Where Does Location Intelligence Fit in An Enterprise Data MiningBI Strategy

1
Where Does Location Intelligence Fit in An
Enterprise Data Mining/BI Strategy?

• Tim Pletcher pletc1ta_at_cmich.edu

2
A Broad Definition of Business Intelligence

• CMU-RC uses the Data Warehousing Institutes
  definition of Business Intelligence (BI) to gain
  insight from data for the purpose of taking
  action.
• This definition encompasses the broad suite of
  business analytics predictive modeling, data or
  text mining, geographic information systems,
  statistical analysis, operations research,
  systems dynamics, simulation, and advanced data
  visualization.

3
Common Applications for BI






4
Value Creation
Time ?When Spatial ? Where
5
Reporting
6
LI Inspired Data for Business Intelligence

• Census/Postal Geography
• Street Networks
• Demographics
• Spatial Segmentation
• Aerial Photos and Land Use Data
• GPS RFID captured/fed updates
• Consumer Expenditure Data
• Retail transactions
• Market Potential Data
• Shipping volumes
• Utility usage
• Traffic Counts

Street and Cartographic Data
Aerial/Imagery Data
Census Geography and Data
Customer Data
Competitor Data
Store Location Data
7
Unique Spatial Techniques

• Market Area Boundaries
• Drive Times
• Desire Lines
• Market Penetration
• Site Selection
• Gravity Models
• ETL for spatial data (Soils volumes/zip to
  census)
• Spatial Queries
• E.g. based on Demographic or Household Data
• Spatial Statistics
• Networks and Process Maps

8
Advanced Visualization
9
Location Intelligence is Evolving with BI
Web Services Networks
Client/Server Systems
Desktop Tools Data
Enterprise Platform
Departmental
Projects
10
Embedded Solutions e.g. SAP Integration
Connectivity Layer
HTTP/SOAP
Tier 3
Presentation Layer
Integration Layer
Web Browser

• JSP/Java Servlets/JSP Tag Libraries
• BSP/BSP Extensions

• Java Connector
• .NET Connector
• GBC
• BC
• XML
• SOAP
• BAPI/RFC
• XI
• Other ..

Tier 2
Web Server
Business Layer
ABAP/J2EE
Application Server
Tier 1
Database Server
Persistence Layer
JDBC/Open SQL
SAP Web Application Server
11
Multiple Solutions That Span the Enterprise
Views
Products
Updates Transactions
Analysis
Mission Critical Applications
12
Enterprise Technology Adoption
Enterprise
Business Unit
Emerging Technology
Economies of Scale
BI
LI
Well Understood
Not Well Understood
13
Enter the BI Competency Center

• A BI Competency Center is a group chartered to
  advocate and bolster the adoption of BI in the
  enterprise.
• Some specific charters
• Generate awareness for executives and line
  managers about the competitive advantage and ROI
  
• Inter Silo-data sharing
• Establish standards and methodologies
• Raise the alarm about the need for data quality
• Ensures that quality analytics and applied

14
Models/Homes for a BI Competency Center

• Possible Structures or Organization Homes
• Project management offices
• Six Sigma Continuous Quality Improvement
• Repurposed Operations Research Teams
• Newly constructed teams at strategic level or in
  IT
• Key Team Characteristics
• Understands the business drivers
• Can work with a process and get results
• Ability to apply technology, but recognizes it is
  not about technology
• Quantitatively competent.. Including spatial
  analysis

15
One Example

• Scenario A large company wanted to understand
  their risk related to warranty on a product.
• Previous attempts using traditional analysis
  continued to miss the mark each quarter (by many
  millions of ).
• There was a physical driver for the defect
  (moisture, soil permeability, temperature, etc.)
• There was a people driver for the claim rate
  (once it started there was a claim fad)
• Result A robust forecast using neural networks
  to score the data and predict the amount of
  claims that would occur during the warranty
  period.

16
Model Results

• The company had three groups do modeling. All
  produced the bottom line result with fairly close
  estimates.
• Example
• XXX,XXX,XXX of future warranty expenses can
  expected to occur during the remaining warranty
  period for the product.
• This result has a 98 confidence interval within
  YYY,YYY,YYY and ZZZ,ZZZ,ZZZ

17
Predictive Modeling
18
Combining LI and BI
19
Actual Claims History
20
Predictions
21
Results
22
Contact Information

• THANK YOU!
• Timothy A Pletcher
• Director of Applied Research
• Central Michigan University Research Corporation
• Phone [phone]
• tim.pletcher_at_cmich.edu