So often, we use machine learning for a limited set of use cases, in customer experience, marketing, advertising, and finance.
In this talk, we step back to understand how big data and machine learning can serve a new class of problems: those for which users need to know "what will be the impact of today's decision, tomorrow?"
2. Question: If technology could solve one
problem for you, that it doesn’t solve
today, what would it be?
3. Question: If technology could solve one
problem for you, that it doesn’t solve
today, what would it be?
Answer: Massive amounts of data,
machine learning, other advanced
technology, but it’s not getting used for the
most important decisions.
8. From… To…
Source: Tibco Jaspersoft
“Isn’t there a better way? I am making big decisions:
is there a way to structure all this data, and to use
machine learning, to get the most value out of it?”
11. Data
Instrumented
Code / Sensors
Data
Management
Analytics
Gap between computer and human
bridged by Data Visualization
Presentation
Demarcation between
automated (computer-centric) and
manual (human-centric)
information processing
12. Gap between computer and human
bridged by Data Visualization
System
Analysis
Decision
Data
Instrumented
Code / Sensors
Data
Management
Analytics
Presentation
Demarcation between
automated (computer-centric) and
manual (human-centric)
information processing
13. What is a the relationship
between data, systems, and
decisions?
Decision Lever
Outcome
Outcome
Outcome
Externals
External Factors
Decision Lever
Intermediates
Intermediates
External Factors
Intermediates
Goal
Goal
Each connector represents a dependency.
Goal
15. Fa“FcACtTiIvVISisM”m
Intervention
Impact
analysis
How can we best
deploy security to
ensure a fair
election?
How can we
maximize the value
of aid to reduce
childhood
mortality?
16.
17.
18. “Poor decision making can cost–
and, in an industry that invests as
much as telecoms, the total cost can
be very large indeed.”
“Our research reveals that, in
the past decade, the average
long-term return on investment
(ROI) has been just 6%—three
percentage points less than the
cost of the capital itself.”
19. “What is critical in today’s
complex world is the ability
to see over the horizon and
around corners to
understand the impact of
today’s decisions on all of
the desired outcomes.”
20.
21. “We are seeing increasing demand for a
C-level executive who understands how
to use data and machine learning to
support business decisions.
This may end up as a role for the CIO,
Chief Data Officer (CDO), or a new role
may emerge: the Chief Decision Officer:
who is in charge of using expertise and
evidence to support the company’s
most important business decisions”
Adam-Bryce, LLC (919) 638-0707
27. TRADITIONAL VIEW
What will
be the
outcome?
What
decisions
can we
make?
Data, Analytics, Big
Data, Reports,
Predictive Analytics,
Spreadsheets
28. DECISION INTELLIGENCE VIEW
What data, analytics,
reports, human
expertise, and other
assets are relevant?
What
outcomes
do we
need or
want to
reach ?
What
decisions
will get us
there?
29. “…our predictions may be more prone to failure in the era of Big
Data.
As there is an exponential increase in the amount of available information, there is likewise an
exponential increase in the number of hypotheses to investigate.
For instance, the U.S. government now publishes data on about 45,000 economic statistics. If you
want to test for relationships between all combinations of two pairs of these statistics–is there a
causal relationship between the bank prime loan rate and the unemployment rate in Alabama?–
that gives you literally one billion hypotheses to test.
But the number of meaningful relationships in the data–those
that speak to causality rather than correlation and testify to how
the world really works–is orders of magnitude smaller.”
—Nate Silver
Who correctly called the outcomes of the 2012 US Presidential election in all 50 states
30. Elements
Need:
1) A systems model (with systems dynamics, feedback loops, etc.)
2) Machine learning
3) Information from experts for when data is missing
4) Simulation
5) Optimization
6) Crystal clear visualization
7) An agency model to add to the information model
8) Interactivity
31. Decision Makers
What will be the impact of
today’s decision, tomorrow?
Machine Learning
Analytics
Data
32. A DECISION INFLUENCES A SYSTEMS MODEL,
WHICH RESULTS IN OUTCOMES
Decision Lever
Outcome
Outcome
Outcome
Externals
External Factors
Decision Lever
Intermediates
Intermediates
External Factors
Intermediates
Goal
Goal
Each connector represents a dependency.
Goal
33. Decision Makers
The CDO’s responsibility is to
fill this gap
Machine Learning
Analytics
Data
34.
35. Lack of
consistent
service
Net
Promoter
Score
Irrelevant
proactive
notifications
Invest in consistent
customer data
Pain of having to
deal with the call
center
Invest in self service
improvement via the
smartphone
How many
people use the
call center
Invest in improving
the call center
Churn
Levers
Outcomes
Externals
Loyalty
Revenue
Better proactive
resolution of issues
Customer
calling behavior
Customer needs
My competitor’s
NPS
Competitor
advertising
36. Cause Effect
Machine Learning rule, from historical data
that captures this link
37. Lack of
consistent
service
Net
Promoter
Score
Irrelevant
proactive
notifications
Invest in consistent
customer data
Pain of having to
deal with the call
center
Invest in self service
improvement via the
smartphone
How many
people use the
call center
Invest in improving
the call center
Churn
Levers
Outcomes
Externals
Loyalty
Revenue
Better proactive
resolution of issues
Customer
calling behavior
Customer needs
My competitor’s
NPS
Competitor
advertising
38. “If I make this decision today, how will it affect my outcomes
in the future?”
Decision Lever
Outcome
Outcome
Outcome
Externals
External Factors
Decision Lever
Intermediates
Intermediates
External Factors
Intermediates
Goal
Goal
Each connector represents a dependency.
Goal
40. Optimization
Decision Lever
Outcome
Outcome
Outcome
Externals
External Factors
Decision Lever
Intermediates
Intermediates
External Factors
Intermediates
Goal
Goal
Goal
41. How do we maximize profit? Should we
generate our own renewable energy? If so,
when is the best time to do so?
Tens of millions of dollars in savings potential
per year for a typical large enterprise
42. How do we best invest in the legal system
to set a developing country on a road to
avoid future conflict?
Should we invest in buildings and books
or motorcycles and paralegals?
Thousands of lives at stake
43. “Is my money better
spent on more
servers or more
iPads?”
45. “Where should I place network equipment to build the next
internet, at the lowest cost and maximum value to customers?”
Typical – 72 homes per Node Optimal – 115 homes per Node
46. “Where should I place wifi hotspots in my town to provide the
best customer service and to maximize revenues?”
47. What to do next?
1. Start thinking about your
organization’s “big decisions”
2. Identify outcomes and goals
3. Identify levers
4. Identify externals
5. Build a decision model picture to
show how they connect
6. Identify data that can be learned from
to analyze cause-and-effect links.
7. Where data is missing, find human
expertise
8. Combine existing learned rules as
parts of the full decision model
9. Assign someone to be responsible
10. Learn about complex systems analysis
11. Learn about optimization and
simulation
Learn more on at http://www.youtube.com/quantellia
Unified resource: http://www.scoop.it/t/decison-intelligence
… or complex decisions affecting human lives and livelihoods
Two ways we use data:
1) fully automated (list use cases)
2) graphs and charts, but then we upload into our brains and make decisions
is limited. 2) requires that we effectively upload the graphs and charts into our brains to make decisions
Steve
Microsoft gave us its innovation award in 2009 to reflect its recognition of the importance of our work
Microsoft gave us its innovation award in 2009 to reflect its recognition of the importance of our work
Microsoft gave us its innovation award in 2009 to reflect its recognition of the importance of our work
Microsoft gave us its innovation award in 2009 to reflect its recognition of the importance of our work