6. «It is a capital mistake to theorize before one has data. Insensibly one begins
to twist facts to suit theories, instead of theories to suit facts»
– Arthut Conan Doyle, Sherlock Holmes
«More data means more information, but it also Means more false information»
– Nassim Nicholas Taleb, Antifragile: Things That Gain from Disorder
«Information is the oil of the 21st century,
and analytics is the combustion engine»
– Peter Sondergaard, Gartner Research
8. DATA PREPARATION
Get as much as possible
Merge into a single database
DATA USE
Advertising
CRM marketing
Usability
DATA MEASUREMENT
Smart attribution
Omnichannel
Econometrics
We are here
9. WORKSHOP SCHEDULE
10.00 - 10.20 - Introduction. Significance of the topic.
10.20 - 11.30 - How to get the most from existing data? Data availability.
11.30 - 11.45 - Coffee break
11.45 - 13.30 - Down the purchase funnel . Marketing activities with unified database.
13.30 - 14.30 - Lunch
14.30 - 15.45 - How to be perfect in e-com marketing activities?
An ideal ad split, smart attribution and the power of econometrics
15.45 - 16.00 - Coffee break
16.00 - 16.45 - Wrapping up
16.45 - 17.15 - Group discussion
10. PART ONE: HOW TO GET THE
MOST FROM THE EXISTING DATA.
DATA AVAILABILITY
14. BASIC DATA FROM THE CLIENT
NAME /
SURNAME
EMAIL /
PHONE
NUMBER
BIRTHDAY SEX INTERESTS
15. z
Give a reward to a client:
Discounts for birthday data,
special occasional discounts for
specific categories, etc.
MORE DATA FROM
THE CLIENT >>
MORE PROFIT
21. GOOGLE BIGQUERY – A FULLY-MANAGED DATA
ANALYTICS SERVICE IN THE CLOUD
22. Google Analytics free account has
sampled reporting and sometimes
produces skewed data
GA STANDARD GA 360 (PREMIUM)
FREE
Sampled reporting from
500 000 sessions
$150 000+ / year
No sampling up to
100 000 000 sessions
23. USER ID TO GOOGLE ANALYTICS
USER/CUSTOMER ID
CRM SITE
GOOGLE
ANALYTICS
24. TYPES OF USERS WHO VISIT THE SITE
CLIENTS,
REGISTERED
CLIENTS,
NON-REGISTERED
NON-CLIENTS
Any info
available
No info
Can be obtained
it there is a
reason to sign in
No info
To create a
purpose for
signing up
25. HOW TO USE UNIFIED DATA
SINGLE DATABASE - ONE
SOURCE OF DATA
ADVERTISING
SITE
PERSONALIZATION
CRM MARKETING
26. TECHNICAL SOLUTION FOR CRM AND ONLINE
DATA UNIFICATION
an official distributor of Google
solutions and Google partner
for Cloud Console
LINA YARYSH
28. PART TWO: HOW TO THE DATA FOR
EXISTING AND PERSPECTIVE CLIENTS
29. DATA PREPARATION
Get as much as possible
Merge into a single database
DATA USE
Advertising
CRM marketing
Usability
DATA MEASUREMENT
Smart attribution
Omnichannel
Econometrics
We are here
30. SUMMARY OF THE FIRST SECTION
01Data from
clients (as
much as
possible)
02User-ID into
the site and
GA
03Online data
streaming
into Google
Big Query
04CRM data
regular
download
into Google
BigQuery
32. In Google and in social media
Existing
customers
Similar audiences
to existing
customers = look-
alike audiences
Video LAL audience
Email list LAL audience
Conversion-based LAL
audience
Page likes LAL audience
LOOK-ALIKE AUDIENCES IN FACEBOOK
43. EMAIL MARKETING AND CUSTOMER MATCH SYNERGY
Online shoe
shop
Email
marketing
Open
Not open
“Reminder: Get
shoes from new
collection! Use
Code XYZ For An
Extra 10% Off!”
“Introducing new
collection of
summer shoes for
your everyday
activities.”
46. PROGRAMMATIC ADVERTISING – HOW IT WORKS
STEP 1
Someone Clicks on
the webpage
STEP 2
The publisher of the
page puts up the ad
impression for auction
STEP 3
The publisher
holds the
auction among
the advertisers
competing for
the impression
STEP 5
The ad is delivered to
the prospective
customer
STEP 4
The advertiser willing to
bid the most for the
impression wins the right
to display their ad
STEP 6
Customer clicks on the
ad and the advertiser
converts them into
a sale and profits
51. CRM MARKETING – MAJOR STEPS IN THE DEVELOPMENT
ONE MESSAGE
TO ALL
1st
STEP
2st
STEP
“PEOPLE WHO BOUGHT
THIS, ALSO BOUGHT THAT”
PERSONALIZED
MESSAGE
3st
STEP
52. This is Ruta
Affluent young mom,
Homeowner
shops at a national clothing retailer online, in the store,
and occasionally via the app
CRM MARKETING – RUTA’S
EXAMPLE
53. RUTA’S TIMELINE
Online purchase
FIRST PURCHASE
• In search of kids’ toys
• Found a toy that was
offered by the site
based on her previous
buying experience
Three days after
MESSAGE ON
A NEW TOPIC
• The retailer sends Ruta
a health-themed email
• Ruta watches a video
about raising healthy
kids
One week later
PURCHASE OF A
NEW PRODUCT
• Ruta receives an iPhone
message with 15 percent
one-day discount on baby
food
• Ruta purchases a bag
with infant fruit smoothie
“People who bought this,
also bought that”
Personalized message
54. CRM MARKETING – RUTA’S NEW
EXPERIENCE
Before
Now
Increase in lifetime value
55. PERSONALIZATION ROADMAP
01 Unite all possible data on the customer
02 Create triggers for messages
03 Craft the offers
04 Deliver messages
56. TRIGGERS – STEP TWO
Buy of a particular item
Visit of a particular page
Particular geolocation
No buy for a particular
period of time but internet
activity (visits) to the site
Birthday
Buy in a particular category
No buy for a particular
period of time
Buy a several
days/weeks/months ago
A LOT OF
62. Site personalization
“If we have
4.5 million
customers,
we shouldn't
have one
store, we
should have
4.5 million
stores”
Jeff Bezos,
1998
63. SITE PERSONALIZATION
+19%
of revenue
+56%
of customers like their
names mentioned
DYNAMIC PERSONALIZATION
Site changes on
the fly
Generates individual
experiences at the site
Content that perfectly fits
user’s session
65. SITE PERSONALIZATION – MAJOR CUSTOMER’S
CHARACTERISTICS NEEDED
• Demography: Age, race, gender, language, etc.
• Context: Situation in which the visitor views your content
• Customer Journey: Previously bought products, product
categories, average transaction value
• Psychographic: Habits, likes, interests, preferences, etc.
• Geo-location: Where the visitor lives and engages with the
content
• Engagement: Past and current visitor’s interactions with your
content across all channels
• Device: Computer, mobile, or tablet
66. SITE PERSONALIZATION – MAJOR TWO APPROACHES
Personalization
based on the
visitor
Personalization
according to the
product/service
67. TECHNICAL SOLUTION FOR SITE
PERSONALIZATION
The world’s first personalization
technology stack
ANDREY TYSCHENKO
69. PART THREE: HOW TO ASSESS
PROFITABILITY OF WHAT WE HAVE
DONE WITH THE DATA
70. DATA PREPARATION
Get as much as possible
Merge into a single database
DATA USE
Advertising
CRM marketing
Usability
DATA MEASUREMENT
Smart attribution
Omnichannel
Econometrics
We are here
72. TYPICAL MARKETING CAMPAIGN
ALLOCATE
BUDGETS
Divide campaign
into two-three parts –
performance, brand-
oriented, engagement
Assess the results:
- leads in performance
- views, clicks, BR in the
others
73. Performance campaign analytics – typical approach
LTV
calculation
for average service user
P&L
analysis
For the profit from the
user’s revenue
LTV and
profit
calculations
for segments of the
audience
Profit for users from
any audience segments
75. Analytics of the segment – example of LTV calculation
Women 25-34, Kaunas
3 439
Meets criteria of Google and Facebook
Four years with
the service
320
Five years with
the service
31
LTV full
9 orders, revenue - 500 Euro, net profit – 175 Euro
Should we pay for a customer:
60 Euro
150 Euro
167 Euro
76. Fewer years – lower cost of lead
4 years • 175
Euro
3 years • 140
Euro
2 years • 110
Euro
1 year • 70
Euro
83. How can brands meet their marketing goals
across the purchase funnel?
What is the optimal allocation for your
campaign?
How can you make your mobile investment
work harder?
We know last click is lame, but what’s the
better alternative?
QUESTIONS FOR ECONOMETRIC MODELS
85. Econometrics – main formula
Базовый уровень
ttt XXDepVar ...2211
Показывают связь между каждым
фактором (т.к. цена) и зависимой
переменной. В качестве примера
“снижение цены на 1%, приведет к росту
продаж на 1.5%”
Всегда имеется ошибка, так как мы не
обладаем совершенной информацией и
поведение человека в определенной степени
непредсказуемо. Ключевым критерием
является отсутствие корреляции между
ошибкой и каждым из факторов Х
Зависимая переменная
(например, продажи или
трафик)
Каждый X является фактором, который имеет
систематическую, причинно-следственную связь с зависимой
переменной. Участвует столько X-ов, сколько требуется. Они
включают в сеюя такие вещи, как цена, ТВ, активность
конкурентов и т.д.
86. Econometrics – main formula
Period Sales Price Distribution Promotion Income Advt.
Constant/
Base
1
2
3
4
5
6
S1
S2
S3
S4
S5
S6
=
=
=
=
=
=
a x P1
a x P2
a x P3
a x P4
a x P5
a x P6
+
+
+
+
+
+
b x D1
b x D2
b x D3
b x D4
b x D5
b x D6
+
+
+
+
+
+
c x Pr1
c x Pr2
c x Pr3
c x Pr4
c x Pr5
c x Pr6
+
+
+
+
+
+
d x I1
d x I2
d x I3
d x I4
d x I5
d x I6
+
+
+
+
+
+
e x A1
e x A2
e x A3
e x A4
e x A5
e x A6
+
+
+
+
+
+
K
K
K
K
K
K
89. FMCG
Econometrics – modelling
graph – real vs forecasted by
model -100,000
0
100,000
200,000
300,000
400,000
500,000
Week Ending Saturday
UnitSales
Price Distribution & Base Promotions TV Advertising
Demonstrations Seasonality Outdoor Advertising
91. Econometrics – modelling
graph – real vs forecasted by
model
What we saw and
what we can take
from this journey
1. How to get customers’ info
How to assemble all info on
customers
How to use it in marketing and
communication with clients
how to analyze marketing
campaigns