You are on page 1of 47

Making Advanced Analytics Simpler:

Challenges, Opportunities, and Value


Fern Halper
TDWI Director of Research, Advanced Analytics
July 23, 2015
@fhalper

Sponsor

Speakers

Fern Halper

John K. Thompson

Research Director,
Advanced Analytics
TDWI

General Manager,
Advanced Analytics, Dell

Advanced Analytics

Advanced analytics provides


algorithms for complex analysis of
either structured or unstructured data.

CHANGE
EXPECTATIONS

Agenda

Changing expectations
Skills needed
Common pitfalls
Best practices for getting started

Changing landscape for analytics

New Users
More users

Ease of use

New
platforms

New data
Changing
expectations

More and disparate data

(source: TDWI 2014)

More advanced
analytics
techniques used

(source: TDWI 2014)

More platforms
tools and
techniques

(source: TDWI 2014)

10

Predictive analytics process


Problem Identification
Framing problem
Identifying data elements

Data Access

Model Management

Platforms

Evaluation/Monitoring
Actual Management

Model deployment

Data preparation

Sharing
Scoring
Operationalizing

Cleansing
Transformation
Exploration

**

**

Model building
Exploration
Collaboration
Validation

**

11

New Users are Emerging


Statistician/Modeler

Moving towards critical


thinker with
knowledge of the
business- e.g. a
business analyst

More users too

(source: TDWI 2014)

13

Democratizing BI
To extend the deployment of BI and analytics
tools to more users in the organization

14

Democratization
What percent of your organization's employees is using
BI and/or analytics tools on any platform?
Don't know

<10%

10-30%

30-50%

31%

50-70%

70-100%
0%

Series1

5%
70-100%
15%

10%
50-70%
16%

15%
30-50%
13%

20%
10-30%
24%

25%
<10%
24%

30%
Don't know
8%

15

Consumability
Able to be used
More accessible results

16

Advanced Analytics Consumability Trends


Ease of Use
UI, Automation
PA as part of BI
package
Collaboration
Platforms

Operationalizing

Model scoring
Embedding
Real time
Platforms

17

Another way to look at it

2. Utilizing
Results

1. Model
Building
18

Is this a good thing?

19

Skills Needed (1)

3.
Domain
Expertise

1. Critical
Thinking

2. Data
Sense

Framing the problem


20

1. Critical Thinking
Ability to formulate a question
Comfortable creatively thinking in numbers
and attributes
Interpretation skills
Inference
Above all: Questioning

21

2. Domain Expertise
Helps in:
formulating good questions
understanding objectives
assessing the model and taking action on it

Understanding relevant data


Dealing with data outliers, missing data, etc.

22

3. Understanding data
Target vs. explanatory variables
Derived variables
Lots of new data types
Documents, graph, location
May require parsing, geocoding

23

Skills Needed (2)

4. Tools
5.
Techniques

6.
Storytelling

Explain/Defend

24

4. Understanding the tools!

25

5. Understanding the techniques


A basic understanding is necessary
Decision trees
Clustering
Regression

26

6. Storytelling
Dont start with the techniques
Begin with the business problem and the
outcome.

(source: vitualspeechcoach.com)

27

Common pitfalls
Underestimating training needs
On tools
On methods/interpretation
On thought process

Data management
Governance
Not thinking through cultural issues

28

Best practices
Build your skills even incrementally
Make sure there are process controls in place
before deploying models
Mentors office hours
CoE or even working groups
Collaboration
Model management

29

Poll Question
Is democratizing analytics a good idea?
Yes, it is best if everyone can build and use models
No, it is too risky to have people who arent trained in analytics
using easy to use tools
Dont know

30

Making Analytics Simpler: Challenges,


Opportunities, and Value
John K Thompson GM Advanced Analytics
Twitter: @johnkthompson60
31

Three Forces Redefining Analytics

Collective Intelligence

Native Distributed Analytics


Redefining the Economics
of Analytics
32

Leveraging the analytics skills & abilities of the


global community is Collective Intelligence

33

The idea is simple, collective intelligence allows


for an exchange of ideas, skills, models & more.

$
Idea &
Information
Exchange

People with good ideas


Business with a need

34

Collective Intelligence (CI) the global community.

35

CI & Statistica management, security, governance.


Chicago

Sao Paolo

Source

Singapore

36

CRAN
CRAN
CRAN
AML
Algo
Aperv
EM
Experfy

Model Type Version

Btree
Btree
Btree
NN
LGR
Ensemble
NN
CART

v1.0
v1.1
v1.2
v10
v5.0
V1.0
V2.0
V3.0

Native Distributed Analytics - v1.0

Statistica
Statistica Big Data Analytics
Neural Net..
Export Model as:
1. Java
2. PMML
3. C
4. C ++
5. SQL

37

Native Distributed Analytics v2.0

JVM

Statistica
Statistica Big Data Analytics

Boomi

Date/Time
Trans type
Velocity
Trigger

Neural Net..
Export Model as:
1. Java
2. PMML
3. C
4. C ++
5. SQL

JVM

JVM

Private
Cloud

38

Native Distributed Analytics v3.0

JVM

Statistica Model Building Environment

SMBE

Statistica
Statistica Big Data Analytics

Boomi

Date/Time
Trans type
Velocity
Trigger

Neural Net..
Export Model as:
1. Java
2. PMML
3. C
4. C ++
5. SQL

JVM

JVM

Private
Cloud

39

Redefining the Economics of Analytics

Dell set out on one of the most


ambitious migration projects since
the company was founded.

40

Redefining the Economics of Analytics

41

~300 users migrated


~70% savings in annual renewal fees
300+ projects across multiple business units
~6months

Dells pragmatic approach helps customers get


started today.
Start with what
you have
Start small

Use the devices and


data you already have.
Build on your current
technology
investments.
Grow based on realworld success.

42

Architect for analytics

Put security first

Plan for analytics-driven


action.

Secure from the data


center to the farthest
Dell endpoint and along
the networks and
clouds in between.

Build on your terms


with modular,
architecture-agnostic
solutions.
Harness the power of
advanced analytics.
Prepare to scale quickly
from pilot to
production.

Protect data wherever it


goes.

Secure for privacy and


compliance.

Dell Analytics Portfolio

43

Key Takeaways
Data Scientists are scarce, leverage yours and
everyone else in the world you can.

Bring analytics to the data, anywhere in the world,


at anytime

An open analytics platform will enable this


operating model and keep you ahead of the curve
and competition.

44

45

QUESTIONS?

46

Contact Information
If you have further questions or comments:
Fern Halper, TDWI
fhalper@tdwi.org
John K. Thompson, Dell
john.k.thompson@software.dell.com

47

You might also like