonomy list contains the definition of these different tags.
What is taxonomy?
Implicit behavior is what visitors are doing on your site, such as what type of pages/categories are they looking at, or what site path they follow.
Explicit behavior is demonstrated when visitors take action or submit data, like voting in a poll, searching for a specific keyword, or filling out a form.
Types of behavior
Content Strategy: Taxonomy and Scores
TaxonomyScoring
Male factor(M minus F)
Motorization Innovation Design Power Eco Technology Safety Exterior Design Interior Design
behavior (clicks)
1 click(s) 0 click(s) 1 click(s) 0 click(s) 0 click(s) 0 click(s)
html(tags)
2 1 1 0 1 0 1 0 0 0
database(%)
100% 50% 50% 50% 50%
VALUE MAIN DIMENSION INTERESTS
SelligentTable
Top 1 VALUETop 1DIMENSION
Top 2DIMENSION
Top 3DIMENSION
Top 1INTEREST
Top 2INTEREST
Top 3INTEREST
prospect #1
MALE MOTOR INNOVATION DESIGN POWERTECHNO-LOGY
ECOPROFILE INFOper Prospect
Emakina Belgium - Audi
Personalization is a three step process:
1. You identify the visitor.
2. The visitor is classified according to which target group or persona he or she most closely matches.
3. All or part of a web page on your site dynamically changes to show relevant content, bringing the visitor closer to the action you would like him or her to take.
Profiling visitors
Identify visitor intent via behavior and target with relevant content and actions
Main concepts
Profiles & Profile Keys-personas, buying phases, etc.
Profile Cards –assign to content
Pattern Cards –create and store behavior patterns for expected browsing behavior
What is predictive personalization?
Automated Personalization automatically identifies meaningful segments based on hundreds of out of-the-box variables, including:
Site behavior (e.g., new/returning visitor, search behavior, browsing behavior)
Temporal (e.g., day of week, time of day)
Referral (e.g., direct/bookmark, link from e-mail, search/display ad)
Environmental (e.g., operating system, device, mobile/pc, browser).
Automated predictive personalisation
Augment profiles with
CRM-data
3rd-party segmentation data
offline sales
social crm data
Automated predictive personalisation
The cornerstone of Automated Personalization is visitor data. That is what allows us to differentiate between visitors and predict their propensities for conversion.
The performance of any machine learning system is only as good as the quality of the available data.
Systems will:
Predict testing (A/B – multivariate) outcome
Propose tests on your page
Evaluate on your skills
Guide marketers with understanding and defining personas and profiles
Automated personalisation