Dr. Lorien Pratt, Quantellia Chief Scientist, challenges you to advance your career by becoming a Decision Engineer. This emerging profession is a natural extension of business intelligence, and Dr. Pratt presents research to show that decision engineers are desperately needed. Learn how to design decisions, and some best practices as you help your organization and clients to gain that maximum value from data, "big data", databases, your expertise, and more.
2. Why Won’t Managers Use My Data?
Or: An Invitation to Become a
Decision Engineer
Dr. Lorien Pratt, Chief Scientist, Quantellia
Mark Zangari, CEO, Quantellia
3. About Me
• Based in Denver
• Former college professor
• Research focus: applied analytics/neural networks
• Wrote Learning to Learn and a lot of articles
• Ran market analyst team with Frost and Sullivan
• Co-founded Quantellia in 2008
• Chief Scientist
• US Government spending
• Community Justice Advisors analysis / Liberia
4. Agenda
1. Decision Engineering: Research showing
the importance of this need
2. Research results for what’s needed to fill
this need
3. How to do it: key steps
5. Global research study:
Q: What is the biggest problem that
technology should be solving, that
it is not?
6. Global research study:
Q: What is the biggest problem that
technology should be solving, that
it is not?
A: Decision making
7. Where all this great
data could be used
Where the
data is
actually used
8. Strong Demand for Better Use of Data
"Better use of our data and analytics could produce substantially
more value (cost savings and/or revenue growth) than it does
today"
Strongly Disagree
Disagree
Neutral
Agree
Strongly Agree
0.0% 10.0% 20.0% 30.0% 40.0% 50.0% 60.0% 70.0%
9. Ineffective Navigation Structure the Norm
"We have an effective business navigation structure in
place, where we make decisions, monitor their outcomes, then
adjust decisions as needed to achieve our business goals"
Strongly disagree
Disagree
Neutral
Agree
Strongly agree
0% 5% 10% 15% 20% 25% 30% 35%
10. Market Research
Environment Pharmaceuticals Financial Services
2% 2% 2%
Human Resources Nonprofit
2% 3%
Manufacturing
3%
Defense
6%
Public Health
7%
Media
10%
Telecommunications Information
52% Technology
11%
Source: Quantellia (2008) Number of samples = 61
11. Decision Making
How carefully do organizations
make decisions today?
We have a
All decisions are formal
made in an ad methodology
hoc manner and we generally
25% follow it
14% Approximately 86% of
organizations do not
consistently follow a
formal methodology for
ensuring sound decisions.
We have a
We follow an formal
informal "rule of methodology
thumb" but we do not
methodology adhere to it very
32% closely or
consistently
29%
Source: Quantellia (2008). N = 28
12. So why don’t managers use my
data?
Because their most essential needs
aren’t met
14. Decision making problems involve many business factors:
especially communication, collaboration, and visualization
What is Difficult in Your Organization About
Making Decisions?
Source: Quantellia (2008) N= 61
15. Decision makers have many needs that are not met by current
decision support systems
Qualitative plus quantitative data together
Need to represent intangibles
Organize information / Help with overload
Iterative Methodology
Social / Value Network Visibility
Templates / pre-canned models and/or data
Need for decision maker to tweak models themselves
User Friendly
Multiple bottom lines / objective functions
High powered quantitative engine
Handle uncertainty, e.g. by visualizing confidence levels
Model Building Wizard
Integrate with Excel
KPI Identification / Dashboard
Sensitivity Analysis
Common Methodology for Visualization
Include domain expertise
Mine Unstructured Data Sources
0%
5%
What features would be most valuable in 10%
software that supports decision making? 15%
Source: Quantellia (2008) N = 61
16. Systematic Decision Making Problems
• “We focus on only one measure, when there are really multiple
objectives.”
• “We make decisions that assume a predictable unchanging future.”
• “Our focus is on short-term goals,
ignoring long-term ones.”
• “We are unableReduce Time We Spend on long
to reason about Reduced Knowledge of our
cause-and-effect chains.”
Customer Care Telephone Calls Customers
• “We ignore intangibles like morale, reputation, trust, and brand. Brand
Cost
• “We plan for only a single future scenario Costs radically different
Lower Customer Care
when Unhappier Customers
courses of action may be appropriate, depending on how the future
unfolds.”
Revenue
“I can barely plan for next
quarter, how can I think about the
Community
future, too?”
Improved Contribution Margin
Service
“Five years from now, the market
for our product will have grown by
30%”
Worse Contribution Greater Customer Churn
Margin
26. Q: So how can I get my data more
widely used?
A: Realize that a decision (like
software) can be engineered, and
apply engineering principles to its
creation and management
27. Analogies from History
What have we done in the past
when the complexity of a
problem eventually exceeded
our ability to manage it?
Example: Construction.
• Small structures require little planning, commit
few resources, and have relatively few
consequences if they fail.
• As we try to build larger structures, we need
more is needed.
• There is a ceiling beyond which the complexity
becomes too great.
• An engineering discipline provides the
organizational and communications tools that
enable much larger structures to be reliably
erected.
29. To overcome the complexity ceiling, we
need to create a structured paradigm for
decision making…
We need Decision Engineering.
30. Previous times we’ve introduced
visual engineering approaches
Software Manufacturing Decision Making
Increasing visualization / interactivity over time
31. “[It is essential] to visualize not just the data used to
support decisions, but also the decisions
themselves. [This is an] essential need in both the
commercial and nonprofit worlds.”
-Lynn Langit, Developer Evangelist at Microsoft and author of the book Smart
Business Intelligence Solutions with SQL Server 2008
Quantellia: Winner of the 2009
Microsoft Windows 7 Innovation
Award
32. "In an age of global complexity, the time for making
decisions is ever-shrinking, and the cost of bad choices
too great to tolerate. Quantellia created a tool for
making the right decisions in this environment.”
-Guy Pfeffermann, former Chief Economist of the International Finance
Corporation (World Bank); Founder and CEO of the Global Business School
Network (www.gbsnonline.org).
33. “Telecommunications companies, along with other businesses challenged
by the rapid pace of a global environment, recognize the competitive
value of applying Business Intelligence and analytic tools to the vast
stores of data they generate. Visual, actionable decision engineering
solutions are the next evolutionary step in BI, to help get at what decision
makers need and how they think, rather than on what data managers can
provide.”
- Susan McNeice, Vice President - Software
Research, Yankee Group (www.yankeegroup.com)).
34. “Anyone facing complex decisions with many participants and
stakeholders, mounds of data, and limited resources to address
the decision-making process, should look closer at visualization
tools … Visualized decision support—decision engineering—is
fast becoming a key part of effective business management.”
-Karl Whitelock, Director Strategy – OSS/BSS, Stratecast, a Division of Frost
and Sullivan (www.frost.com ).
38. Key Elements of a Decision Model External Factors: impact the
outcome but over which we
Decision Data
have no control Examples:
• Competitor price
Levers External • Market demand
f Factors Goals: targets against
Decision Predictive
analytics outcomes. Example: 5%
margin growth in 2 years.
Levers
Decision levers: Factors over
which we have control. f Analytics
f Outcome
Examples: Analytics #1
• Price of a product
• Features of a product
f
• Investment in sales
• Investment in marketing
Intermediate Values
Outcome
• Investment in OSS f
#2
Dependencies: how one part of
the model depends upon f Analytics f
another, through cause-and-
effect or other flows. Intermediate Values: Facts Outcome
Examples: and values that are
How does MTTR respond to investment in
CSR training?
calculated along the way to #3
determining outcomes
How does brand respond to sales staff
expertise level? Examples: sales
Note: these can be determined through volume, mean time to Outcomes: Measures of success
traditional analytics, staff expertise, or respond, sales expertise Examples: Margin, Brand, Share
industry benchmarks
level, fallout rate Price
44. Apply best practices of the
engineering lifecycle
Quality Assurance
Objectives
Security
Planning
Phase Specification
Design
Alignment
Implementation Execution &
Phase Monitoring
Change
Management
45. Beware the Whack-a-Mole
“When I lower costs in one part of
my business, it ends up creating
bigger problems in another.”
46. My decision is only as
good as the data that
supports it
47. My decision is only as
good as the data that
supports it
Not
48. Good Decisions from
Imperfect Data
How:
Since only 10% of the data impacts 90% of
the decision, problems with the 90% matter
much less. Know which is which
Use sampling / statistical to extract excellent
analytics from messy data
Use human expertise when data is
imperfect
50. Follow the decision value chain /
connect the dots
Customer Changes to
Improvement Improvement demand curve: More revenue for
experience sell same product the same cost
to a KPI to brand
investment at a higher price
Keep asking why
52. Decision vs. Operational
Engineering Monitoring
• Like automobile design • Like monitoring a working
• Key competency: being able to vehicle
understand how the system will • Key competency: detecting
work problems accurately and quickly
• Key competency: using • Key competency: diagnosis
judgment where data is missing
53.
54.
55.
56. Data Is a key element, because
Situational Data + Decisions + Time =
Outcomes
57. Decision Engineering is the Next
Generation of Business Intelligence
Wanted: Decision Engineers.
Decision
An invitation: change the world.
(or, just do the next cool thing) Engineeri
Predictive Analytics
ng
Reporting/Business Intelligence
Data Management
58. THANK YOU.
Lorien.pratt@quantellia.com
303 589 7476
@LorienPratt
Please fill out the evaluation and turn it in to this session’s
host.
#GMSQL
Editor's Notes
What kinds of question need this kind of forward view? Here are some examples.
…
…
…
These are just a few of the kinds of decisions that need better support.
Let’s return to this gap. How do we fill it?
Microsoft gave us its innovation award in 2009 to reflect its recognition of the importance of our work
This demonstration shows how decision models are built. The environment has the ease-of-use of PowerPoint or Excel, but allows users to easily create a simulation of complex decisions.
Before we turn to the demonstrations, an important point. Decision modeling allows us to make highly confident decisions with incomplete or messy data. This is important because data management often presents a barrier for projects: it is expensive, risky, and time-consuming. So we must be strategic about the data we use. Decision modeling provides a way to identify the most important data for a problem.
Before we turn to the demonstrations, an important point. Decision modeling allows us to make highly confident decisions with incomplete or messy data. This is important because data management often presents a barrier for projects: it is expensive, risky, and time-consuming. So we must be strategic about the data we use. Decision modeling provides a way to identify the most important data for a problem.
Before we turn to the demonstrations, an important point. Decision modeling allows us to make highly confident decisions with incomplete or messy data. This is important because data management often presents a barrier for projects: it is expensive, risky, and time-consuming. So we must be strategic about the data we use. Decision modeling provides a way to identify the most important data for a problem.
Decision makers need two things: understanding of the present/past and a view towards the future. This is what decision engineering does.