alison wagonfeld: marketing meets science
TRANSCRIPT
Marketing Meets Science
Alison WagonfeldVP Marketing, Google Cloud
greenlight.com
California Research Center
My journey
Maps Chrome Android
Marketing demands constant innovation
The marketer’s role is changing
Traditional campaign execution
Hyper-relevant, Real-time engagement
Data capture
Retroactive performance analysis
Data-backed customer insights
Performance-led strategy
Thoughtful, targeted, scalable
Proactive
FROM TO
Big Data Analytics Machine Learning
Gives computers the ability to learn without being explicitly programmed. Machine Learning
Machine Learning Opportunities in Marketing
IDENTIFYCustomers
TARGET & APPROACH
Customers
GROWCustomers
OPTIMIZE SPEND
1. Acquisition
2. Customer growth
Examples of Machine Learning in marketing at Google Cloud
Which campaigns are driving the most paid G Suite users?
Business Challenge #1
Example 1:
Customer Acquisition for G-Suite
FREE TRIALSCAMPAIGN CONVERSIONS
ANALYSIS
Traditional marketing model requires 45+ days to optimize
45 DAYS
OPTIMIZE
START END PAID SEATS
Channels
Example 1:
Customer Acquisition for G-Suite
COURSE CORRECT
2-day marketing optimization model
PREDICT
2 DAYS
PAID SEATS
UPDATE MODEL
FREE TRIALSCAMPAIGN CONVERSIONS
ANALYSIS
Channels
Example 2:
Customer acquisition for G-Suite with ML
3
Looking under the hood
Data is collected.
Data is analyzed & visualized.
New free trial accounts get scored.
1 2 4 5 6
ML model is created and trained.
Historical data
Paid media campaigns begin.
ML makes automated paid media campaign adjustments.
Real-time data
What is the optimal “next” product for each Google Cloud Platform (GCP) customer?
Business Challenge #2
Channels
Example 2:
Customer account growth for GCPTraditional product correlation model
A
B
C
B
D
X
Recommendation engine
ACCOUNT BEHAVIOR
Example 2:
Customer growth for GCP
RECOMMENDATION ENGINE
A
LEARNING MODEL
A X
COMMUNICATIONCHANNELS
MOST RELEVANT
3
Example 2:
Customer Growth for GCP with MLLooking under the hood
Data is collected.
Data is analyzed & visualized.
Recom-mendation campaigns begin across channels.
1 2 4 5 6
ML model is created and trained.
Historical data
ML scores accounts, suggesting products and offers.
Consumer takes action, informing the ML model.
Real-time data
ML Model suggests optimal outreach for each account
In-console message
Direct mailEmail
How do I try this at home?
Where to start
1. Team: marketer + engineer + data analyst
2. Products:a. Data & Analysis: BigQuery,
Data Studiob. Machine Learning: Cloud ML
Engine, TensorFlow
1. Data integrity
2. Experimental mindset
3. Select business problem to address
Setting your team up for success
Common pitfalls
1. Incomplete Data
2. Privacy & Legal
3. Bias
4. Resources
5. Creative
Is Machine Learning the “Holy Grail” for Marketing?
Alas, no such thing.
We still need the “Marketing” to go with the “Science.”It’s all about the combo.
Q&A