Business Intelligence/ Decision Models Week 2 IT Infrastructure & Marketing Database Design and Implementation.

Business Intelligence/Decision Models

Week 2IT Infrastructure

& Marketing Database Design and Implementation

Outline Issues with Mkt Databases DBMS Database Design and Schemas Data Integrity and Hygiene Demo and Lab: Table redundancy and Queries

DB Marketing Problems Lack of a marketing strategy. Focus on promotions instead of relationships. Failure to have a 3600 picture of every customer. Failure to personalize your communications. Building a DB and sending e-mails in house. Getting the economics wrong. Failure to use tests and controls. Lack of a forceful leader. Bad DB architecture Corrupted data  

DB Environment

Traditional Environment:Silo Approach

Source: Laudon and Laudon 2012

Data Warehouse Technology

Marketing Datamart

Data Warehouse Architecture

Data Warehouse Architecture

Metadata

Database Management Systems (DBMS)

Flat Files Sequential

Fixed or variable length record

A B C D

Name Address Transactions1 2 3

A

DBMS with VSAM Index

QC

ON

ON

QC

ON

TNNENBIPEQCONMBSKABBC

Hierarchical IndexedDirect Access DBMS

Cust_id

Name Purchases Products

Top down

Indexed Direct Access DBMS

Key Record107 4110 6145 1167 2234 5267 3

Records1 145 ……….2 167 ……….3 267 ……….4 107 ……….5 234 ……….6 110 ……….

Reversed Hierarchical DBMS

Cust_id

Name Purchases ProductsPsyte CodeLifestyle

Bottom up/Top down

Reversed Hierarchical DBMSNAME PSYTE PURCHASES

Dubé 18 120Smith 34 130Bertrand 18 150White 56 200Harris 34 50Habib 18 300Jones 34 430

PSYTE NAMES

18 Dubé; Bertrand; Habib

34 Smith; Harris; Jones56 White

Relational Database

CUSTOMERS   ORDERS   PRODUCTS

Customer ID PK Order ID PK Product ID PK

Cust First Name Customer ID FK Product Name

Cust Last Name Product ID FK Product Description

Street Order Date  

City Order Amount

State  Zip

1

Relational DBMSMultiple Tables

Source: Laudon and Laudon 2012

Relational DBMSwith Query

Source: Laudon and Laudon 2012

Relational Design

An Unnormalized Relation For Order (flat file)

An unnormalized relation contains repeating groups. For example, there can be many parts and suppliers for each order. There is only a one-to-one correspondence between Order Number and Order Date.

Source: Laudon and Laudon 2012

Normalized Tables Created From Order

Pros: Data integrity and updating Cons: Processing speed for large data sets

Source: Laudon and Laudon 2012

Charitable Contributions

Source: Kishore-jaladi-DW.ppt

The “Classic” Star Schema

A single fact table, with detail and summary data

Fact table primary key has only one key column per dimension

Each key is generated Each dimension is a single

table, highly de-normalized

Tradeoff between data integrity, updating and speedSome alternatives: Star and Snowflake structure

Benefits: Easy to understand, easy to define hierarchies, reduces # of physical joins, low maintenance, very simple metadata

PERIOD KEY

Store Dimension Time Dimension

Product Dimension

STORE KEYPRODUCT KEYPERIOD KEY

DollarsUnitsPrice

Period DescYearQuarterMonthDayCurrent FlagResolutionSequence

Fact Table

PRODUCT KEY

Store DescriptionCityStateDistrict IDDistrict Desc.Region_IDRegion Desc.Regional Mgr.Level

Product Desc.BrandColorSizeManufacturerLevel

STORE KEY

Data Integrity and Hygiene

Illustrating Data Hygiene  Quantities   Response Response RateCustomers 2,000,000   29,000 1.45%  Undel. 15% 1,700,000 15% 29,000 1.71%Dup. 20% 1,360,000 20% 29,000 2.13%

    Cost   CPOCPM = $500 2,000,000 $1,000,000 29,000 $34.48    1,700,000 $850,000 29,000 $29.31  1,360,000 $680,000 29,000 $23.45    Revenue   Profit ROIPrice = $60 2,000,000 $870,000 29,000 -$130,000 -13%GM 50% 1,700,000 $870,000 29,000 $20,000 2%  1,360,000 $870,000 29,000 $190,000 28%

BE = FC / (P-C) 1,000,000 / 30  $       33,334 BE = FC / (P-C)    850,000 / 30  $       28,334 BE = FC / (P-C)    680,000 / 30  $       22,667 

Data Hygiene Processes (1) Standardize names

Title, First name, Initials, Family name, Suffix Standardize addresses

Address 1, Address 2, City, Province, Postal Code Abbreviations (apt., ave, p.o., province) Replace prestige names with postal addresses (i.e.

Commerce Court) Scrubbing

Ex. c/o, co, c/o Delivery

FSA/LDU, Postal walk Address change database

Data Hygiene Processes (2) Data Comparison

Duplicate (cost, abuse) Householding

• Hyphenated Names, Maiden Names, Spouse’s Name• Recomposed Families, Roommates

Consolidation (merge/purge)• Multiple Accounts (financial Services)• Multiple policies (insurances)• Multiple phone numbers (telco)• Multiple divisions within firm

Wrap-up Issues with Mkt Databases DBMS Database Design and Schemas Data Integrity and Hygiene Demo and Lab: Table redundancy

and Queries

Next Week Data Import Data Preparation Data Transformation