Transcript BI de-Mystified

Business Intelligence
De-Mystified
Ben Bor
NZ Ministry of Health
1
Ben Bor
 Over 20 years in IT, most of it in Information Management
 Oracle specialist since version 5
 Involved in Business Intelligence for over 10 years
 Consulted the world’s largest corporations
 Presents regularly on Information Management
 Was annual Guest Lecturer at Sussex University
2
Session Objectives
 Understand the need for Business Intelligence and its
role in the enterprise information strategy
 Understand the role of the various Business Intelligence
technologies and tools
 Understand the BI components and the importance of
Data Quality
3
Contents
 Business Intelligence (BI) – Definition and Examples
 Data Warehousing (DW) – Definition and Architecture
 BI Challenges
 The BI Promise
 OLAP
 Data Mining
 Dashboards
 Alerts
4
Business Intelligence Ingredients
 Data Warehousing
 Data Marts
 OLAP
 Data Mining
 Data Quality
And others
5
Business Intelligence – Definition
‘Business Intelligence is the art of
gaining business advantage
from data’
 Who are my best and worst customers?
 What parameters affect my sales?
 What advantages does my business offer customers?
 Analyse my products by any parameter.
6
Some BI Success Stories
How much am I
spending?
What do I know about Joe
Bloggs?
7
Integrated view
of Customers & Suppliers
Business Intelligence Quadrants
Who Needs Business Intelligence? (Gartner Group)
Volume
of
Information
The Captive Customer
BI utilized by limited
numbers of experts to
reduce costs of
delivering services to
large numbers of
customers. No
competitive threats exist.
Global 2000
BI critical to understand
complexity of business,
leverage customer and
supplier relationships
and grasp and exploit
new opportunities.
The “Candy Store”
BI capabilities are of
limited utility. Decisions
made based on personal
management
observations of
customer trends and
markets.
The “e” Startup
Extreme need to
understand competition,
market and customer
trends. BI is pervasive as
a competitive weapon.
Business Pace
8
Data Warehousing – Definition 1
Accepted definition:
‘Subject Oriented, Integrated, Non-volatile, and Time
Variant Collection of Data in Support of
Management’s Decisions’.
Bill Inmon
‘Building a Data Warehouse’,
2nd edition, wiley 1996.
9
Data Warehousing – Definition 2
My definition:
‘A Data Warehouse is the enterprise single point of
access to its data’
10
Data Mart – Definition
A Data Mart is a project that uses Data Warehousing
techniques, but covers only a selected part of the
enterprise data
 Examples:
 Accounting Data Mart
 Sales Data Mart
11
Data Warehousing - How
a set of technologies:
Data Analysis
Access Different Data Sources
Data Cleansing & Normalising (ETL)
Data Storage
Presentation
12
Data Warehouse Architecture
Access Tools
Layer
Web-based
Reports
Semantic
Layer
Core Data
Store (CDS)
Views over CDS
Stand-alone
(legacy) schemas
Dashboards
Tool-specific
Business Model
(i.e. BO universes)
Metadata
Views
Data
Exploitation
Services (DES)
OLAP
IC-maintained
Data Marts
(Physical)
Federation
User-maintained
Data Marts
Joining Structures
External
Databases
Reference data
Including person,
company,
address,
household,
etc’
Joined-up data
ODS
Staging
(Data acquisition)
Persistent Staging
(with history)
Non-Persistent Staging
Data Quality Profiling
Extracted
13 files
Oracle
Streams
Log
Mining
XML
Risk
Engine
Inmon and Kimball
Inmon
Centralised
Kimball
Distributed
Fundamental Model
Entity-Relation
Dimensional
ODS
Separate
System
Part of DW
Data Mart
Derived &
Aggregated
Independent &
Atomic
Basic
Architecture
14
Dimensional Modelling
A design method that is
 Not entity-relationship modelling
 Not normalised
 Easily understood by users
 More efficient for BI
15
Dimensional Modelling Example
Consultants submit timesheets, showing the number
of hours, the rate and their expenses per project
per day.
Managers (AD) are responsible for projects and
consultants.
16
Entity Relationship Design
Consultant
Rates
Manager
Expense
Type
Client
Time
Sheet
Project
Expense
17
Project
Task
Project
Code
Project
Staff
Dimensional Modelling Example
(Star Schema)
Consultant
Dimension
Project
Dimension
Activity
Facts
Time
Dimension
Client
Dimension
Manager
Dimension
18
Dimensional Modelling Example
(Snowflake Schema)
Sector
Division
Team
Client
Consultant
Project
Activity
Facts
Time
Client
Manager
Industry
Quarter
Division
19
Month
Year
OLAP
On-Line Analytical Processing
 A data presentation method that allows the users to
interactively change the criteria, the level and the
contents
 Usually based on a multi-dimensional model
 Allows for drill-down, drill-up and drill-across
20
OLAP - Multi Dimensional Cube
Product Mgr. View
M
M
AA
RR
K
KE
ET
T
SALES
Regional Mgr. View
TIME
Financial Mgr. View
21
Ad Hoc View
OLAP Methods
 ROLAP
 Relational OLAP
(Business objects)
 MOLAP
 Multi-dimensional OLAP
(Hyperion)
 HOLAP
 Hybrid OLAP
(Cognos)
22
OLAP DEMO
23
Data Mining - Definition
A method for automatically deducing knowledge from data:
 Patterns, clusters, rules, decision trees etc’
24
Classification Tree
age < 35
salary >
80000
sex = M
bal < 6300
IM for Data Classification Results
Interpreting Tree Induction Results
age >= 35
sex = F
marital = S
2 classes who purchase luxury cars
age < 35
age < 35
sex = M
sex = F
salary > 80000
marital = S
bal > 6300
IBM Software Solutions
25
Executive Dashboards
26
The BI Assimilation Lifecycle
Alerts
C
o
m
p
l
e
x
i
t
y
OLAP
Bulk
Reports
Exceptio
n
Reports
Time
n months
27
n months
n months
n months
Balanced Scorecard
A method of organisational performance measurement.
Performance of an organisation from four perspectives:
 Customer perspective (how do customers see us?)
 Internal capabilities perspective (what must we excel at?)
 Innovation and learning perspective
(can we continue to improve
and create value?)
 Financial perspective (how do our owners/shareholders see us?)
28
Information Quality
Information is Data in context.
Information Quality is Data Quality in context with meaning.
The ability to trust the information
 Data Quality
 Reliable and repeatable testing
 Metadata
29
Main Challenges in Business Intelligence
Business intelligence projects are
 User-oriented
 Large
 “Complex simplicity”
 Continuously evolving
 Require deep technical and business knowledge
 Stretch all limits:
 Time, storage capacity, CPU, machine
communications, human communication, human
perception, and teamwork.
 Data Quality
30
What’s Happening in the BI world?
 BI is becoming a norm
 Mature, off-the-shelf tools
 Combine structured and non-structured data
 A small number of main players
 New uses (Data Webhouse)
 Real-time
 Hosted BI
 Open Source BI
31
Summary - Business Intelligence
Data Warehouse
Information Quality
Data Mart
Dimensional
Modelling
Star Schema
ETT/ETL
Snowflake Schema
Balanced Scorecard
ODS
OLAP
Data Webhouse
32
Data Mining
33
Thank you !
I can be contacted at [email protected]
34