BI 1 Data Warehousing
-
Upload
nidhi-kumar -
Category
Documents
-
view
218 -
download
0
description
Transcript of BI 1 Data Warehousing
Dhruv Nath
BI 1Data Warehousing for
CRM
BI 1Data Warehousing for
CRM
The CRM System
CRMSystem
CRMSystem
SharedSharedData BaseData Base
Marketing
Sales
Service
Management
Customer
Operational CRM
Analytical CRM
Reading
• The Nuts and Bolts of CRM– Ch 8 - 11
Case : Data Warehousing 1a
Data Warehouse
vs Operational Database (OLTP Database)
The CRM System
CRMSystem
CRMSystem
Marketing
Sales
Service
Management
CustomerOperational CRMAnalytical
CRM
SharedSharedData BaseData Base
The CRM System
Marketing
Sales
Service
Management
CustomerOperational CRM
SharedSharedData BaseData Base
DataDataWarehouseWarehouse
AnalyticalCRM
AnalyticalCRM
OperationalCRM
OperationalCRM
Analytical CRM
Real-Time Updates ?
No. Snapshots at pre-defined frequency
Granularity
• The Database has detailed data
• What about the Data Warehouse ?
• Grain Size : Fine vs Coarse / Large
• Pluses / Minuses ?– Speed– Disk space– Answerable queries
The Data Warehouse Could be Multi-Level
OperationalOperational(OLTP)(OLTP)
DatabaseDatabase
Warehouse with coarse grain size
Historical Data detailed
Tape
Warehouse with fine grain
size
Could be one or more levels
Users use the level of detail they need
Snapshot of data : Detailed
Exercise : OLTP Database vs. Data Warehouse….. list differences
• Current - Real Time• Current data• Updates reqd
– Volatile
• Full details• Low volume• Response time :
Fraction of a second
• Snapshots• Historical data• No updates
– Non-volatile (Why ?)
• Summaries• Huge volume• Several seconds ->
minutes / hours
When do we take snapshots ?
• Examples :– Airtel– Hindustan Lever Sales data– Citibank– BHEL Sales data– Indianoil Sales Data
Depends on the frequency / volume of transactions
Typically at night. Why ?
Doesn’t affect speed of the OLTP database
ICICIBank
Savings Account
Credit Card
Share broking
Private Equity
Does the Management need information across these product lines? Examples ?
Multiple Services / Products
Typical Management Queries
• Check the list of all credit card holders, and take out those with good credit history. These may be potential candidates for home / auto loans
• Of the customers who use our share broking services, which ones are good candidates for private equity ?
• If a customer drops one of our products, is he likely to drop all products ?
What do we need for these queries ?
The DW needs to pick up data from multiple sources : OLTP databases / Excel files / Manual files / Registers, etc. etc.
Disc : Examples from your organisations ?
The Data Warehouse Architecture
DataDataWarehouseWarehouse
Application Oriented
Subject Oriented
Problems with Multiple Sources of
Data ???
Source Data (OLTPSource Data (OLTPDatabases / Files….)Databases / Files….)
?
Process for DW Creation (Disc)
• Extract
• Transform– Clean– Standardise– Store these transformations for the future
(METADATA)
• Load
ETL
Data Mart - disc.• Subset of the Data Warehouse• Benefits ?
– Lower volume of data => Speed– Simpler than the DW
• Examples ?– Customer Data split into Data Marts
• by Region
• by Category of Customer (Corp, Govt, Retail)
• By Product…………..
• Are these Data Marts Exclusive ?– Maybe / maybe not
The Data Warehouse Architecture
Data MartsData Marts
DataDataWarehouseWarehouse
OLTPOLTPDatabasesDatabases
Application vs Subject Oriented
Why can’t we simply convert each Data Source into a Data Mart, instead of going through a Data
Warehouse ?
Where do Data Marts fit into this Architecture ?
Data Marts determine the queries you can ask + Speed + Complexity
Therefore Managers and Analysts need to carefully decide on Data Marts at Design time
The Data Warehouse Architecture
• When are Data Marts updated ?
• At the same time as the Data Warehouse
• Time Consuming– Therefore don’t have
unnecessary Data Marts
Data MartsData Marts
DataDataWarehouseWarehouse
OLTPOLTPDatabasesDatabases
Dhruv Nath
BI 1Data Warehousing for
CRM
BI 1Data Warehousing for
CRM