https://www.bigdataspain.org/2016/program/fri-from-data-to-ai-with-the-machine-learning-canvas.html https://www.youtube.com/watch?v=Qpo5XzlUrwE&index=27&list=PL6O3g23-p8Tr5eqnIIPdBD_8eE5JBDBik&t=12s

2. From Data to AI with the
MACHINE LEARNING
CANVAS
@louisdorard #bigdataspain

3. “A breakthrough in machine
learning would be worth ten
Microsofts”
–Bill Gates

4. “In the next 20 years, machine
learning will have more impact
than mobile has.”
–Vinod Khosla

5. 4

6. Skiing down the Gartner hype cycle
with Waldo & the Machine Learning Canvas

8. W
H
O
S
Y
O
U
R
P
A
P
I
RED
ICT
IVE
PPLI
CATI
ON
ROGR
AMMI
NG
NTER
FACE
?

10. What is ML?

13. Bedrooms Bathrooms Surface (foot²) Year built Type Price ($)
3 1 860 1950 house 565,000
3 1 1012 1951 house
2 1.5 968 1976 townhouse 447,000
4 1315 1950 house 648,000
3 2 1599 1964 house
3 2 987 1951 townhouse 790,000
1 1 530 2007 condo 122,000
4 2 1574 1964 house 835,000
4 2001 house 855,000
3 2.5 1472 2005 house
4 3.5 1714 2005 townhouse
2 2 1113 1999 condo
1 769 1999 condo 315,000

14. Bedrooms Bathrooms Surface (foot²) Year built Type Price ($)
3 1 860 1950 house 565,000
3 1 1012 1951 house
2 1.5 968 1976 townhouse 447,000
4 1315 1950 house 648,000
3 2 1599 1964 house
3 2 987 1951 townhouse 790,000
1 1 530 2007 condo 122,000
4 2 1574 1964 house 835,000
4 2001 house 855,000
3 2.5 1472 2005 house
4 3.5 1714 2005 townhouse
2 2 1113 1999 condo
1 769 1999 condo 315,000
last column = output (by convention)

17. 16
Some use cases
• Real-estate
• Spam filtering
• City bikes
• Reduce churn
• Anticipate demand
property price
email spam indicator
location, context #bikes
customer churn indicator
product, store, date #sales
Zillow
Gmail
BikePredict
ChurnSpotter
Blue Yonder
RULES

18. @louisdorard

19. 1. Descriptive analysis
2. Predictive analysis
3. Prescriptive analysis
4. Automated decisions
18
(Big?) Data analysis
“big data”, reporting
& old-school BI…
now we’re talking!

20. Decisions from predictions

22. 1. Show churn rate against time
2. Predict which customers will churn next
3. Suggest what to do about each customer
(e.g. propose to switch plan, send promotional offer, etc.)
21
Churn analysis

23. • Who: SaaS company selling monthly subscription
• Question asked:“Is this customer going to leave within 1
month?”
• Input: customer
• Output: no-churn or churn
• Data collection: history up until 1 month ago
22
Churn prediction

24. Assume we know who’s going to churn. What do we do?
• Contact them (in which order?)
• Switch to different plan
• Give special offer
• No action?
23
Churn prediction prevention

25. “3. Suggest what to do about each customer”
→ prioritised list of actions, based on…
• Customer representation + context (e.g. competition)
• Churn prediction
• Uncertainty in predictions
• Revenue brought by customer
• Constraints on frequency of solicitations
24
Churn prevention

26. • Taking action for each TP (and FP) has a cost
• For each TP we“gain”: (success rate of action) *
(revenue /cust. /month)
• Imagine…
• perfect predictions
• revenue /cust. /month = 10€
• success rate of action = 20%
• cost of action = 2€
• What is the ROI?
25
Churn prevention ROI

27. Machine Learning Canvas

28. 27
The Canvas Concept

29. 28
The Machine Learning Canvas
The Machine Learning Canvas (v0.4) Designed for: Designed by: Date: Iteration: .
Decisions
How are predictions used to
make decisions that provide
the proposed value to the end­user?
ML task
Input, output to predict,
type of problem.
Value
Propositions
What are we trying to do for the
end­user(s) of the predictive system?
What objectives are we serving?
Data Sources
Which raw data sources can
we use (internal and
external)?
Collecting Data
How do we get new data to
learn from (inputs and
outputs)?
Making
Predictions
When do we make predictions on new
inputs? How long do we have to
featurize a new input and make a
prediction?
Offline
Evaluation
Methods and metrics to evaluate the
system before deployment.
Features
Input representations
extracted from raw data
sources.
Building Models
When do we create/update
models with new training
data? How long do we have to
featurize training inputs and create a
model?
Live Evaluation and
Monitoring
Methods and metrics to evaluate the
system after deployment, and to
quantify value creation.
machinelearningcanvas.com by Louis Dorard, Ph.D. Licensed under a Creative Commons Attribution­ShareAlike 4.0 International License.

30. • Not an adaptation of the Business Model Canvas
• Describe the Learning part of a predictive system / an intelligent
application:
• What data are we learning from?
• How are we using predictions powered by that learning?
• How are we making sure that the whole thing“works”through
time?
29
The Machine Learning Canvas

31. –Ingolf Mollat, Principal Consultant at Blue Yonder
“The Machine Learning Canvas is providing our
clients real business value by supplying the first
critical entry point for their implementation
of predictive applications.”

32. 31
Cross Industry Standard Process for Data Mining
By Kenneth Jensen -
Own work, CC BY-SA 3.0
ML Canvas

33. The Machine Learning Canvas (v0.4) Designed for: Designed by: Date: Iteration: .
Decisions
How are predictions used to
make decisions that provide
the proposed value to the end­user?
ML task
Input, output to predict,
type of problem.
Value
Propositions
What are we trying to do for the
end­user(s) of the predictive system?
What objectives are we serving?
Data Sources
Which raw data sources can
we use (internal and
external)?
Collecting Data
How do we get new data to
learn from (inputs and
outputs)?
Making
Predictions
When do we make predictions on new
inputs? How long do we have to
featurize a new input and make a
prediction?
Offline
Evaluation
Methods and metrics to evaluate the
system before deployment.
Features
Input representations
extracted from raw data
sources.
Building Models
When do we create/update
models with new training
data? How long do we have to
featurize training inputs and create a
model?
Live Evaluation and
Monitoring
Methods and metrics to evaluate the
system after deployment, and to
quantify value creation.
machinelearningcanvas.com by Louis Dorard, Ph.D. Licensed under a Creative Commons Attribution­ShareAlike 4.0 International License.
background
specifics
LEARNPREDICT
EVALUATE
GOAL
(what, why, who)
how how
how well

34. The Machine Learning Canvas (v0.4) Designed for: Designed by: Date: Iteration: .
Decisions
How are predictions used to
make decisions that provide
the proposed value to the end­user?
ML task
Input, output to predict,
type of problem.
Value
Propositions
What are we trying to do for the
end­user(s) of the predictive system?
What objectives are we serving?
Data Sources
Which raw data sources can
we use (internal and
external)?
Collecting Data
How do we get new data to
learn from (inputs and
outputs)?
Making
Predictions
When do we make predictions on new
inputs? How long do we have to
featurize a new input and make a
prediction?
Offline
Evaluation
Methods and metrics to evaluate the
system before deployment.
Features
Input representations
extracted from raw data
sources.
Building Models
When do we create/update
models with new training
data? How long do we have to
featurize training inputs and create a
model?
Live Evaluation and
Monitoring
Methods and metrics to evaluate the
system after deployment, and to
quantify value creation.
machinelearningcanvas.com by Louis Dorard, Ph.D. Licensed under a Creative Commons Attribution­ShareAlike 4.0 International License.
background
specifics

35. The Machine Learning Canvas (v0.4) Designed for: Designed by: Date: Iteration: .
Decisions
How are predictions used to
make decisions that provide
the proposed value to the end­user?
ML task
Input, output to predict,
type of problem.
Value
Propositions
What are we trying to do for the
end­user(s) of the predictive system?
What objectives are we serving?
Data Sources
Which raw data sources can
we use (internal and
external)?
Collecting Data
How do we get new data to
learn from (inputs and
outputs)?
Making
Predictions
When do we make predictions on new
inputs? How long do we have to
featurize a new input and make a
prediction?
Offline
Evaluation
Methods and metrics to evaluate the
system before deployment.
Features
Input representations
extracted from raw data
sources.
Building Models
When do we create/update
models with new training
data? How long do we have to
featurize training inputs and create a
model?
Live Evaluation and
Monitoring
Methods and metrics to evaluate the
system after deployment, and to
quantify value creation.
machinelearningcanvas.com by Louis Dorard, Ph.D. Licensed under a Creative Commons Attribution­ShareAlike 4.0 International License.
background
specifics

36. The Machine Learning Canvas (v0.4) Designed for: Designed by: Date: Iteration: .
Decisions
How are predictions used to
make decisions that provide
the proposed value to the end­user?
ML task
Input, output to predict,
type of problem.
Value
Propositions
What are we trying to do for the
end­user(s) of the predictive system?
What objectives are we serving?
Data Sources
Which raw data sources can
we use (internal and
external)?
Collecting Data
How do we get new data to
learn from (inputs and
outputs)?
Making
Predictions
When do we make predictions on new
inputs? How long do we have to
featurize a new input and make a
prediction?
Offline
Evaluation
Methods and metrics to evaluate the
system before deployment.
Features
Input representations
extracted from raw data
sources.
Building Models
When do we create/update
models with new training
data? How long do we have to
featurize training inputs and create a
model?
Live Evaluation and
Monitoring
Methods and metrics to evaluate the
system after deployment, and to
quantify value creation.
machinelearningcanvas.com by Louis Dorard, Ph.D. Licensed under a Creative Commons Attribution­ShareAlike 4.0 International License.
background
specifics
LEARNPREDICT
EVALUATE
GOAL
(what, why, who)
Domain
Integration
Predictive
Engine

37. The Machine Learning Canvas (v0.4) Designed for: Designed by: Date: Iteration: .
Decisions
How are predictions used to
make decisions that provide
the proposed value to the end­user?
ML task
Input, output to predict,
type of problem.
Value
Propositions
What are we trying to do for the
end­user(s) of the predictive system?
What objectives are we serving?
Data Sources
Which raw data sources can
we use (internal and
external)?
Collecting Data
How do we get new data to
learn from (inputs and
outputs)?
Making
Predictions
When do we make predictions on new
inputs? How long do we have to
featurize a new input and make a
prediction?
Offline
Evaluation
Methods and metrics to evaluate the
system before deployment.
Features
Input representations
extracted from raw data
sources.
Building Models
When do we create/update
models with new training
data? How long do we have to
featurize training inputs and create a
model?
Live Evaluation and
Monitoring
Methods and metrics to evaluate the
system after deployment, and to
quantify value creation.
machinelearningcanvas.com by Louis Dorard, Ph.D. Licensed under a Creative Commons Attribution­ShareAlike 4.0 International License.
background
specifics
#1 #2#3#4 #5
#6
#7#8 #9#10

38. The Machine Learning Canvas (v0.4) Designed for: Designed by: Date: Iteration: .
Decisions
How are predictions used to
make decisions that provide
the proposed value to the end­user?
ML task
Input, output to predict,
type of problem.
Value
Propositions
What are we trying to do for the
end­user(s) of the predictive system?
What objectives are we serving?
Data Sources
Which raw data sources can
we use (internal and
external)?
Collecting Data
How do we get new data to
learn from (inputs and
outputs)?
Making
Predictions
When do we make predictions on new
inputs? How long do we have to
featurize a new input and make a
prediction?
Offline
Evaluation
Methods and metrics to evaluate the
system before deployment.
Features
Input representations
extracted from raw data
sources.
Building Models
When do we create/update
models with new training
data? How long do we have to
featurize training inputs and create a
model?
Live Evaluation and
Monitoring
Methods and metrics to evaluate the
system after deployment, and to
quantify value creation.
machinelearningcanvas.com by Louis Dorard, Ph.D. Licensed under a Creative Commons Attribution­ShareAlike 4.0 International License.
On 1st day of every month:
• Filter out ‘no-churn’
• Sort remaining by
descending (churn prob.) x
(monthly revenue) and
show prediction path for
each
• Solicit customers
Predict answer to “Is this
customer going to churn in
the coming month?”
• Input: customer
• Output: ‘churn’ or ‘no-
churn’ class (‘churn’ is the
Positive class)
• Binary Classification
Context:
• Company sells SaaS with
monthly subscription
• End-user of predictive
system is CRM team
We want to help them…
• Identify important clients
who may churn, so
appropriate action can be
taken
• Reduce churn rate among
high-revenue customers
• Improve success rate of
retention efforts by
understanding why
customers may churn
• CRM tool
• Payments database
• Website analytics
• Customer support
• Emailing to customers
Every month, we see which of
last month’s customers
churned or not, by looking
through the payments
database.
Associated inputs are
customer “snapshots” taken
last month.
Every month we (re-)featurize
all current customers and
make predictions for them.
We do this overnight.
Basic customer info at time t
(age, city, etc.)
Events between (t - 1 month)
and t:
• Usage of product: # times
logged in, functionalities
used, etc.
• Cust. support interactions
• Other contextual, e.g.
devices used
Every month we create a new
model from the previous
month’s customers.
• Monitor churn rate
• Monitor (#non-churn among solicited) / #solicitations
Customer retention Louis Dorard Sept. 2016 1

39. The Machine Learning Canvas (v0.4) Designed for: Designed by: Date: Iteration: .
Decisions
How are predictions used to
make decisions that provide
the proposed value to the end­user?
ML task
Input, output to predict,
type of problem.
Value
Propositions
What are we trying to do for the
end­user(s) of the predictive system?
What objectives are we serving?
Data Sources
Which raw data sources can
we use (internal and
external)?
Collecting Data
How do we get new data to
learn from (inputs and
outputs)?
Making
Predictions
When do we make predictions on new
inputs? How long do we have to
featurize a new input and make a
prediction?
Offline
Evaluation
Methods and metrics to evaluate the
system before deployment.
Features
Input representations
extracted from raw data
sources.
Building Models
When do we create/update
models with new training
data? How long do we have to
featurize training inputs and create a
model?
Live Evaluation and
Monitoring
Methods and metrics to evaluate the
system after deployment, and to
quantify value creation.
machinelearningcanvas.com by Louis Dorard, Ph.D. Licensed under a Creative Commons Attribution­ShareAlike 4.0 International License.
On 1st day of every month:
• Filter out ‘no-churn’
• Sort remaining by
descending (churn prob.) x
(monthly revenue) and
show prediction path for
each
• Solicit customers
Predict answer to “Is this
customer going to churn in
the coming month?”
• Input: customer
• Output: ‘churn’ or ‘no-
churn’ class (‘churn’ is the
Positive class)
• Binary Classification
Context:
• Company sells SaaS with
monthly subscription
• End-user of predictive
system is CRM team
We want to help them…
• Identify important clients
who may churn, so
appropriate action can be
taken
• Reduce churn rate among
high-revenue customers
• Improve success rate of
retention efforts by
understanding why
customers may churn
• CRM tool
• Payments database
• Website analytics
• Customer support
• Emailing to customers
Every month, we see which of
last month’s customers
churned or not, by looking
through the payments
database.
Associated inputs are
customer “snapshots” taken
last month.
Every month we (re-)featurize
all current customers and
make predictions for them.
We do this overnight.
Basic customer info at time t
(age, city, etc.)
Events between (t - 1 month)
and t:
• Usage of product: # times
logged in, functionalities
used, etc.
• Cust. support interactions
• Other contextual, e.g.
devices used
Every month we create a new
model from the previous
month’s customers.
• Monitor churn rate
• Monitor (#non-churn among solicited) / #solicitations
Customer retention Louis Dorard Sept. 2016 1

40. • We predict customer would churn but they don’t…
• Great! Prevention works!
• Sh*t! Data inconsistent…
• (Store which actions were taken?)

42. The Machine Learning Canvas (v0.4) Designed for: Designed by: Date: Iteration: .
Decisions
How are predictions used to
make decisions that provide
the proposed value to the end­user?
ML task
Input, output to predict,
type of problem.
Value
Propositions
What are we trying to do for the
end­user(s) of the predictive system?
What objectives are we serving?
Data Sources
Which raw data sources can
we use (internal and
external)?
Collecting Data
How do we get new data to
learn from (inputs and
outputs)?
Making
Predictions
When do we make predictions on new
inputs? How long do we have to
featurize a new input and make a
prediction?
Offline
Evaluation
Methods and metrics to evaluate the
system before deployment.
Features
Input representations
extracted from raw data
sources.
Building Models
When do we create/update
models with new training
data? How long do we have to
featurize training inputs and create a
model?
Live Evaluation and
Monitoring
Methods and metrics to evaluate the
system after deployment, and to
quantify value creation.
machinelearningcanvas.com by Louis Dorard, Ph.D. Licensed under a Creative Commons Attribution­ShareAlike 4.0 International License.
On 1st day of every month:
• Randomly filter out 50% of
customers (hold-out set)
• Filter out ‘no-churn’
• Sort remaining by
descending (churn prob.) x
(monthly revenue) and
show prediction path for
each
• Solicit customers
Predict answer to “Is this
customer going to churn in
the coming month?”
• Input: customer
• Output: ‘churn’ or ‘no-
churn’ class (‘churn’ is the
Positive class)
• Binary Classification
Context:
• Company sells SaaS with
monthly subscription
• End-user of predictive
system is CRM team
We want to help them…
• Identify important clients
who may churn, so
appropriate action can be
taken
• Reduce churn rate among
high-revenue customers
• Improve success rate of
retention efforts by
understanding why
customers may churn
• CRM tool
• Payments database
• Website analytics
• Customer support
• Emailing to customers
Every month, we see which of
last month’s customers
churned or not, by looking
through the payments
database.
Associated inputs are
customer “snapshots” taken
last month.
Every month we (re-)featurize
all current customers and
make predictions for them.
We do this overnight.
Basic customer info at time t
(age, city, etc.)
Events between (t - 1 month)
and t:
• Usage of product: # times
logged in, functionalities
used, etc.
• Cust. support interactions
• Other contextual, e.g.
devices used
• Monitor churn rate
• Monitor (#non-churn among solicited) / #solicitations
Customer retention Louis Dorard Sept. 2016 1
Every month we create a new
model from the previous
month’s customers.

43. The Machine Learning Canvas (v0.4) Designed for: Designed by: Date: Iteration: .
Decisions
How are predictions used to
make decisions that provide
the proposed value to the end­user?
ML task
Input, output to predict,
type of problem.
Value
Propositions
What are we trying to do for the
end­user(s) of the predictive system?
What objectives are we serving?
Data Sources
Which raw data sources can
we use (internal and
external)?
Collecting Data
How do we get new data to
learn from (inputs and
outputs)?
Making
Predictions
When do we make predictions on new
inputs? How long do we have to
featurize a new input and make a
prediction?
Offline
Evaluation
Methods and metrics to evaluate the
system before deployment.
Features
Input representations
extracted from raw data
sources.
Building Models
When do we create/update
models with new training
data? How long do we have to
featurize training inputs and create a
model?
Live Evaluation and
Monitoring
Methods and metrics to evaluate the
system after deployment, and to
quantify value creation.
machinelearningcanvas.com by Louis Dorard, Ph.D. Licensed under a Creative Commons Attribution­ShareAlike 4.0 International License.
On 1st day of every month:
• Randomly filter out 50% of
customers (hold-out set)
• Filter out ‘no-churn’
• Sort remaining by
descending (churn prob.) x
(monthly revenue) and
show prediction path for
each
• Solicit customers
Predict answer to “Is this
customer going to churn in
the coming month?”
• Input: customer
• Output: ‘churn’ or ‘no-
churn’ class (‘churn’ is the
Positive class)
• Binary Classification
Context:
• Company sells SaaS with
monthly subscription
• End-user of predictive
system is CRM team
We want to help them…
• Identify important clients
who may churn, so
appropriate action can be
taken
• Reduce churn rate among
high-revenue customers
• Improve success rate of
retention efforts by
understanding why
customers may churn
• CRM tool
• Payments database
• Website analytics
• Customer support
• Emailing to customers
Every month, we see which of
last month’s customers
churned or not, by looking
through the payments
database.
Associated inputs are
customer “snapshots” taken
last month.
Every month we (re-)featurize
all current customers and
make predictions for them.
We do this overnight.
Basic customer info at time t
(age, city, etc.)
Events between (t - 1 month)
and t:
• Usage of product: # times
logged in, functionalities
used, etc.
• Cust. support interactions
• Other contextual, e.g.
devices used
Every month we create a new
model from the previous
month’s hold-out set (or the
whole set, when initializing
this system).
We do this overnight (along
with making predictions).
• Monitor churn rate
• Monitor (#non-churn among solicited) / #solicitations
Customer retention Louis Dorard Sept. 2016 1

44. The Machine Learning Canvas (v0.4) Designed for: Designed by: Date: Iteration: .
Decisions
How are predictions used to
make decisions that provide
the proposed value to the end­user?
ML task
Input, output to predict,
type of problem.
Value
Propositions
What are we trying to do for the
end­user(s) of the predictive system?
What objectives are we serving?
Data Sources
Which raw data sources can
we use (internal and
external)?
Collecting Data
How do we get new data to
learn from (inputs and
outputs)?
Making
Predictions
When do we make predictions on new
inputs? How long do we have to
featurize a new input and make a
prediction?
Offline
Evaluation
Methods and metrics to evaluate the
system before deployment.
Features
Input representations
extracted from raw data
sources.
Building Models
When do we create/update
models with new training
data? How long do we have to
featurize training inputs and create a
model?
Live Evaluation and
Monitoring
Methods and metrics to evaluate the
system after deployment, and to
quantify value creation.
machinelearningcanvas.com by Louis Dorard, Ph.D. Licensed under a Creative Commons Attribution­ShareAlike 4.0 International License.
On 1st day of every month:
• Randomly filter out 50% of
customers (hold-out set)
• Filter out ‘no-churn’
• Sort remaining by
descending (churn prob.) x
(monthly revenue) and
show prediction path for
each
• Solicit customers
Predict answer to “Is this
customer going to churn in
the coming month?”
• Input: customer
• Output: ‘churn’ or ‘no-
churn’ class (‘churn’ is the
Positive class)
• Binary Classification
Context:
• Company sells SaaS with
monthly subscription
• End-user of predictive
system is CRM team
We want to help them…
• Identify important clients
who may churn, so
appropriate action can be
taken
• Reduce churn rate among
high-revenue customers
• Improve success rate of
retention efforts by
understanding why
customers may churn
• CRM tool
• Payments database
• Website analytics
• Customer support
• Emailing to customers
Every month, we see which of
last month’s customers
churned or not, by looking
through the payments
database.
Associated inputs are
customer “snapshots” taken
last month.
Every month we (re-)featurize
all current customers and
make predictions for them.
We do this overnight.
Basic customer info at time t
(age, city, etc.)
Events between (t - 1 month)
and t:
• Usage of product: # times
logged in, functionalities
used, etc.
• Cust. support interactions
• Other contextual, e.g.
devices used
Every month we create a new
model from the previous
month’s hold-out set (or the
whole set, when initializing
this system).
We do this overnight (along
with making predictions).
• Accuracy of last month’s predictions on hold-out set
• Compare churn rate & lost revenue between last month’s
hold-out set and remaining set
• Monitor (#non-churn among solicited) / #solicitations
• Monitor ROI (based on diff. in lost revenue & cost of
solicitations)
Customer retention Louis Dorard Sept. 2016 1

45. The Machine Learning Canvas (v0.4) Designed for: Designed by: Date: Iteration: .
Decisions
How are predictions used to
make decisions that provide
the proposed value to the end­user?
ML task
Input, output to predict,
type of problem.
Value
Propositions
What are we trying to do for the
end­user(s) of the predictive system?
What objectives are we serving?
Data Sources
Which raw data sources can
we use (internal and
external)?
Collecting Data
How do we get new data to
learn from (inputs and
outputs)?
Making
Predictions
When do we make predictions on new
inputs? How long do we have to
featurize a new input and make a
prediction?
Offline
Evaluation
Methods and metrics to evaluate the
system before deployment.
Features
Input representations
extracted from raw data
sources.
Building Models
When do we create/update
models with new training
data? How long do we have to
featurize training inputs and create a
model?
Live Evaluation and
Monitoring
Methods and metrics to evaluate the
system after deployment, and to
quantify value creation.
machinelearningcanvas.com by Louis Dorard, Ph.D. Licensed under a Creative Commons Attribution­ShareAlike 4.0 International License.
On 1st day of every month:
• Randomly filter out 50% of
customers (hold-out set)
• Filter out ‘no-churn’
• Sort remaining by
descending (churn prob.) x
(monthly revenue) and
show prediction path for
each
• Solicit customers
Before soliciting customers:
• Evaluate new model’s
accuracy on pre-defined
customer profiles
• Simulate decisions taken
on last month’s customers
(using model learnt from
customers 2 months ago).
Compute ROI w. different #
customers to solicit &
hypotheses on retention
success rate (is it >0?)
Predict answer to “Is this
customer going to churn in
the coming month?”
• Input: customer
• Output: ‘churn’ or ‘no-
churn’ class (‘churn’ is the
Positive class)
• Binary Classification
Context:
• Company sells SaaS with
monthly subscription
• End-user of predictive
system is CRM team
We want to help them…
• Identify important clients
who may churn, so
appropriate action can be
taken
• Reduce churn rate among
high-revenue customers
• Improve success rate of
retention efforts by
understanding why
customers may churn
• CRM tool
• Payments database
• Website analytics
• Customer support
• Emailing to customers
Every month, we see which of
last month’s customers
churned or not, by looking
through the payments
database.
Associated inputs are
customer “snapshots” taken
last month.
Every month we (re-)featurize
all current customers and
make predictions for them.
We do this overnight (along
with building the model that
powers these predictions and
evaluating it).
Basic customer info at time t
(age, city, etc.)
Events between (t - 1 month)
and t:
• Usage of product: # times
logged in, functionalities
used, etc.
• Cust. support interactions
• Other contextual, e.g.
devices used
Every month we create a new
model from the previous
month’s hold-out set (or the
whole set, when initializing
this system).
We do this overnight (along
with offline evaluation and
making predictions).
• Accuracy of last month’s predictions on hold-out set
• Compare churn rate & lost revenue between last month’s
hold-out set and remaining set
• Monitor (#non-churn among solicited) / #solicitations
• Monitor ROI (based on diff. in lost revenue & cost of
solicitations)
Customer retention Louis Dorard Sept. 2016 1

46. The Machine Learning Canvas (v0.4) Designed for: Designed by: Date: Iteration: .
Decisions
How are predictions used to
make decisions that provide
the proposed value to the end­user?
ML task
Input, output to predict,
type of problem.
Value
Propositions
What are we trying to do for the
end­user(s) of the predictive system?
What objectives are we serving?
Data Sources
Which raw data sources can
we use (internal and
external)?
Collecting Data
How do we get new data to
learn from (inputs and
outputs)?
Making
Predictions
When do we make predictions on new
inputs? How long do we have to
featurize a new input and make a
prediction?
Offline
Evaluation
Methods and metrics to evaluate the
system before deployment.
Features
Input representations
extracted from raw data
sources.
Building Models
When do we create/update
models with new training
data? How long do we have to
featurize training inputs and create a
model?
Live Evaluation and
Monitoring
Methods and metrics to evaluate the
system after deployment, and to
quantify value creation.
machinelearningcanvas.com by Louis Dorard, Ph.D. Licensed under a Creative Commons Attribution­ShareAlike 4.0 International License.
Before soliciting customers:
• Evaluate new model’s
accuracy on pre-defined
customer profiles
• Simulate decisions taken
on last month’s customers
(using model learnt from
customers 2 months ago).
Compute ROI w. different #
customers to solicit &
hypotheses on retention
success rate (is it >0?)
Predict answer to “Is this
customer going to churn in
the coming month?”
• Input: customer
• Output: ‘churn’ or ‘no-
churn’ class (‘churn’ is the
Positive class)
• Binary Classification
Context:
• Company sells SaaS with
monthly subscription
• End-user of predictive
system is CRM team
We want to help them…
• Identify important clients
who may churn, so
appropriate action can be
taken
• Reduce churn rate among
high-revenue customers
• Improve success rate of
retention efforts by
understanding why
customers may churn
• CRM tool
• Payments database
• Website analytics
• Customer support
• Emailing to customers
Every month, we see which of
last month’s customers
churned or not, by looking
through the payments
database.
Associated inputs are
customer “snapshots” taken
last month.
Every month we (re-)featurize
all current customers and
make predictions for them.
We do this overnight (along
with building the model that
powers these predictions and
evaluating it).
Basic customer info at time t
(age, city, etc.)
Events between (t - 1 month)
and t:
• Usage of product: # times
logged in, functionalities
used, etc.
• Cust. support interactions
• Other contextual, e.g.
devices used
Every month we create a new
model from the previous
month’s hold-out set (or the
whole set, when initializing
this system).
We do this overnight (along
with offline evaluation and
making predictions).
• Accuracy of last month’s predictions on hold-out set
• Compare churn rate & lost revenue between last month’s
hold-out set and remaining set
• Monitor (#non-churn among solicited) / #solicitations
• Monitor ROI (based on diff. in lost revenue & cost of
solicitations)
Customer retention Louis Dorard Sept. 2016 1
On 1st day of every month:
• Randomly filter out 50% of
customers (hold-out set)
• Filter out ‘no-churn’
• Sort remaining by
descending (churn prob.) x
(monthly revenue) and
show prediction path for
each
• Solicit as many customers
as suggested by simulation

47. • Assist data scientists, software engineers, product and business
managers, in aligning their activities
• Make sure all efforts are directed at solving the right problem!
• Choose right algorithm / infrastructure / ML solution prior to
implementation
• Guide project management
• machinelearningcanvas.com
46
Why fill in ML canvas?

49. –Jeremy Howard
“Great predictive modeling is an important
part of the solution, but it no longer stands on its
own; as products become more sophisticated, it
disappears into the plumbing.”

50. Example: send promotional offers via email to SaaS customers, based on predicted churn
• Compute feature values for given input (a.k.a.“featurize”; involves merging data sources,
aggregating data…)
• Collect training data (inputs and outputs)
• Provide predictive model from given training set (i.e. learn)
• Provide prediction against model for given input (context)
• Provide optimal decision from given contextual data, predictions, uncertainties,
constraints, objectives, costs
• Apply given decision
49
Software components in real-world ML / AI

51. Software components for automated decisions:
• Compute feature values for given input (a.k.a.“featurize”; involves merging data sources,
aggregating data…)
• Collect training data (inputs and outputs)
• Provide predictive model from given training set (i.e. learn)
• Provide prediction against model for given input (context)
• Provide optimal decision from given contextual data, predictions, uncertainties,
constraints, objectives, costs
• Apply given decision
50
Application-specific component

52. Software components for automated decisions:
• Compute feature values for given input (a.k.a.“featurize”; involves merging data sources,
aggregating data…)
• Collect training data (inputs and outputs)
• Provide predictive model from given training set (i.e. learn)
• Provide prediction against model for given input (context)
• Provide optimal decision from given contextual data, predictions, uncertainties,
constraints, objectives, costs
• Apply given decision
51
Optimization / Operations Research component

53. Software components for automated decisions:
• Compute feature values for given input (a.k.a.“featurize”; involves merging data sources,
aggregating data…)
• Collect training data (inputs and outputs)
• Provide predictive model from given training set (i.e. learn)
• Provide prediction against model for given input (context)
• Provide optimal decision from given contextual data, predictions, uncertainties,
constraints, objectives, costs
• Apply given decision
52
Machine Learning components

54. Software components for automated decisions:
• Compute feature values for given input (a.k.a.“featurize”; involves merging data sources,
aggregating data…)
• Collect training data (inputs and outputs)
• Provide predictive model from given training set (i.e. learn)
• Provide prediction against model for given input (context)
• Provide optimal decision from given contextual data, predictions, uncertainties,
constraints, objectives, costs
• Apply given decision
53
Data Engineering components

55. 54

56. follow us: @papisdotio
WE’RE HIRING!