CS5504/CS4504: Business Intelligence Systems, Infrastructures and Technologies

Surgery 3 – Sample Solution

Q1
The manager of an airline company wants to know, daily, the number and revenue of flight
bookings made, per flight, at the different travel agencies.
a. Design an appropriate dimensional star schema to satisfy the above requirement.
Explicitly state the grain of such schema and devise at least two non-key attributes for
each of its dimensions; such attributes should make sense. Also, provide a sample row
of all tables of the proposed schema illustrating the respective data interrelationships.
b. Redesign your schema to a snowflake scenario of your choice.
Sample Answer

Grain: Flight Bookings, per flight, per travel agency, per day.

1
Date
Date_Key Date Description Day of Week Month Year
1 21/02/2014 February 21st, Friday February 2014
2014
Travel Agency
TravelAgency_ Name Description Address City Country
Key
1 XPT Travel Central 18, Bond St London UK
London
Branch
Flight
Flight_ Description Destination Origin Capacity
Key
1 European London Lisbon 320
Flight
Flight Bookings Facts

TravelAgency_Key Date_Key Flight_Key No of Bookings Total Revenue

1 1 1 5 £2 100

Example Snowflake
Split Travel Agency (City, Country)

2
Q2. Sample Answers in Italics
You are given the following dimensional schema of a BI system. The schema is derived from the
OLTP system of a company that sells a variety of products, to several customers worldwide. The
company products belong to different categories and come from various suppliers. Customers
are located worldwide, therefore products have to be shipped to them.

Study the dimensional model above and answer the following questions:
1. What is the granularity of the model? What are the measures?
Grain is at the OrderLineItem (per Customer, Employee, etc). Measures: Unit Price,
Quantity, Discount, Sales Amount, Freight
2. Do snowflake elements/structures exist in the schema? If yes, which ones?
Yes, City/ State/ Country/Continent, and Product/Category
3. Do degenerate dimensions exist in the schema? If yes, which ones?
Yes, Order No, OrderLineNo

4. Do role playing dimensions exist in the schema? If yes, which ones?

Yes, Dimension Time OrderDate, DueDate, ShippedDate
5. Are there bridge tables in the schema? If yes, please specify.
Yes, Territories Table (many-to-many relationship between Employees and Cities)
3
6. Are there natural keys (source system keys) in the schema?. If yes, please specify.
Yes, CustomerID (Customer Table)
7. Identify some potential slowly changing dimensions.
8. Are there any mini-dimensions in the schema? If yes, please specify.
No, there are not.
9. Do any ragged hierarchies exist in the schema? If yes, please specify.
Yes, City, State, Country
10. Propose an additional fact table that would make sense to include in such a model so
that to form a fact-constellation schema?
Fact PurchasesFromSuppliers related to Products, Time, Supplier

4
Q3.

The manager of the cards department of a bank wants to know on a daily basis the total number
and total amount (i.e. total value) of credit and debit card purchases performed at the
Point-of-Sale (POS) machines of the bank across the country.

Design an appropriate dimensional schema to satisfy the above requirement. The granularity of
such schema should represent every transaction (purchase) performed at each POS of the bank.
Devise at least two non-key attributes for each of the dimensions of the schema.

Provide also a sample row of all entities of the proposed schema illustrating the respective data
interrelationships.
Sample Answer

Date
Date_Key Date Description Day of Week Month Year
1 21/02/2016 February 21st, Sunday February 2016
2016

POS Machine
POS_ POS Manufacturer Merchant Address City Region
Key Code ID
1 322 ABC Co ZARA 18, Bond St London London
03

5
 Card
Card Card Number Card Network BIN Cardholder Name Expiry Date
Key Type
1 5167320011923457 Debit Visa 5167 John Smith 17/04/2017

POS Purchases Facts

POS_Key Date_Key Card_Key Purchase Number Total Amount

1 1 1 123457534 £35.09

Q4. Sample Answers in Italics

Consider the following STAR schema of a supermarket chain.

The chain has 300 stores and approx. 40 000 products. Also, there are 500 products per brand.
Out of the 40 000 products, assume that only 4000 sell in each store daily, and that a sold item
may be in only one promotion per store, per day. Finally, there is at least one sale per product,
per store, per week.

6
a. What is the maximum number of fact table rows in this schema? What is the (maximum)
number of fact table records that are actually stored in this schema?
40 000x300x1825 = 21,9 billion
4 000 x 300 x 1825 x1 = 2,19billion
b. Estimate the number of fact table’s rows to be retrieved by the following database
queries.
▪ Query involving 1 product, 1 store, 1 week.
1 fact table row
▪ Query involving 1 product, all stores, 1 week.
300 fact table rows

▪ Query involving 1 brand, 1 store, 1 week.

500 fact table rows
▪ Query involving 1 brand, all stores, 1 year.
300x500x52 = 7,800,000 fact table rows