Busting Big Data Myths with an Analytics-First Strategy KIRK BORNE Principal Data Scientist, Booz Allen Hamilton Booz | Allen | Hamilton @KirkDBorne
Busting Big Data Myths
with an Analytics-First
Strategy
KIRK BORNE Principal Data Scientist, Booz Allen Hamilton
Booz | Allen | Hamilton @KirkDBorne
2
Find competitive
advantage for the
business with machine
learning and AI
Side-step the Big Data
hype bandwagon and
derive Big Value from
your data assets
Think Big, Start
Small, Learn Fast
with DataOps
Go for Analytics-First
by focusing on
purpose, products,
and outcomes
Adopt a Culture of
Experimentation
Acquire, nurture,
benefit from, and
retain key data
science talent
Machine Learning and AI are
big scary things
c
Data Science is a side project for data scientists
Data-first is the right strategic
posture for success
Three Responses Three Challenges Three Myths
Booz | Allen | Hamilton @KirkDBorne
Busting Big Data Myths – part 1:
Demystifying AI, Machine Learning, Data Science, and
DataOps
Data-informed , Analytics-driven
Innovation
3
Booz | Allen | Hamilton @KirkDBorne
Source for graphic: https://www.forbes.com/sites/chunkamui/2016/01/03/6-words/
“The distinction between
success and failure
in innovation efforts
boils down to six words:
Think Big,
Start Small,
Learn Fast.”
- Chunka Mui
innovation advisor
Are you ready for DataOps? … Agile Data Science and a Fail-fast, Learn-fast Culture of Experimentation!
The Learn Fast
culture of DataOps
helps you to avoid
an episode of
“Data Oops!”
4
Booz | Allen | Hamilton @KirkDBorne
DataOps – Agile Data Science
10
… Incremental, Iterative, Continuous, Agile
… Nurtures a Culture of Experimentation
… Builds The Learning Organization
… Focus on POVs (Proofs of Value), not POC (proof of concept)
… Think Big, Start Small = the MVP (Minimally Viable Product)
and the MLP (Minimally Lovable Product)
… Fail-fast Learn-fast!
DataOps — DevOps for Data Analytics
https://oreil.ly/2zZWRvk
5
Data Science: 4 Types of Discovery from Data! Which are you doing?
1)Class Discovery: Finding new classes of objects (population segments), events, and behaviors. This includes: learning the rules that constrain the class boundaries.
2)Correlation (Predictive and Prescriptive Power) Discovery: Finding patterns and dependencies, which reveal new governing principles or behavioral patterns (the “customer DNA”).
3)Novelty (Surprise!) Discovery:
Finding new, rare, one-in-a-million objects / events.
4)Association (or Link) Discovery: Finding unusual (“interesting”) co-occurring associations.
6
Booz | Allen | Hamilton @KirkDBorne
5 Levels of Analytics Maturity in Data-intensive Applications
1) Descriptive Analytics
– Hindsight (What happened?)
– Asks the required questions.
2) Diagnostic Analytics
– Oversight (Real-time / What is
happening? Why did it happen?)
3) Predictive Analytics
– Foresight (What will happen?)
4) Prescriptive Analytics
– Insight (How can we optimize what
happens?) (Follow the dots!)
5) Cognitive Analytics
– Right Sight (the 360 view; what is the right
action, right decision, right now, for this set
of data within this specific context.
– Moves beyond simply providing answers, to
generating new questions and hypotheses.
7
Metaphorical Use Case of Data Science, AI, Machine Learning, DataOps and Agile Analytics in a Data-Driven System
The Mars Rover : • intelligent data-gatherer
• mobile data mining agent
• autonomous decision system • A self-driving “enterprise”
8
Mars Rover:
9
Metaphorical Use Case of Data Science, AI, Machine Learning, DataOps and Agile Analytics in a Data-Driven System
Booz | Allen | Hamilton @KirkDBorne
Busting Big Data Myths – part 2:
Becoming the Data and Analytics Catalyst
Data-informed , Analytics-driven
Innovation
10
Booz | Allen | Hamilton @KirkDBorne
Source for graphics: https://bit.ly/2zF2MUY
The Role of Data and Analytics Catalysts :
Be the agent of change in your organization!
Culture is the key ingredient to analytics success.
11
The mature data science organization…
1) …democratizes all data and data access.
2) …uses Agile for everything and leverages DataOps.
3) ...leverages the crowd and works collaboratively (hackathons, etc.)
4) …follows rigorous scientific methodology (i.e., experimental, disciplined,…).
5) …attracts & retains diverse participants; grants them freedom to explore.
6) …relentlessly asks the right questions, and searches for the next one.
7) …celebrates a fast-fail collaborative culture.
8) …shows insights through illustrations and tells stories.
9) …builds proof of value, not proof of concepts.
10)…personifies data science as a way of doing things, not a thing to do.
How to Attract, Nurture, and Retain Key Talent
Booz | Allen | Hamilton @KirkDBorne
12
Booz | Allen | Hamilton @KirkDBorne
Busting Big Data Myths – part 3:
Taking “Data to Action” for Big Value through “Analytics by Design”
Data-informed , Analytics-driven
Innovation
13
Booz | Allen | Hamilton @KirkDBorne
Analytics By Design – (a) Organizational Posture
Analytics-first Posture: Focus on Business Outcomes (Products) –
this focus explicitly induces the corporate messaging, culture, and
strategy to be better aligned with what matters => Outcomes!
Deliver business value from the products of Data Science, AI, and
Machine Learning – products deliver ROI and Value from your
data assets.
Examples of products: enriched data sets, curated open data,
APIs, applications, models, cloud services, models, data science
notebooks, open source tools, …
Analytics-first is not the same as Data-first. (Data are the input.
Analytics are the output.)
14
Booz | Allen | Hamilton @KirkDBorne
#2: Identify Desired Results: outcomes, priorities, purpose,
strategic objectives
#3: Determine Acceptable Evidence (proofs): data, KPIs,
measurement instruments
#4: Plan and Design Activities: machine learning applications,
data experiences, data products, areas of AI and automation
#1: Adopt a Culture of Experimentation – “test or get fired!”
https://bit.ly/2JPFQIN
https://en.wikipedia.org/wiki/Understanding_by_Design
Analytics By Design – (b) Organizational Principles
Analytics By Design avoids the 2 biggest problems: (a) FTH due to
FOMO (Following The Hype due to Fear Of Missing Out); (b) Being
activity-oriented (i.e., focused on “busy work” instead of outcomes).
15
16
Find competitive
advantage for the
business with machine
learning and AI
Side-step the Big Data
hype bandwagon and
derive Big Value from
your data assets
Think Big, Start
Small, Learn Fast
with DataOps
Go for Analytics-First
by focusing on
purpose, products,
and outcomes
Adopt a Culture of
Experimentation
Acquire, nurture,
benefit from, and
retain key data
science talent
Machine Learning and AI are
big scary things
c
Data Science is a side project for data scientists
Data-first is the right strategic
posture for success
Three Responses Three Challenges Three Myths
Booz | Allen | Hamilton @KirkDBorne
Nurture and empower your analytics talent within a culture of
experimentation: A data-driven experimental orientation (which is
the essence of Data Science and DataOps) is an essential
“innovation best practice.”
The organizational cultural change (including democratized data
access) that is required to adopt data science as a way of
doing things (and not just a thing to do) is perhaps a greater
challenge than the technological challenges.
Demonstrating value and ROI (Return On Innovation) from small
implementations and POVs (Proofs of Value) will inspire the
cultural change needed for the larger implementations that will
come.
Take-away Messages
Image Credit: Qubole
DataOps
17
Booz | Allen | Hamilton @KirkDBorne
Thank you!
KIRK BORNE Principal Data Scientist Booz Allen Hamilton
@KirkDBorne https://bit.ly/2qbqa7l
18
Adopting a culture of
experimentation is good data
science, and adopting an
analytics-first big data
strategy is good business.