To watch the entire webinar replay, please visit:
http://www.mintigo.com/predictive-marketing-the-science-behind-marketing/
Title: "Predictive Marketing: The Science Behind Marketing"
Description:
One of the hottest trends in marketing and lead generation is predictive marketing. But what does it mean and how does it really work? Can it be implemented by mere mortals in marketing? Or does it require an army of big data scientists and a black box model?
Implementing predictive data for decision making surrounds us today. The challenge is providing advanced analytics without the need for a team of programmers. Join Tal Segalov from Mintigo as he shows how to quickly build predictive models and how to visualize the results for B2B businesses.
In this webinar you will learn:
- Who is already using predictive marketing all around you
- How 20% of your leads give 80% of your business and we have the proof
- The most efficient way to share predictive scores for optimal engagement
- The importance of clean data for building predictive models and constructing visualizations
About The Speaker:
Tal Segalov, COO and Co-Founder at Mintigo
Tal brings more than 15 years of experience in software development. Prior to Mintigo, Tal was AVP Research and Development for modu, the modular mobile handset company. His previous experience includes developing complex, large scale data analysis systems. He holds a B.Sc. EE and a B.A in Physics from the Technion – Israel’s leading school of technology. He also holds an executive MBA from Tel Aviv University.
2. Host:
Tony
Yang
Director
of
Marke2ng
at
Min2go
Presenter:
Tal
Segalov
COO
&
Co-‐Founder
at
Min2go
3. Agenda
• Why
do
Marketers
need
Science?
• What
is
Predic2ve
Marke2ng
and
what
can
it
do
for
me?
• How
to
build
a
Predic2ve
Marke2ng
machine
• What
do
you
get
from
Predic2ve
Marke2ng?
4. Netflix is commissioning original content because it
knows what people want before they do. “There are 33
million different versions of Netflix”
• Goofy
Comedies
• Cri2cally-‐acclaimed
Movies
• Because
you
watched
Curious
George
• Foreign
Movies
• …
6. Predic2ve
Marke2ng
The
ability
to
discover,
target,
and
engage
the
customers
and
prospects
who
are
the
most
likely
to
buy
based
on
your
current
customer
aNributes
including
their
digital
ac2vity.
7. A
Major
Problem
For
Marketers
You
are
constantly
running
campaigns,
but
not
seeing
desired
sales
results
at
end
of
the
pipeline.
8. What
If
You
Could…
• Discover
the
profile
of
your
ideal
customers?
• Target
the
prospects
who
are
most
likely
to
become
buyers?
• Respond
to
the
best
opportuni2es
faster?
• Shorten
your
sales
cycles?
• Spend
less
on
lead
gen
campaigns?
10. Predic2ve
Marke2ng
Name
Title
Email
MAP
SaaS?
Webinar?
Score
Tal
Segalov
COO
tal@min2go.com
Marketo
Yes
Yes
98
Mickey
Mouse
None
of
your
business
jdoe@gmail.com
0
John
Smith
VP
Marke2ng
js@qualiware.com
None
No
Yes
23
Bob
A
Director
Demand
Gen
boba@zen2st.net
Hubspot
Yes
No
78
Bill
Silver
HR
Specialist
bills@apitech.com
Eloqua
No
Yes
34
11. Predic2ve
Marke2ng
Name
Title
Email
MAP
SaaS?
Webinar?
Score
Tal
Segalov
COO
tal@min2go.com
Marketo
Yes
Yes
98
Mickey
Mouse
None
of
your
business
jdoe@gmail.com
0
John
Smith
VP
Marke2ng
js@qualiware.com
None
No
Yes
23
Bob
A
Director
Demand
Gen
boba@zen2st.net
Hubspot
Yes
No
78
Bill
Silver
HR
Specialist
bills@apitech.com
Eloqua
No
Yes
34
12. Predic2ve
Marke2ng
Name
Title
Email
MAP
SaaS?
Webinar?
Score
Tal
Segalov
COO
tal@min2go.com
Marketo
Yes
Yes
98
Mickey
Mouse
None
of
your
business
jdoe@gmail.com
0
John
Smith
VP
Marke2ng
js@qualiware.com
None
No
Yes
23
Bob
A
Director
Demand
Gen
boba@zen2st.net
Hubspot
Yes
No
78
Bill
Silver
HR
Specialist
bills@apitech.com
Eloqua
No
Yes
34
13. Pareto
–
Your
Hero
Vilfredo
Pareto,
Italian
Economist
observed
that
80%
of
the
land
in
Italy
was
owned
by
20%
of
the
families…and
the
80/20
rule
was
born.
17. Ingredients
• Fast sparse data store
• Company & people MI data
• Input identification & matching to DB
• Jobtitle analysis and clustering
• Input validation (field validity)
• Machine learning engine
18. MI Data
• Rich, accurate company & people data
– Coverage vs. Accuracy dilemma
– Granular MIs – specific attributes stronger
• Measuring the data quality:
– Amount of data – number of “True” in DB
– Data accuracy
– Measuring data usage in actual models
19. Example of An Indicator - Salesforce User
Hiring
Org-‐chart
Website
Scripts
3rd
Party
Sites
• Salesforce
2tle?
• Descrip2on
seman2c
processing
• Salesforce
admin
in
company?
• Lead
forms
CRM
connec2on
• Eco-‐system
products
(Marketo,
Pardot,
…)
• Dreamforce
men2on
• Case
studies
• PR,
news
‘Salesforce
User’
Never
Run
a
Bad
Campaign
Again!
20. Matching Algorithm
• Key element is matching input to DB data
• Identify input fields and map to internal
fields
• Match companies and people to MI DB:
– Use email, company, name, etc.
– Measure match rates on valid input data
– Measure match accuracy
21. Jobtitle Classification
• Jobtitles are one of the strongest
indicators
• Classify titles by semantic methods:
– Interpret jobtitle keywords
– Standardization of titles (e.g. dir.,
Director are the same)
– Distance metric between titles
22. Input Validation
• Validations on all available input fields
• Validations used for:
– Cleansing customer DB
– Modeling input
• Provide a “quality” metric for each lead
– Valid input (name, domain, email, etc.)
– Contact details validity
23. Machine Learning Model
• Modeling based on:
– 100’s of “positive” leads
– 10,000’s of “houselist” leads
– 1000’s of attributes (MIs)
• Strong feature selection
• Pareto – 80%-20% performance goal
24. Mixing It Together
Sparse
Data
Store
MI
Data
Houselist
Input
Connector
Matching
to
DB
iden22es
Input
Valida2on
Machine
learning
model
Predic2on
&
Enrichment
Job2tle
Classifier
Data
Partners’
Leads
New
leads
in
DB
27. The
Marke2ng
Machine:
Knobs
and
Levers
More
fuel
for
sales
Shortcut
through
my
funnel
28. The
Marke2ng
Machine:
Knobs
and
Levers
More
fuel
for
sales
Shortcut
through
my
funnel
Cross
sell
/
re-‐sort
my
lists
29. The
Marke2ng
Machine:
Knobs
and
Levers
More
fuel
for
sales
Shortcut
through
my
funnel
Cross
sell
/
re-‐sort
my
lists
Cleanse
my
list
("brush
teeth")
34. How
to
Measure
Success?
1. Qualifica2on:
– Build
model
based
on
Q4/2013
data
• Posi2ves
–
Opportuni2es
/
Closed
won
• Houselist
–
Marke2ng
list
– Run
predic2on
on
Q1/2014
data
• Predict
on
Q1/14
Marke2ng
list
– Test
results:
• What
percentage
of
posi2ves
are
in
top
scoring
20%?
• How
many
leads
were
appended
with
relevant
info?
• What
is
the
amount
of
bad
leads
cleaned
from
the
list?
35. Predic2ve
Model
–
Results
0%
5%
10%
15%
20%
25%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
recall
accuracy
%
Leads
%
of
Posi'ves
Accuracy
To
Sales
Nurture
Tracks
Bad
Fit
37. How
to
Measure
Success?
2. On-‐going
Usage:
– Enriched
data
based
campaigns
• 2-‐4x
liu
in
response
rates
– MQL
genera2on
rate
–
top
scores
vs.
the
rest
• Typical
–
3-‐5x
liu
– Bullet
train
through
the
funnel
-‐
reduce
lead
decay
• Typical
–
15-‐20%
decay
per
quarter
• Shortening
the
funnel
by
a
quarter
wins
back
this
decay
38. With
Predic2ve
Marke2ng
And
Min2go,
DocuSign
Discovered…
• 23.8%
engagement
on
acquired
targets
10X
improvement
• More
than
35
sales
opportuni2es/live
deals
• Genera2ng
over
$1M
TCV
• Higher
%
of
opps
leading
to
Closed
Won
• Faster
39. Conclusions
• Pareto
is
your
Hero!
– 20%
of
your
efforts
generate
80%
of
results
• Elements
of
successful
predic2ve
marke2ng:
– Good
data
collec2on
– Appending
data
to
your
DB
– Priori2zing
efforts
based
on
scores
– Quick
valida2on
and
qualifica2on
– Apply
across
all
Marke2ng
ac2vi2es