Digital Media Analytics January 30, 2014
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Digital Media Analytics January 30, 2014
Agenda
• Introduction
• Module 1: Brief Overview of Digital Media Ecosystem
• Module 2: Competencies and Technology Needed to Succeed
• Module 3: Discussion of Data Created within Ecosystem
• Lunch
• Module 4: Customer Level Targeting
• Module 5: Attribution and measurement
• When You Get Back to Work….
• Exercises
2 A Glossary has been provided as a separate handout
“Should we increase or decrease the
spend on Video Advertising?”
“We will like to get results of digital attribution every quarter”
“Can we understand which banners
ads should be served to different
customers?”
“How will the results match with
my Marketing Mix model?”
“If the digital media is not meeting
the benchmarks, should we
change the creative?”
“What is the Rx impact of my Digital campaign in first Half of the year?”
Questions We Hear About Digital Analytics
“Do I have all the data I need for digital attribution?“
“Can you assess the performance
of Un-branded and Branded Digital
Campaigns?”
3
“Are my paid media campaigns properly set up for tracking, reporting, and measurement?”
Age of the Customer
Digital Media Ecosystem
`
There Is A Power Shift Happening in
Customer and Health Branding Today
1950s+
AGE OF
BRAND creation of
national brands
via TV
Brands
at
Scale
BRAND in control
1980s+
AGE OF
BIG BOX physically get
closer to
customer
Big Box
Format
RETAILER in control
2010+
AGE OF PATIENT-
CENTRICITY low cost 1:1
personalized
engagement
Digitization
CUSTOMER in control
1995+
AGE OF
INTERNET direct-to-
consumer
business model
eCommerce
CHANNEL in control
Macro-trends are changing the
landscape…
Digitization of everything
Social networks at scale
Consumer mobility
…and Customers are responding
Shift in media consumption patterns for Patients,
Caregivers, HCPs and Payers
Changing Customer Behaviors
Mass consumer to consumer engagement
Know Me then
Amaze Me
5
The Always
Addressable
Customer
And marketers are
responding by
And consumers
are changing
Mass consumer to
consumer engagement
Changing consumer
purchase behaviors
Shift in media
consumption patterns
Focusing on big data and
its ability to drive value
Embracing digital media
and channels to enhance
customer experience
Putting the customer at the
center of business strategy
Social networks
at scale
Macro-trends are changing the landscape
The Age of the (Addressable) Customer
Mobility at
scale
Addressability at
scale
6
Migration of Customer from Offline to Digital
2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018
7.0——
6.5 —
6.0——
5.5 —
5.0——
4.5 —
4.0——
3.5 —
3.0——
2.5 —
2.0——
1.5 —
1.0 ——
0.5 —
2008-2013
Social media 0.1 to 1.1
Digital content 1.7 to 2.4
Consumer hours spent
per day on non-digital channels are decreasing,
while use of digital channels
are steadily increasing
Hours
2008-2013
TV 3.8 to 3.1
Radio 1.6 to 1.4
Print 0.7 to 0.4
7
Advertising Budgets Are Shifting
Mobile Advertising
Increase Stay about the same Decrease Don’t use
69% 20%
70% 29%
64% 24%
23% 50%
19% 48%
13% 20%
Social Media Advertising
Video Advertising
Rich Media Advertising
Standard Display Advertising
Connected TV / IPTV
Survey on the amount marketers will increase/decrease
their budgeting in these forms of advertisement
8
Social and Mobile Trends – Always Connected
9
The Rapidly Growing Digital Advertising Market
Current and projected market share from 2011-2015
Internet
Outdoor
Cinema
Radio
Television
Magazines
Newspapers
7.3%
2011 2015
20.3% 15.9%
9.4%
7.3%
39.9%
40.0%
7.1%
6.6%
6.7%
6.3%
16.1%
23.4%
0.5
% 0.6%
10
Digitization is Creating Massive Amounts of Data
For Digital Marketing
“There were 5 exabytes (5 million terabytes) of information created
between the dawn of civilization through 2003 but that much
information is now created every 2 days, and the pace is increasing.”
- Eric Schmidt, CEO, Google
182 billion e-mail messages
are sent each day
29.8 billion ads served by
Google each day
70 billion pieces of content
shared on
Facebook every
month
400,000 bid requests per
second processed
on the AppNexus
ad platform
Merkle manages over
3 peta-bytes of marketing data,
increasing by approximately
10TB/month”
Storage for one client includes
8,824,526,619 page views (to
be exact) and over 24
terabytes in a single database
11
Programmatic Media Has Now 50%+ Share in US Digital
Media Market with RTB the Fastest Growing Area
76%
62%
47% 36%
27% 21% 17%
13%
18%
25%
29%
32% 32%
31%
11% 19%
28% 34%
41% 47% 52%
100%
80%
60%
40%
20%
0% 2011 2012 2013 2014 2015 2016 2017
US: Programmatic Share (% of Digital media transactions)
RTB Non-RTB Non-Programmatic Source: Magna Global
Overall Digital Media Marketplace - $61B by 2017
$8B in RTB media by 2017 growing at 59% CAGR
$8B in “Custom Audience” by 2017
Over half of all digital media today is bought programmatically
12
Targeting &
Personalization
• Research
• Persona
• Segmentation
• Retargeting
• Individual level activity,
next best product,
customer value
Measurement • Impressions vs. Calls
• Last click
• Digital attribution
• Media Mix Modeling
• Accurately assigning
“partial credit” cross-media
and channels
Channels &
Media
What’s The Opportunity?
Personas Known Anonymous
Individual
Data generated by digitization is driving addressability at scale across audience
platforms at the individual level
13
The Addressability Spectrum
Anonymous Partially Identified Identified Identification:
Unknown Some Knowledge Well Known Knowledge:
Defining Addressability
Addressability is the degree to which customer data (anonymous or identified) can be
used to increase the target-ability and personalization of marketing impressions and
experiences
14
The consumer clicks on
an online ad, which
conveys their city through
the IP address.
Definition: The degree to which you can use customer (anonymous or identified) data
to increase the targetability and relevance of marketing impressions and experiences
Le
ve
l o
f id
en
tifica
tio
n
An
on
ym
ous
Pa
rtia
l ID
F
ull
ID
Low Medium High
Level of knowledge
The Addressability Spectrum
Interest in
product
Location
Definition: The degree to which you can use customer (anonymous or identified) data
to increase the targetability and relevance of marketing impressions and experiences
Le
ve
l o
f id
en
tifica
tio
n
An
on
ym
ous
Pa
rtia
l ID
F
ull
ID
Low Medium High
Level of knowledge
The Addressability Spectrum
The consumer submits
their email address, which
helps point the way to
data about them that is
elsewhere online.
Interest in
product
Location
Public online activity
Definition: The degree to which you can use customer (anonymous or identified) data
to increase the targetability and relevance of marketing impressions and experiences
Le
ve
l o
f id
en
tifica
tio
n
An
on
ym
ous
Pa
rtia
l ID
F
ull
ID
Low Medium High
Level of knowledge
The Addressability Spectrum
Highest value
The consumer logs into
Facebook, providing an
exact name and identity.
Interest in
product
Location
Public online activity
Real
name
Definition: The degree to which you can use customer (anonymous or identified) data
to increase the targetability and relevance of marketing impressions and experiences
Le
ve
l o
f id
en
tifica
tio
n
An
on
ym
ous
Pa
rtia
l ID
F
ull
ID
Low Medium High
Level of knowledge
The Addressability Spectrum
The consumer reads a story
on a news site, which logs
their IP address.
Interest in
topic
Region
Definition: The degree to which you can use customer (anonymous or identified) data
to increase the targetability and relevance of marketing impressions and experiences
Le
ve
l o
f id
en
tifica
tio
n
An
on
ym
ous
Pa
rtia
l ID
F
ull
ID
Low Medium High
Level of knowledge
The Addressability Spectrum
The consumer’s IP is cross-
referenced with publicly available
data, showing the general
location where they live.
Interest in
topic
Region
Specific
location
Le
ve
l o
f id
en
tifica
tio
n
An
on
ym
ous
Pa
rtia
l ID
F
ull
ID
Low Medium High
Level of knowledge
The Addressability Spectrum
High value
Later, the same IP is logged at a
shopping site. Linking the news story,
location, and retailer allows a targeted
ad to be served to the consumer.
Interest in
topic
Region
Specific
location
Interest in
specific
product
15
Determining Addressability and Value of
Qualified Patients and Health Care Providers
Au
tom
atio
n
Addressability Phenomenon Is Reaching New Levels of
Sophistication Due to Rise of Audience Platform
An Audience Platform is a technology that enables automated, real-
time delivery of targeted, personalized experiences to individuals
(known and anonymous) at scale utilizing first and/or third party data
Customer Data
Marketer 1st Party Data
3rd Party Data
Au
tom
ation
Individual Level
Delivery
Audience Platforms
Audience Owners
16
The evolution of digital media in recent years is a good example
Contextual Real-Time Bid (RTB) CRM Targeted
RTB
How many people visit site, and
their profile
Cookie & Third Party data on
unknown individuals
Addition of data repository including
all media and CRM data
Targeting • Based on inferred match
between audience &
publisher
• Anonymous user behaviors
• Retargeting
• Anonymous user behaviors
tied to CRM data
• 1st party data targeting
(name and address)
Optimization • Publisher performance • Anonymous cookie data Performance by customer
• 1st & 3rd Party Cookie
• Device ID
• Known individual
Place
Anonymous and Individual
CRM Data Driven
Place
Anonymous Individual
Place Level of insight
Digital & Social Media Evolution 2008 - 2013
Platforms &
Format
• PC
• Banners
• PC
• Display & Video
• Channel specific
• Cross Media/Cross Device
• PC
• Mobile
• Tablet
One Great Example Of This Is The Rapid
Evolution of Addressability in Digital Media
17
But Addressability Is Not Just About Display Media
– Search is Evolving Very Quickly As Well
Universal Platform Differentiated by
Device
Integrated Media and
Site Targeting
How many people are searching
and for what terminology
Increased options controlled by the
search engines for delivery
Use of Remarketing programs in
search and display to customize
Targeting • Based on exact keyword
search behavior with not
personalization
• Anonymous by device type
and carrier
• Geography, day of week and
time functions
• Audience profile data from
prior site visitation
• Unique experiences based on
user profiles
Optimization • Match type and keyword • Extended match types
• Device Targeting for Mobile
• Location specific ads and
costs
Performance by customer
• Segment
• Value
• Intent
Keyword
Audience Driven
Keyword
Location/Device/Time
Keyword Level of insight
Search Evolution 2010 - 2013
Formats • Pure text only • Site Links
• Video Ads
• Image/Logo ads
• Click to Call
• Video
• Form Extensions
• Product Price Ads
• Maps/Location Extensions
Location/Device/Time
18
Some Search Marketing Ads still
Struggle in Medical Legal Review
• Broad Match (Branded) – Many PharmaCo’s still fail to approve SEM submissions beyond Exact Match
• Unbranded Ads – Cannot use co‐morbid or off‐label indications to target keywords
• Patient Common Terms for Symptom Key Words – Approvable but still must be qualified and on label
• Unbranded URL’s within unbranded ads – Approvable but the url cannot include any product representation
• Branded Ads on Competitive Brand Name – No MLR barrier but requires an alignment with Commercial Team
19
Industry Example
Landing Page Best Practices
20
Clicking Here Lands Here
Landing pages should contain:
• Strong CTAs
• Content Highlighting Special Offers
Landing pages should not:
• Differ within an ad group
Industry Example
Ad Copy Best Practices
21
• Ad text should include copy related to
the user’s Search Query
• Utilize multiple variations of ad copy
– Brand awareness (with full
generic name)
– Highlight Special Offers – Pay no
more than $25
– Non brand description with non
brand destination URL
– Call to action to learn more
• Rotate ad copy throughout
campaign based on keyword sets to
determine highest producing click-
through and conversion rates.
The Big Trends – Where This Is All
Heading
Digital media
convergence –
digital media has
taken on forms in
custom content,
video, social, and
mobile , search
Known individual
level targeting
reaches massive
scale as the core of
the strategy
Mass adoption of
social log-in will reach
huge scale and open
massive addressability
opportunities
Cross device
targeting
maturity will
accelerate –
Google and
Twitter to lead
the way (Apple,
smart TV)
1st party audience expansion and extension
creates massive addressable scale opportunity
Commerce - Data driven inventory will expand dramtically
as large commerce expand business model to monetize first
party data assets – eBay, Amazon
Publisher - Google and Facebook off platform extension -
stack integration will create massive addressable audience
scale (Atlas and DART acquisitions used to drive ubiquity
and integration of advertiser, social, and paid media
targeting and reach
Advertisers will have to deal
with complexity of the closed
media platforms as large
players such as Google and
Facebook create “walled data
gardens” through their stack
acquisitions
Search will be
the next big
addressable
platform
Programmatic
media buying
explodes
Unique content
will drive growth
of video-on-
demand which
opens yet another
big addressable
platform at scale
(Netflix – Orange is
the New Black)
The info-mediary
starts to take shape
- Consumer influence
into their own
experience – the
value exchange
22
Some of these platforms are creating addressability
beyond the domain of their own native platforms
This has not happened yet, but the connections can
be made at scale
As endemic health buys fall out of acceptable ROI targets, updating media plans to reflect more efficient, targeted media through audience buying is essential.
23
Some of these platforms are creating addressability
beyond the domain of their own native platforms
…and like
clockwork, a
week later,
makes this
announcement
24
June
2012
October
2012
May 3
2013
July 11
2013
August 9
2013
August 22
2013
July
1941
July 1941
The first ever television ad,
a ten-second Bulova watch
spot, airs prior to a Brooklyn
Dodgers and Philadelphia
Phillies game.
Sept
1998
1994
March
2012
August 22
2013
July
1941
March
2012
June
2012
October
2012
May 3
2013
July 11
2013
August 9
2013
Sept
1998
1994
The first display ad from
AT&T
1994 July
1941
March
2012
June
2012
October
2012
May 3
2013
July 11
2013
August 9
2013
August 22
2013
1994
September 1998
Google launches service
Sept
1998 July
1941
June
2012
October
2012
May 3
2013
July 11
2013
August 9
2013
August 22
2013
Sept
1998
1994
March 2012
March
2012 July
1941
October
2012
May 3
2013
July 11
2013
August 9
2013
August 22
2013
Sept
1998
1994
March
2012
June 2012
June
2012 July
1941
June
2012
May 3
2013
July 11
2013
August 9
2013
August 22
2013
October 2012
October
2012 Sept
1998
1994
March
2012 July
1941
June
2012
October
2012
July 11
2013
August 9
2013
August 22
2013
May 3, 2013
May 3
2013
Sept
1998
1994
March
2012 July
1941
June
2012
October
2012
May 3
2013
August 9
2013
August 22
2013
July 11, 2013
July 11
2013
Sept
1998
1994
March
2012 July
1941
June
2012
October
2012
May 3
2013
July 11
2013
August 22
2013
August 9, 2013
August 9
2013
Sept
1998
1994
March
2012 July
1941
June
2012
October
2012
May 3
2013
July 11
2013
August 9
2013
August 22, 2013
August 22
2013
Sept
1998
1994
March
2012
Innovation in the Platforms is Picking up
Significant Speed and Volume
25
BIG Digital Trends in Health – Where Is
It All Heading?
The audience platform has become highly ADDRESSABLE and is reaching MASSIVE
SCALE
Marketers seeking growth and competitive advantage will now be “leaning in” hard on these
platforms MOVING HUGE AMOUNTS OF BUDGETS from mass media and traditional direct
marketing
We are already seeing marketers moving that budget at scale and seeing 20-40% LIFT IN
MEDIA PERFORMANCE … and we are just getting started
But the challenge is that YESTERDAY’S MARKETER DOES NOT HAVE THE SKILLS AND
TOOLS to really go beyond the haphazard “bag of tactics” and gimmicks to really leverage
the opportunity here
We need to evolve … introducing THE HEALTH PLATFORM MARKETER
26
• 1%-3% increase in customer acquisition over 5 years • Increased response/conversion from digital media efficiency by connecting Anonymous Data
to CRM data for better targeting, measurement and segmentation • Increased effectiveness of remarketing and personalization (offer/package) in Search, Display,
Site and Email
$3 MM $8.6MM $14.6MM $16.4MM
We Believe This Has Value To Our Brands in
Tens of Millions
27
$7.8MM $20.8MM $34MM $36.4MM
Acquisition of new customers
Improvement in Patient Adherence
Improvement in Intent to Prescribe/ Intent to Ask My Doctor over Baseline*
NPV of Revenue Impact
• Roughly 1bp (~31k) improvement in Adherence YOY over next 5 years • Combine Customer service, and contact data to improve
$4.8MM $12.1MM $19.4MM $20MM
• Use Connected data to predict/model Improvement in Intent to Prescribe for HCPs and Intent to Visit and Ask my doctor
+5 +10 +5 +5
YEAR 1 YEAR 2 YEAR 3 YEARS 4+5
What’s Changing and Why You Should Care?
Introducing The Health Platform Marketer
Introducing…The Health Platform Marketer
• Decision science PhD
• Audience platform expert
• Marketing technologist
• Programmatic media buyer
• Endemic Media Expert
• Addressability expert
• Measurement and attribution expert
• Chief Patient/ HCP/ Payer Economist
Health Platform Marketer wears many hats & embodies the
competencies needed to successfully operate in today’s digital world
• Change advocate and champion
• Consumer experience designer
• User experience expert
• Creative advocate
• Consumer privacy & preference advocate
• Multi-channel program strategist
• Direct Marketer
• Segment portfolio manager
29
Health Platform Marketer Represents a Dramatic Shift From
Traditional Marketing Skills and Competencies
The Traditional Marketer The Health Platform Marketer
Big Data informs Big Ideas in CRM Channels
Media buying driven through the tech stack and
audience platforms
Integrates consumer experience across media and
channels at segment and individual level
Big Idea informs DTC or HCP Campaign
Programs disconnected from the lived customer
experience
Marketing moves in internet speed through
programmatic approach to decisioning and execution
Media buys reliant on buying clout and scale
Marketing moves at the speed of human
decision
30
Twitter handle
Cookie 3rd party
ID address
Digital set top ID
Mailing
address
IB ID
Pinterest ID
GooglePlus ID
Device ID
70’s 80’s 90’s Today
12 Main Street
Philadelphia, PA
Email Phone
number
00’s
617-555-0728
12 Main Street
Philadelphia, PA 617-555-0728
12 Main Street
Philadelphia, PA
617-555-0728
12 Main Street
Philadelphia, PA
#JohnnyDoe
01100100010
0110001 //asdohs.hhd.net 617-555-0728
12 Main Street
Philadelphia, PA
#JohnnyDoe
01100100010
0110001
//asdohs.hhd.net
Pinterest: jdoe
JD’s iphone
011001000
100110001
01100100010
0110001
//asdohsd.asiudhscns/html
The Platform Marketer
knows he/she must
maximize his addressable
market through high
coverage of consumer
identifiers and knowledge in
the database
This requires mastery
of consumer
addressability in the
database and
constant collaboration
and leadership with
technology
It also requires deep consumer
insight and experience skills to design
and implement the experiential “value
exchange” that incents consumers to
provide data (e.g. why should one
identify on a site with one’s facebook
log-in?)
Platform marketers bring addressable data skills that
facilitate the exchange of identity and data for
personalization and relevance.
Health Platform Marketer – The
Addressability Expert
31
The Platform Marketer knows customer segmentation intimately and uses it as a
core strategic tool to better understand opportunities and risks in the market
Health Platform Marketer – The Segment
Portfolio Manager
YoungTrendse er
EnlightenedConsumer
TimeSavvyMom
Thri er
$1.5 $10M
$5 $30M$15 $10
$4.5 $12M$3
ShopperMarke ng
Tradi onalMedia
Promo onalSpend
$8.5 $3.5$6Social/DigitalMedia -- $18M
TotalMarke ngDollars
$4 $70M
$4
$3.5$2 $3
--
<$1
$14.5$22.5$29
StoreFootTraffic
PricesforUpSell
CustomerLoyalty
MassMarke ng
ROI $8M$14M$38M$48M $100M
Customer value analysis highlights the right investments that
should be made to each segment and the return that can be
expected
Segmentation needs to be fully
operationalized, reported on
and tracked over time
32
Segmentation Gives us the Ability to Feed
Individualized Moments to Change Health Behavior
PERSONAL
EXPERIENCE
CONNECTED CUSTOMER
PROFILE
SEGMENTATION
message
treatment
IP
context
offer
33
Convert
Purchase
Activate
Triggered
Direct Mail
CONNECTED CUSTOMER TM PROFILE
DRIVES MESSAGING
Social
Incentive
Review
Incentive
Recommend
Pairing
Re-Purchase
Re-Activate
Over Time, Pharma Customers Receive Smart Messaging
Based on Customer Preference & Segment Behavior
34
Concept In-Action
Web Mobile SMS
Printable
Integrated Experience Delivery
• Personalized Email and SMS CRM Program
• Printable & Mobile Coupons
Step One
Industry Example 35
Concept In-Action
36
Step Two
Mobile Coupon
Printed Coupon
Redeemed at
pharmacy
Customer
Segment
identified
Tailored email
Tailored SMS
Industry Example 36
Industry Example
Case Example: BrandX Adherence Program
• BrandX mobile adherence program provides personalized education, emotional support and smoking cessation tips.
• Message frequency, mix and content is continuously updated based on input received from patients in real-time.
37
User Texts “URGE” To Receive Tips
RedShop Rx Sends Coping Tips Via SMS
CONTENT &
CONTEXT
INTENT &
BEHAVIOR
ANONYMOUS
INDIVIDUAL
IDENTIFIED
INDIVIDUAL
context
custom content
intent
geo-location
device ID
anonymous cookie
1st party cookie
name & email
behavior
3rd party segments
probabilistic ID
The Platform Marketer is a master of the ever evolving Audience Platform targeting and
optimization capabilities
The Health Platform Marketer – The Audience Platform Expert
38
A Short History Of Digital Media Buying
Audiences
aggregates by
scaling niche
content
Audience
aggregated by
third party data
Audiences
aggregated by
content
Differentiation created
by Media Skills Differentiation Created
by Optimization
Differentiation Created
by Technology
Audience
aggregated
using known
relationships
Differentiation Created
by Data Integration and
Analytics
1995-2005 2005-2009 2009-2012 2013+
39
Digital Media 10 Years Ago
Buying is relationship based with targeting and optimization done at a very coarse level.
“Transparency line” ends at the network and publisher level – what falls below the line is “black box” .
Buying is done across numerous platforms without the ability to manage frequency and cost resulting in significant waste .
Just as bad (or worse), targeting capability does not allow for targeting the right individuals.
“Black box” ad networks “Black box” ad networks “Black box” ad networks Direct sales force
“Remnant” inventory
“Remnant” inventory
“Remnant” inventory
“Premium” inventory
Publisher Publisher Publisher Publisher Publisher Publisher Publisher Publisher
Agency
Approved Campaigns
40
Digital Media Today – Challenges
Remain For Life Sciences Adoption
Buying is done using a data-driven targeting skill-set and mind-set.
Consolidated buying platforms allow for complete transparency and granular targeting – no more black box.
Real-time-bid environment allows for access to premium and remnant inventory that gets bid on auction-style based on the value to the advertiser.
Direct buys and paid social leverage data and technology to cross multiple channels while remaining customer-centric.
Publisher Publisher Publisher Publisher Publisher Publisher Publisher Publisher
Integrated Media Management Platform
Data & enabling technology
Real-time bidding auction
Paid Social Direct Buys Programmatic (DSP)
41
Targeting Framework C
on
solid
ated
Bu
yin
g P
latf
orm
(D
SP)
Tra
din
g D
esk
Lookalike Modeling
Match converted consumers to anonymous ID
and create look-alike predictive
model to identify “like” cookies/
placement opportunities through RTB
Online Audience Segments
Identify high performing
online audience segments (“auto intenders”) and
target these anonymous users through the DSP
Re-Targeting
Identify users visiting site
(anonymous or authenticated)
and target customized
impressions after they leave the
site
Online-Offline Direct Match
Match offline “top deciles” to cookies through
third party match providers and target known
consumers on a 1-1 level
42
Health Platform Marketer – Programmatic
Media Buyer
The Platform Marketer brings programmatic buying skills to the enterprise
43
Platform Marketer – The
Stack Expert
Platform marketers has strong expertise in state of the art and emerging
marketing technology and how it drives business value
44
In the last 18 months, we see advanced marketers
rationalize this technology into a unified stack
Audience
Platforms
Platform
Marketer
Stack
Name & address 3rd party cookie 3rd party segment Context
1st party cookie Device ID Geo-location Social ID / handle
Execution
Currencies
Campaign
Management
DMP
Attribution &
Insights
On-boarding
Ad Serving &
Tag Management
Identity
Management
Marketing
Database
45
Market Forces Require A Different Type of Analyst
• “Marketers must be able to keep pace with their customers and react to changes in customer behavior instantly” Forrester
• Batch analytics is no longer sufficient
• More ads are targetable at a user level than ever before through, display, social, video, and mobile.
• 1/3 of US online adults are always addressable through digital media Forrester
• Yet, advanced user-level attribution is not widely adopted, most emails are batch, and organizations are not unlocking the value of user-level ad and site targeting.
Analytic methods and tools need a big data
reboot
Analytics is not matching up to real-
time marketing
Data scientists are critical to drive digital and offline data
integration
Analytics is a constraint as media becomes more
targetable
• “The data science toolkit is more varied and more technically sophisticated than the BI toolkit” Green plum
• “There is a shortage of talent necessary for organizations to take advantage of big data.” McKinsey
• Analysts must help marketers and technologists figure out what data is valuable and how it should be integrated
• Managing and integrating data from a variety of sources is the top challenge preventing organizations from making use of customer analytics. Forrester Customer Analytics as a
Marketing Competitive Advantage
46
SUMMARY
The Health Platform Marketer
Addressability at
scale has and will
create competitive
advantage
The new
addressable
platforms will
require new
analytical
competencies!
Massive budgets
are already being
shifted to take
advantage of this
opportunity
47
What data is created and how it can be connected to create value?
Digital Data and Data Integration
Optimized Channel
Experience
(Targeting and
Personalization)
Value is Unlocked Within The Digital
Marketing Value Chain
49
First
Party
Data
Second
Party Data
Third
Party
Data
Inte
gra
tio
n
Measurement and
Budget Allocation
(Attribution)
Connected Customer
Insight
(Data Integration)
The Digital Marketing Value-Chain
Anonymous Behavior Tracking
Anonymous User Identifiers • Cookies • IP address • Device fingerprints (Probabilistic Ids) • Mobile Device ID • Social handle
User
Data Collection Methods • JavaScript • Pixels/Beacons • Packet Sniffing • Web Server
• Cookie ID: 43Jx41LKs980s
• IP address: 192.168.2.49
• Device fingerprint:
34x43292jk2395kls9ef876
50
How A Website Works
51
HTTP
• HTTP (HyperText Transfer Protocol) – Protocol for requesting and responding to requests for web pages (hypertext)
• Request/Response – Methods (GET, POST, PUT, DELETE,...) – Response codes
• Stateless protocol
• Request line, Response status line, Header, Body info – Host, User Agent, Referrer, Cookies
HTTP Request (from client) GET / HTTP/1.1 Host: www.linkedin.com User-Agent: Mozilla/5.0 (Windows NT 5.1; rv:21.0)
Gecko/20100101 Firefox/21.0 Accept: text/html,application/xhtml+xml,application/xml; Accept-Language: en-US,en;q=0.5 Accept-Encoding: gzip, deflate Cookie: leo_auth_token=... Connection: keep-alive [optional request body, e.g. when posting data from a form]
HTTP Response (from server) HTTP/1.1 200 OK Server: Apache-Coyote/1.1 Content-Encoding: gzip Vary: Accept-Encoding Content-Type: text/html;charset=UTF-8 Content-Language: en-US Date: Fri, 07 Jun 2013 01:49:26 GMT Connection: keep-alive Set-Cookie: _lipt=deleteMe... [response body; e.g. html content goes here]
52
How Web Data Capture Generally Works
Typical site visitor
1
• IP address:
192.168.2.49
Looking for ways to donate food in her community. Does search on local food banks and clicks on paid search ad for
Feeding America
2 Lands on food bank search page
1 GA sees that browser coming Google.com paid seach has no cookie, drops 1st party cookie on browser, and counts browser as a new site visitor
Google Analytics is web analytics tool for Feeding America (Javascript on all pages)
2 GA records all actions taken by user on site in Google collection server. Java script instructs what data to send.
3
• Cookie ID: 43Jx41LKs980s
• IP address: 192.168.2.49
When she leaves the site the session is marked as complete and session metrics such as time on site, etc. are calculated
3 Next time she comes to the site GA recognizes the browser based on cookie ID
53
A Data Flow View of Data Capture
Web Browser
Feeding America Web Server
Firefox
Google Analytics Collection Server
1
Web server notifies Google analytics collection server of request
2
3 Collection server looks to see if user has a cookie and drops cookie if no cookie exists
Collection server captures behavior on site per pre-configured collection rules
4 Web browser requests content from Feeding America
Note: Pages load for user regardless if collection server can complete their actions. If user leaves page before collection script completely loads then no data capture will happen.
54
What Data Is Passed To The Collection Server?
• Cookie ID (Assuming browser accepts cookies) • IP + user agent data
• Contextual information (Where you are) • http://espn.go.com/mens-college-basketball/ • Note: this data is sometimes masked on third party sites
• Referrer (where did you come from) • Behavioral (What you did)
• Basic—clicked on ad (Beacons) • Extensive – watched 1/3 of video (Javascript)
Source: http://www.whatsmyuseragent.com
55
Cookie Background
What is a cookie?
• Small snippets of plain text containing a key, value pair, and saved within the browser, that are used to maintain state throughout your visit to a website (HTTP is a stateless protocol)
• Cookies can only be read and written by the domain to which they belong (i.e. cross-domain cookie access is not allowed by your web browser)
There are two flavors of cookies important to this discussion
• First-party cookies – Belong to the same domain as the requested web page (Example: NIKE assigning a cookie to browser of NIKE.com)
• Third-party cookies – Belong to domains other than the domain of the requested web page. These are read and written by separate third-party HTTP requests on the web page, commonly for advertising and tracking purposes, but also for providing 3rd party content. (Example: Google assigning a cookie to a browser on NIKE.com )
56
IP Addresses
• IP Address (Internet Protocol Address) – A unique address for finding any machine connected to the Internet. This is how client requests and server responses are sent by routers to the correct location over the Internet.
• IPv4 address – 32 bits => 232 = 4,294,967,296 unique addresses
• IPv6 address – 128 bits => 2128 = 340,282,366,920,938,463,463,374,607,431,768,211,456 unique addresses
– Went live 6/6/2012, there will be several years of transition – Every machine will be able to have a unique public IP address in the future – http://www.pcworld.com/article/257037/ipv6_five_things_you_should_know.html
• Static vs. Dynamic IP addresses – There are a limited number of IPv4 addresses which can be assigned by ISPs to machines that connect to the Internet – Most home IP addresses are dynamic and are periodically reassigned (usually assigned at the home router level, and the router
tracks your machines on the internal home network using separate private IP addresses)
• Composition of IP addresses – Generally, the part on the left corresponds to the network, and the part on the right corresponds to the specific machine – Allocated in hierarchies of blocks that read from general to specific, left to right – There is no set of rules or patterns to read these blocks (like there is with a zip code for example), instead there are databases
maintained for looking up IP allocations – GeoIP lookup databases are maintained by various services for identifying geo location by IP address.
57
Death of The Cookie?
• This is really a conversation about 3rd party cookies, not first party
• In general, third party cookies have a shorter shelf life than first party cookies
• Recent studies suggest that about 40% of devices don’t accept third party cookies. Upwards of 60% of cookies may be deleted within 30 days (including mobile devices)
• Third party cookies are most often not deleted by user, but by spyware or antivirus software
58
What About Cookie Tracking on Mobile
Devices?
• Third party cookies have limited tracking usage for mobile devices
– Most mobile devices don’t accept cookies by default
– Concern as well that long term viability of these cookies may be in question for PCs
• In April 2013 Apple exposed a new device ID for tracking at user level (IDFA) within IOS6. Users can opt out of tracking.
• Vendors are emerging that are creating persistent device IDs for targeting and attribution
• Vendors are emerging that are creating persistent device IDs for targeting and attribution
– Vendors include Ad Truth, BlueCava, Tapad and others
– ID persistence length varies by device
– Vendors use combination of deterministic and probabilistic ids
– 80%+ mobile device coverage/accuracy is possible today
• Many ID tracking can be used in conjunction with ad privacy compliance solutions (ex. TRUSTe)
59
What About IP Addresses?
• It is harder to find published data IP uniqueness.
• Most of what I have learned has come from confidential communications with IP data providers and demand side platform vendors (DSPs)
• About
• 85-90% of US IP addresses can be accurately tracked back to a DMA
• 60% of devices with an IP address can be traced back to a known SCF and about 45% to the zip level
• 25 to 35% of IPs can be reliably tracked back to a residence over at least one month’s time
• This is likely to get worse before it gets better as we are “running out of IPv4 addresses”
60
Device Fingerprinting
• Device fingerprinting is emerging as one way to resolve third party cookie deletion issue
• Originated out of fraud detection and has migrated to marketing
• We estimate that many fingerprint technologies are more than 90% accurate. Click here https://panopticlick.eff.org/
• Biggest issue is privacy and adoption to date is still relatively low
• Companies such as Bluecava, and Iovation specialize in this area
61
Our Observations About Digital Data
Landscape
• First-party customer data generally has the highest marketing value
– There are many opportunities for companies to collect first-party digital data across digital medias and
channels
– Most companies do not a cohesive plan for utilizing first party digital data
• Third-party digital data is still in its infancy resulting in opportunity and risk
– Shirting legal environment has huge implications for using third party data (Ex. Internet Explorer Do
not Track). Legal should be involved in strategy development
– Difficult to determine the quality and integrity of digital data providers
– Audience scaling is still a big challenge
– Quantitative approach is necessary to locate and extract value from third-party sources (Example
Merkle Digital Data Optimization Lab)
• Three capabilities are critical to companies creating competitive advantage within digital data
– Ability to effectively identify and extract digital data with business value
– Ability to integrate across digital and offline data sources
– Ability to utilize both online and offline customer data in real time interactive environments
62
Digital Data Sources (Digital media and Channels)
Site Display Social*
Party Identifiers
First Party Data Capture
(Example)
DATA “DEPTH”
DATA “BREADTH”
Primary First Party Data Systems DATA GENERATORS
Third Party Data Providers
(Example) DATA MARKETPLACE
63
Digital Data Sources (Digital media and Channels)
Site Display Social*
Party Identifiers
First Party Data Capture
(Example)
Third Party Data Providers
(Example)
• Cookie ID (Primary) • IP Address • Order ID, Cust#, Profile ID
• Cookie ID (Primary) • IP Address • Order ID, Cust#, Profile ID
• Social Handle (Primary) • Email (Facebook)
• Browser User agent (IP geo, OS, browser type, etc)
• Referral site
• Campaign data (SEO, SEM, Banner clicks, email clicks)
• Internal site search
• Engagement on site (clicks, views, downloads, etc)
• Conversion on site (email signup, purchases, quotes, information requests, etc)
• Browser User agent (IP geo, OS, browser type, etc)
• Ad impressions
• Ad campaign meta data
• Ad clicks
• Ad site conversions—post ad view or click (quote, purchase, etc)
• FB user profile data including likes, interests, geo, etc
• FB friends email/profile data
• FB own site wall posts
• FB custom social engagement (site, apps, etc)
• Other engagement based on specific social network (Twitter, Linkedin, etc)
Primary First Party Data Systems
• Web Analytics tools (Omniture, Coremetrics, etc)
• Ad servers (DFA, Atlas) and DSP (Media Math, Turn, [X+1])
• Social Networks (Facebook*, Twitter) and social platforms
64
Digital Data Sources (Cont.)
Party Identifiers
First Party Data Capture
(Example)
Third Party Data Providers
(Example)
• Device ID (Primary) • Cookie ID (Primary) • Order ID, Cust#, Profile ID
• Cookie ID (Primary) • IP Address • Order ID, Cust#, Profile ID
• Email (Primary)
• SMS send and click
• Mobile site browsing
• Campaign data (SEO, SEM, Banner clicks, email clicks)
• Geo location
• Custom App engagement data
• Search Ad clicks
• Search campaign meta data (keywords, bid amount, cost, creative, etc)
• Ad site conversions- post ad click (quote, purchase, etc)
• Email Send
• Email open and click
• Email campaign metadata
Primary First Party Data Systems
• SMS platforms (iloop),Web analytics, apps
• Web analytics platforms, search ad platforms (Kenshoo, Marin)
• Social Networks (Facebook, Twitter) and social platforms
Mobile Search Email
65
The Customer Event Stream Connects Cross-channel
and Media Interaction Data
The Customer Event Stream is enabled as the customer engages with the brand
DM
Delivered
2/1/2012 Patient
Home Address
Email Address
Mobile #
Cookie ID
Ad ID
Shown Display Ad
2/2/12 3:05pm
Visits branded site
and signs up for
free voucher.
Provides Email
2/2/12 3:06 pm
Sent Email
2/2/12 5:05pm
Opens Email
2/2/12 9:30 pm
Visits clinic and
receives
brochure for
compliance
program 2/6/12
9:00 pm
Signs up for
patient program
via mobile
2/6/12 9:15 pm
66
User ID Date Time Event ID Event Description
1234 2/1/2012 DM437 DM Delivered
1234 2/2/2012 3:05 pm DI9076 Display Impression
1234 2/2/2012 3:06 pm CC068 Signed up on site
for free voucher
1234 2/2/2012 5:05 pm EM087 Sent Email
1234 2/2/2012 9:30 pm EM088 Opened Email
1234 2/2/2012 9:30 pm EM089 Clicked Email
1234 2/6/2012 9:00 pm PS674 Clicks Paid Search
1234 2/6/2012 9:15 pm Q8740 Mobile Enrollment
User Event Table
Event Meta Data
Event ID EM087
Creative A2346 Fight depression
Offer OI92365 30 day trial
Product P978 Rx Description
Customer Event Stream Activates Cross-Channel
and Media Interaction data
67
DM
Delivered
2/1/2012
Shown Display Ad
2/2/12 3:05pm
Visits branded site
and signs up for
free voucher.
Provides Email
2/2/12 3:06 pm
Visits clinic and is
given brochure
for compliance
program 2/6/12
9:00 pm
Sent Email
2/2/12 5:05pm
Opens Email
2/2/12 9:30 pm
Signs up for
patient program
via mobile
2/6/12 9:15 pm
Patient
Connected Recognition Enables the customer Event Stream
Granular attribution allow us to fractionally assign credit to
each touch point into event stream prior to conversion
68
15% 20% 20% 40% 5%
Event Date Cost
Attributed
Credi
t Value
DM Delivered 2/1/2012 $.35 .05 $100
Display Impression 2/2/2012 $.001 .20 $400
Microsite engagement 2/2/2012 $13.20 .30 $600
Sent Email 2/2/2012 $.02 .10 $200
Clinic Brochure 2/2/2012 $12.50 .15 $300
Mobile Enrollment 2/2/2012 $.03 .20 $400
Predicted Customer LTD Value: $2,000
Customer Level Attribution
Program Level Attribution
This scenario represents success in that the predicted customer value is
realized/confirmed and there is a strong program ROI.
Campaign Display-DSP
Spend $10,000
Impressions 1,000,000
Inc TRx 1,320
Inc NRx 102
Value per Rx $30
Total Value $42,660
ROI 327%
Patient
Measure
Assess
Tune
Fire trigger email based
on website interactions
User receives email with
important information
about their disease state
with link to web page
with discussion points for
their visit with physician
Patient program
brochure picked up in
physicians office
Contact Management
Manages user interaction strategy and rules
Value is Unlocked as We Can Influence the
Customer’s Future Behavior
69
DM
Delivered
Shown
Display Ad Visits branded
site and signs
up for free
voucher.
Provides Email
2/2/12 3:06 pm
2/1/2012 2/2/12
3:05pm
2/2/12
3:06 pm
Personalization
Dynamically assembles personalized
communication package
Intervention strategy and rules are used to aid customer to next step in conversion process
Mobile call to action
enables customer to
easily sign up for patient
program while leaving
physician’s office. User
immediately receives
mobile coupon.
Patient
Opportunity to Drive Smarter Planning and Messaging
at the Segment and Customer level
70
Which individuals and segments should we target?
What channel should this individual be communicated
through?
Given the potential value of this customer how much
should I spend to impact behavior?
How often and in what sequence should I
communicate with this prospect?
Given their history what offer, service, or
communication should be delivered?
What product would this individual most likely be
interested in?
What is the best way to engage with this customer?
How frequent should contacts be?
Targeting
Best Media/ Channel
Allowable Spend
Contact
Optimization
Offer Optimization
Product / Disease
State
Messaging
How data can be used to drive more targeted communications
Digital Targeting and Personalization
Value is Unlocked Within The
Digital Marketing Value Chain
72
First
Party
Data
Second
Party Data
Third
Party
Data
Inte
gra
tio
n
Optimized Channel
Experience
(Targeting and
Personalization)
Measurement and
Budget Allocation
(Attribution)
Connected Customer
Insight
(Data Integration)
The Digital Marketing Value-Chain
Today, Consumers are…
Engaged in an ever expanding number of channels, which is challenging business leaders to broaden channel reach & execution capabilities
Barraged by an increasing number of messages and communications
Expecting personalized and relevant interactions; they are self-selecting to engage with brands that provide relevance and timeliness
Assuming brands are aware of their past interactions and expect brands to use this data to manage a worthwhile relationship dialogue
73
But, Most Business Leaders Approach
To Personalization Is Patchy…
Source: Forrester Research “Use Customer Analytics to Get Personal”, by Srividya Sridharan February 17, 2012
• Emphasize digital channels only
• Focus on superficial customer attributes
• Fail to determine causal impact of personalization
• Project aggregate group behavior to individuals
• Rely on asynchronous customer data
• Encourage channel myopia
• Missing the real-time dimension in their approach, thinking and capabilities
74
2000 2013 2006 2003 2009
The Evolution of Market Leaders in
Personalization
SOLUTION-FOCUSED CHANNEL-FOCUSED CUSTOMER-FOCUSED
Capability • Recommendation systems
−Collaborative Filtering −Content-based Filtering −Ensemble Learning
• Content/offer optimization −Segmentation −1:1 Predictive Modeling
Data - Limited to a small set of relevant customer interactions Experience - Isolated personalization interaction
Capability - Personalization execution silo'd in channel specific tools (web, email, display advertising, search)
Data - Channel specific customer
interaction and profile data Experience
• Relevant communications • Inconsistent experience across
channels
Capability • Integrated mobile and social • Ability to optimize timing and
delivery • More control for companies
to optimize decision logic Data - Integrated customer
interaction data across online and offline channels
Experience
• Improved relevance • Consistent experience across
channels
Isolated techniques limited to one or two personalized interactions
Disparate approaches to personalization primarily achieved in channels separate from each other
Coordinated solution across multiple interactions and channels that leverages a complete view of the customer
75
Levels of Personalization Maturity
Level 5 Leading Edge
Optimal Personalization [Contextual Relevance] • Combines multiple personalization enablers to give a multi-dimensional understanding at a
journey stage • Not only informs but also influences the customer’s mindset • Delivers a unique and competitive customer interaction • Addresses customer values: Contains the prevailing emotional criteria that best informs
customer decision behavior • Considers what will motivate: Has behavior stimulus that best connects with the Customer
Values to deliver the necessary response from the customer
Level 4 Consistent Best Practice
Moderate Personalization [General Relevance] • Provides data-driven, relevant content and product offers based on general customer
attributes • Timely; applicable content or offer; addresses customer preference(s) • Aligned with Brand/Promise drivers
Level 3 Industry Competitive
Limited Personalization • Focuses primarily on segment’s channel preferences • Delivery is appropriate and optimized for the media; channel, platform; addresses heuristics,
Level 2 Developing, Inconsistent
Sporadic Personalization • Mass-only attributes considered for content, limited to general versioning (region, language) • Inconsistently optimized for media or channel; only occasional personalization; limited
measurement; lack of data-driven content
Level 1 Limited to No Capability
No Personalization • Not optimized for media, channel or platform; no personalization; not measured for
performance; static content, no versioning
CA
PAB
ILIT
Y
High
Low
76
Personalization Is A Process, Not An
Outcome
Source for Image and Quotes: Forrester, February 17, 2012, “Use Customer Analytics To Get Personal”
Where do we personalize next? Do we understand all of the decisions that are in
place and by who?
Do we have the right data? Is that system integrated yet? Is it fast enough and does it scale?
Is the content written? How do we manage changes?
Will this rule conflict with existing rules? How do we manage so many different yet related rules?
Is the experience consistent across all channels?
How do we know this is working? Is it working
because of what we are doing or someone else?
How can we continuously improve? How do we react quickly and confidently?
77
Analytical
Marketing
Database
Decision Services
Interactive Conductor
Web Service
API
Da
ta
Insig
hts
Decision
Management
Business
Rules Testing Optimization
Decision Management Components
Underlying technology architecture supporting channel-specific technologies enabling consistent, personalized customer experience across touch points.
CC DM EM Display Search Social Mobile Site Agent
Campaign Feed
Batch Lists
Channel-specific interface File delivery - latent
Insig
hts
Da
ta
Benefits:
• Rules engine to govern customer interactions
• Integration with AMD for insight driven communications
• Open environment for Omni-channel connectivity
• Real-time capability for timely communications
• Testing and machine learning for continuous learning and tuning
78
Industry Example
Dr. Jones
Dr. Jones, a sub-optimized HCP with growth
potential, receives a rep call to discuss ease of use...
79
Industry Example
Sarah is a CV patient at risk of a serious health event. She searches
Google for treatment options after her PCP visit…..
80
Industry Example
Customer Journey Driving Awareness For New Hospital Facility
Sally ….
81
• Customer Data Integration
• Single View of the Consumer
• Data Insights
• Opportunity Discovery
• Decisioning Development
• Testing and Optimization
• Behavioral Impact
• Personalization Priority
• Interaction Design
• Media Planning
• Channel Integration
• Asset/Media Development
• Delivery Requirements
• Planning
• Setup and Decision Configuration
• Reporting and Monitoring
Integrated Personalization Solution Overview
• Data Management Platform
• Integrated Marketing Data Warehouse
• Predictive Analytics Tools (e.g. SAS, R)
• Decision Engine
• Testing Module
• Content Management System Integration
• Channel Personalization Tools/Plug-ins
• Campaign Management Tool
• Reporting and Dashboards
TEC
HN
OLO
GY
DATA ANALYTICS EXPERIENCE EXECUTION
PR
OC
ESSE
S
82
Personalization Engine Analytic Methods
• Cross channel personalization engines should support a variety of analytical methods
• Most single channel recommendation products just rely on information filtering since it is easiest to automate
Method Description Examples
Information Filtering (Highly automated & self learning)
Machine learning based techniques • Content based filtering – utilize discrete characteristics of
an item in order to recommend additional items with similar properties [More limited-easier to get started]
• Collaborative based filtering- “User behaves like this (or has preferences such as this) look like another user who likes/ purchases xyz” [More robust-cold start problem]
• Hybrid filtering- Combination of content and collaborative approaches [Best but most complex]
Decision Rules-Tree (Very custom, sequencing)
• Trigger actions- If user does this then do that • Adaptive rules- Next action or content varies based on
sequence of actions taken by user
Propensity Models (Custom models for few important decisions)
Statistical modeling based techniques • Next best product/offer/action modeling- Used in cases where
fewer offers, products, or options but rich consumer history
83
How data can be used to drive more targeted communications
Cross Channel Measurement
History of Marketing Mix Modeling and Attribution
MMO begins as
custom one-off
projects
1940s-1970s 1980s 1990s 2010s
• Mainly academic
until 1970s
• First MMO
product in 1979
• MMO is panel
based, similar to
attribution today
Low adoption,
lack of data, lack
of computing
power
Audiences
aggregated
by content
Audiences
aggregated
by content
Audiences
aggregated
by content
MMO (top down)
and Attribution
(bottom up)
unify
Digital media
disrupts MMO
industry.
Recovers by late
2000s
MMO scales
outside CPG to
include Auto,
Finance and
Pharma
Modern MMO
emerges in CPG
Industry
2000s
Syndicated
scanner data
revolutionizes
industry
Mathematics of
paid digital fixed,
computation cost
falls to $0
Computer power
increase (still
mainframes
though)
Mathematics of
digital need to be
created.
• MMO becomes
the standard
approach for CPG
• Models become
more complex
• First digital
models in 1999
• Bayesian,
Markov, agent-
based and other
models emerge
• First attribution
models in 2005
(based on 1979
panel models)
• Computing
problematic
• Regression
• Panel
approaches fade
(will remain as
forecasting tools)
• Focus on speed
and actionability
• Implementation
becomes
implementation
• Social become
the next frontier
85
3% 14% 3% 5% 5% 5% 15% 5% 5% 40%
0% 100%
Merkle Recommends a Modeled Attribution
Approach Across All Media
Day 8-30 Day 1-7 Day 0-1
New
Customer
Actual
experience
Credit over applied to bottom
of funnel touches. Other
touches often ‘invisible’
Creates flawed financial
view of performance
Direct or
Rules Based
Modeled Model-adjusted interaction
Most accurate and
actionable
$
TV view Direct mail sent Print view Display view Social visit Website visit Paid search click
Mass and Offline Digital
Assess media performance by measuring the incremental impact of each marketing activity
86
Granular attribution allow us to fractionally assign credit to
each touch point into event stream prior to conversion
87
15% 20% 20% 40% 5%
Event Date Cost
Attributed
Credi
t Value
DM Delivered 2/1/2012 $.35 .05 $100
Display Impression 2/2/2012 $.001 .20 $400
Microsite engagement 2/2/2012 $13.20 .30 $600
Sent Email 2/2/2012 $.02 .10 $200
Clinic Brochure 2/2/2012 $12.50 .15 $300
Mobile Enrollment 2/2/2012 $.03 .20 $400
Predicted Customer LTD Value: $2,000
Customer Level Attribution
Program Level Attribution
This scenario represents success in that the predicted customer value is
realized/confirmed and there is a strong program ROI.
Campaign Display-DSP
Spend $10,000
Impressions 1,000,000
Inc TRx 1,320
Inc NRx 102
Value per Rx $30
Total Value $42,660
ROI 327%
Patient
87
Promotion Mix Solution
(Top Down Approach)
PR
OM
OT
ION
RE
SP
ON
SE
AN
AL
YS
IS
Shar
e C
han
ge/V
olu
me
Direct Mail
Mobile
Display / Search
Rep Details
Samples
Tele-Detailing
Managed Care
Physician & Patient Demographics
Promotion Mix Modeling
Insights: •Impact of Personal Promotion
• Impact of Non-Personal and Other Promotion
•Promotion Response curves
• Total and Marginal ROIs
Channel Inputs:
Brand Managed Care Status Competitor Share Competitor Managed Care Status Physician Attributes
Market Factor Controls
NRx Volume/
Market
Share
= Carryover
Effects
Personal
Promotion
Efforts
Non-Personal
Promotion Efforts TREND &
Others + + +
Speaker
Programs/
Seminars/
Journal Advtg.
+
• Promotion Mix Modeling is an econometric technique used to quantify the impact of promotion
spend on sales. It uses historical time-series data to measure the promotion impact
88
Output Data
0
10
20
30
40
50
60
1 3 5 7 9 11 13 15 17 19 21 23
Reve
nue
Time
Sales Decomposition
TV
Direct Mail
Radio
Base
Base51%
Print14%Radio
8%
Direct Mail11%
TV16%
Segment 1
p
i
t
p
ititixy
1
-
100
200
300
400
500
600
Segment 1 Segment 2
280 298
73 73 42 46 62 49
87 71
Reve
nue
TV
Direct Mail
Radio
Base
Statistical Models
Brand Sales = “B”*Units of Touchpoint
Example: For every direct mail piece sent via iConnect, sales increase by 2.5 Rx and for every
email sent, sales increase 0.3
Sales = 2.5 * Direct Mail Units + 0.3 * Emails
0
50
100
150
200
250
300
350
0 5 10 15 20 25 30
Time
TV
0
200
400
600
800
1000
1200
0 5 10 15 20 25 30
Time
Direct Mail
0
50
100
150
200
250
0 5 10 15 20 25 30
Time
Radio
0
20
40
60
80
100
120
140
160
180
0 5 10 15 20 25 30
Time
Input Data
Details
Mobile
Segment 1: Sales = 3.1* Direct mail + 0.8 * Email units + 0.5 * (PDEs)2
Segment 2: Sales = 2 * Direct mail + 0.1 * Email units + + 0.4 * (PDEs)2
Segment 3: Sales = 0.5 * Direct mail + 1.5 * Email units + + 0.2 * (PDEs)2
HYPOTHETICAL DATA Segment view enabled through use of random effects in mixed modeling approach
Model Selection:
Output Benefits Based on Data Structure
89
Example: Channel Contributions
Details 30%
Samples/Detail 3%
Spot TV 1% National TV
19%
Print (News) 3%
Digital Display 3%
Paid Search 2% Speaker Programs
2%
Direct Mail 1%
iConnect DM 1%
Email 1%
Carryover 34%
Top Down Model provides a High Level Contribution Allowing Us to Allocate Total Spend
Budget and Assess Historical Performance
90
We Need Both Top Down And Bottom Up
Measurement Methods
Bottom-up (Customer Level Data)
Top-down (Aggregated Data)
Display
$60
Video
$80
Search
$91
Direct mail
$75
Social
$113
Branded - $87 Video 1 - $121
Video 2 - $35
Video 3 - $213
Video 4 - $23
Not branded - $99
Social 1 - $50
Social 2 - $163
Social 3 - $456
Remarketing - $12
Programmatic - $80
Guaranteed - $130
National media (TV & radio)
$140
Local media (TV & radio)
$200
Direct mail
$180
Digital
$83
DM 1 - $11
DM 2 - $93
DM 3 - $210
DM 4 - $235
Integrated
measurement
All measurement levels
(Media, platform,
campaign, placement)
All measurement dimensions
(Customer segment, product,
geography)
91
Integrated Attribution Provides Output Within
Measurement Levels and Dimensions
• More accurate view into media
performance
• Important input into ongoing
budgeting and planning processes
Media-level results
Monthly
• Visibility into how each tactic was
driving new customers by segment
• Important data to feed into
customer experience to drive better
personalization and targeting by
tactic and segment
Segment-level results
Monthly
• Visibility into “why” different
programs were and were not
performing
• Diagnostic data markers can use to
adjust existing programs
Campaign diagnostics
Daily
$0
$50
$100
$150
$200
$250
$300
$350
$400
January February March April May
CPA by Channel
DM Brand TV Display
Email Organic Search Paid Search
Segment Penetration (Index)
Campaign Segment 1 Segment 2 Segment 3 Segment 4 Segment 5
Display 1 120 90 100 120 85
DM 1 95 75 55 95 105
Alt Media 3 130 50 90 130 114
Display 1 120 95 50 120 87
DM 4 55 150 140 55 79
Social 1 95 200 143 95 100
Display 1 85 75 22 85 75
Search 1 200 98 100 200 97
Social 1 75 101 75 75 120
Display 2 30 130 120 30 115
Overall 100 100 100 100 100
Performance Drivers
CampaignEffective
CPM
Contact
Frequency
(Within
Media)
Contact
Frequency
(Across Media)
Unique
Response
Rate
%
Remarketing% Exclusive
DM 1 5.28$ 3.0 12.0 0.1565 10% 10%
DM 2 1.69$ 3.9 15.0 0.0552 20% 20%
Alt Media 3 18.87$ 3.1 7.0 0.6122 15% 15%
Alt Media 7 0.34$ 39.4 45.0 0.1096 10% 10%
Search 1 0.39$ 5.3 6.0 0.0130 20% 20%
Social 1 1.71$ 66.0 78.0 0.5884 15% 15%
Email 1 1.44$ 2.1 5.0 0.0162 10% 10%
Search 1 44.99$ 5.3 10.0 1.2132 20% 20%
Display 4 96.41$ 2.5 30.0 1.3916 15% 15%
Display 2 0.63$ 1.5 18.0 0.0049 10% 10%
Total 0.66$ 4.0 22.6 1.812% 10% 10%
92
Analytics and Modeling - Things to consider
• Estimation of touch point effects
• Estimation of paid search clicks
• Time effects
• Repeated touch points
• Retargeting
• Sequencing and assists
• Validation
• Attribution Formula
93
Insights Portal
Centralized location to access understand historical performance, plan for future, set up targeting, and analyze
customers
Centralized Insights Portal
* Note: all data changed to protect confidentiality
Media Targeting and Personalization
Customer Insights
Measurement and Attribution
Planning and Forecasting
Integrated Dashboards
14 * Note: all data changed to protect confidentiality
94
Enterprise Measurement Platform Requirements
Solution is Focused on An Enterprise Measurement
Platform, Not Just a Digital Attribution Tool
Support all media Best KPI accuracy at any level Ability to report out KPIs based on key, customized
business dimensions
Integrated and customizable performance reporting
Robust decision support Tight integration with marketing database
Scalable and flexible solution architecture
Action support, not just product support
Digital, mass and offline direct Media, campaign,
placement/keyword Segment, Product, Geography,
and Time
One place to go for program performance, insights, and
analysis
Ability to run what if scenarios and machine
optimization to create the best plan at any level
Ability to push data from marketing database to
attribution platform and back into database for action
Big data platform with robust customer identity
management and ability to absorb frequent changes to
data and requirements
Help lead change management and ongoing
insights extraction and action processes
95
Output Module
Attribution + Media Planning/Optimization Engines Inbound Channels Insights Portal
Store
Display Sms
Social DM
Analytic
Modeling
Data Collection
Transform Event
Stream
Reports/
Dashboards Interface
Recommendations
/ Targeting
Optimization
• Fraction allocation model with comparative techniques
• Multi-stage statistical modeling approach (logistic regression)
• Use of all available information; principal components on
300+ variables
• Forward looking scenario planning capability
• Target individuals and apply
recommendations
• Integration with DSPs, search
platforms, etc.
• Portal Interface Wrapper
• Tableau Reporting Solution
Merkle Attribution Solution –Reference Architecture
Planning
Tools
Outbound Media
Call
Center
Site
Search
TV
Radio Email
Input Intelligence Action
96
Input
Merkle Attribution Platform Physical View - Modules & Process
CR
Intelligence Action
97
Measurement Output Must be Easily Integrated Into
Targeting Algorithms
Cookie
Keyword/
Cookie
User ID Conversion
ID
Event
ID
Attribution
Weight
1234 C76532 DM437 .05
1234 C76532 DI9076 .32
1234 C76532 PS674 .11
1234 C76532 Q8740 .25
Model-based
attribution weights
Digital
platforms
Targeted
ads
Real-time
bidding
Publisher
Publisher
Engine
Engine
Attribution data
1000101110101
0100111001110
Demand Side
Platform
(DSP)
Search Bid
Platform
An
on
ym
ou
s ta
rge
ting
Anonymous Data
98
SUMMARY
Financial management must evolve to
an enterprise-wide initiative to be most
effective
Enterprise scale
is necessary
Analytics and technology must be
tightly integrated to create these
solutions
Analytics alone
is not enough
Significant value can be created by
taking even a few steps forward in the
evolution of the four Financial
Measurement capabilities
Value potential
is enormous
99
Attribution
CASE STUDY
Customer level modeling used to understand relative contribution of each marketing touch to the ultimate conversion activity.
Attribution Modeling Approach
Probability • Assemble conversion
sequences across direct and
digital media
• Given any sequence of
interactions, calculate the
probability of conversion for
that sequence
• By comparing these
conversion probabilities for
interaction sequences, isolate
the individual impact of each
of the interactions and assign
a weight to it
$
Conversion for the sequence with display 1 interaction
Conversion for the sequence without display 1 interaction
Response
Probability $ Display
1
Weight for D1 = [ Probability(conversion for the sequence) - Probability (conversion for the
sequence without D1)
Display 2
Search 1
Display 2
Search 1
Response
Probability
101
Industry Example
Example: Data Flow
Doubleclick (Display, mobile, video,
email, mobile video)
Client M site log (Conversion events, landings
from natural search)
Search providers (Paid search)
Cookie ID with all touchpoints, conversion events
User Shorthand TP stream
Conversion
293832704 DDVSMPVC Converted
99920125 DSVPNV Didn’t convert
220M TPs per week
130M TPs per week, 13K conversions per week
180K TPs per week
102
Industry Example
Example: Model Details
Primary Conversion event Design Yours
Secondary Conversion event Online purchase
Channels Display Paid Search Video Social Mobile Mobile Video Email
Measurement Modeled attribution (attributes each conversion event among that user’s touchpoints)
Model Inputs
Exposure Recency and Decay
Interactions between media Control for Sequence
Frequency
Media Type
Funnel Stage Model remarketing pool expansion Model search
Conversion Type
103
Industry Example
Example: Model Insights –
Significant Predictors
======================== Predictor Sig Non ======================== creativeid 183 22 pageid 693 128 buyid 3 0 siteid 67 1 countryid 49 95 state 57 9 browserID 11 1 browserVersion 41 26 osID 9 2 adid 111 135 creativetype 8 0 creativesize 12 0 DayofWeek 7 0 timeofday 4 0 mediaName 7 1 ======================== Total 1262 420 Gini: 45% +/- 1% Week of 10/13/2013
104
Industry Example
Model Insights - Transference Maps
Purchase
105
Industry Example
Example: Reports
Reports tend to look like typical media optimization reports
106
Industry Example
Example: Overall Results
Based on initial results, by optimizing digital media spend, Company M is able to:
• Reduce expenses by 12.4%. Total expenses for the time period was $22 MM.
• Increase sales and/or site engagements by 10-20% for all future products when using advanced attribution.
107
Industry Example
Example: Recommendations
• Leverage advanced attribution methods
– Out performs last click optimization.
– Estimates more accurate CPA’s over time.
• Optimize between and within media channels
– In addition to the media spend costs savings, there is ample opportunity to improve within channel optimization (10-15%).
• Optimize on intent metrics
– Non-branded paid search drives intent but very few online purchases.
– Mobile advertising drives intent but no online purchases.
– Expand paid search keywords: Select Script Fill keywords drive Intent at decent CPAs (<$130):
• Improve display efficiency
– 10% of total impressions goes to a small sample that received 30+ impressions. Capping frequency can provide significant cost savings.
• Use the xyz Network as the anchor and round out media spend by supplementing with other not overlapping
– The xyz Network in particular has a significant overlap with other publishers
108
Industry Example
Example: Insights
• Segmentations – Chrome is the most common browser driving intent. – Safari users are less likely to be interested caregiver content. – Although only 4% of all impressions were served on an iPad, iPad represented
10% of attributed impressions associated with script fills.
• Best and worst performing sites – In general, sites that drive intent also drive conversion. – Site 1 and Network A are an exceptionally strong performers for intent and
conversion.
• Remarketing (Initial insights) – Remarketing is not as effective for driving intent as it is for driving purchases. – Remarking CPA’s are better than overall CPAs. The purchase CPA is 9 times
lower than average
• Channel drivers/funnel – Display is higher in the funnel and drives all other channels. – Video is an important part of the user journey. The relative value of video
changes from week to week
109
When You Get Back to Work …..
Realities of Working With Digital Data
Working With Your Agency
Realities of Working With Digital Data
• Gathering Data is the big challenge – 80% to 90 % of work may be data related – 10% to 20% of work may be modeling and insight development
• Common data Challenges:
– Agencies won’t share data (or don’t know how to share)
– Data will be missing in certain channels (especially social)
– Costing data is incorrect
– Data won’t mean what you think it means
• Determining meaningful conversion events is second biggest challenge – Mapping digital behaviors to Rx
• Recommended Best Practices
– Do beta engagements now
• Test conversion events
• Attempt to pass and interpret data
– Set expectation correctly on timing (at least a Quarter)
– Level set with agencies – you (the client) owns the data
111
Working With Your Agency
• Demand better ROI Modeling from your agencies – Although this requires work, good agencies like proof of ROI
– If you agencies are uncomfortable, this may be the sign of bigger problems
• Review Your Measurement Frameworks – What is your approach to attribution and ROI?
– Is this socialized through your organization?
• Insist your agencies collaborate – Agencies work best with clearly defined roles and Direct and clear clients
– Agencies prefer strong leadership
• Plan a beta test – Get access to raw data (test that agencies can deliver it or you have access)
– Try to analyze it yourself
– Review attribution marketplace
– Do a one-off pilot project
112
Working With Your Agency
• Get segmentation on your paid and owned experiences – Ex: How do caregivers behave on your website vs. script holders?
– Ex: How do you drill your media plan to different segments?
– Challenge yourself to understand the techniques for instrumenting segmentation
• Learn programmatic media – Not huge in Pharma today
– But it will be huge soon – Get Ready!
113
Three Exercises
114
Related Documents
