Bases de Dados e Armazéns de Dados

Departamento de Engenharia Informática (DEI/ISEP)

Paulo Oliveira
[email protected]

Dimensional Data Modeling

14

1
 Dimensional Modeling
 Build around the numerical
measurements of the business
- Fact tables contain measurements
- Dimension tables contain the
context surrounding measurements
- Measurements are taken at the
intersection of all dimensions
- List of dimensions defines the grain
of the fact table

DATE

15

15

Fact Table
 Is the primary table in a dimensional model
 Holds the measurements of the business
 Composed by a set of foreign keys that connect to
the dimension tables
 Its primary key is made up by the set or a subset
of the foreign keys

 Role of a normalized n-ary associative entity

 All measurements in a fact table must be at the
same grain

16

16

2
 Fact Table doesn’t store “non-events”
 Very important not to try to fill the fact table with
zeros representing “nothing happened”
- If there is no sales activity on a given day, in a given store, for
a given product, the record must be left out of the fact table

 By only including true activity, fact tables tend to be

quite sparse

 Despite their sparsity, fact tables usually make up

90% or more of the total space consumed by a
dimensional database

17

17

Sparsely Fact Table

Date
Time

month
week
Product

brands
categories
CUSTOMER

Region

Fact by:
day
date
product
Customer
customer

18

18

3
 Dimension Tables
 Define the details of each transaction

 Dimension tables answer the “who”, “what”,

“when”, “where” and “why” of a business
event
-For example, a sales transaction may be defined by
a number of components:
 Customer: who made the purchase
 Product: what was sold
 Store: where it was sold
 Date: when it was sold
 Promotion: why it was sold

19

19

Dimension Tables
 Provide the context for fact tables, that is, the
context for all the measures

 Dimension tables have many columns or attributes

- Usual for a dimension to have between 50 to 100 attributes
- Relatively small in terms of the number of rows
- Usually much smaller than fact tables

 Best attributes are textual and discrete

 Entry points into the fact table

20

20

4
 Dimension Attributes
 Serve as the primary source of query constraints,
groupings, and report labels

 Key to making the DW usable and understandable –

DW is only as good as the dimension attributes

 Each dimension is defined by its single primary key

– surrogate key, which serves as the basis for
referential integrity with the fact table(s) to which it
is joined

21

21

Surrogate Keys
 Joins between dimensions and fact tables should be
based on meaningless integer surrogate keys
- Other names: integer keys, no natural keys, artificial keys,
synthetic keys

 Must be assigned sequentially

 Benefits:
- Performance advantages
- Protects the DW from operational changes
- Allow the integration of data from multiple operational source
systems

22

22

5
 The Grocery Store

23

Grocery Store Business - Brief Description

 The business has 500 large grocery stores spread over the
country. Each store is divided by departments such as grocery,
frozen foods, dairy, meat, bakery, floral, drugs,… Each store has
roughly 60000 individual products (called Stock Keeping Units –
SKUs) on its shelves. About 40000 SKUs come from outside
manufactures and have bar codes called Universal Product Codes –
UPCs.
 The remaining 20000 SKUs come from departments like the meat,
bakery, or floral departments and don’t have UPC codes.
Nevertheless, as a grocery store, these products also have SKU
numbers assigned to them.
 At the grocery store, management is concerned with the sales of the
products as well as maximizing the profit at each store. The most
significant management decisions have to do with pricing,
promotions and good visibility of promotions.

24

24

6
 Kimball Dimensional Modeling Steps
1.Identify the business process
- Business process is a major operational process supported by
some computational system(s) from which data can be
collected for the purpose of data warehousing (e.g.: orders)

2.Identify the level of detail (grain)

- Detail level of the data to be represented in the fact table
- Determines the dimensionality of the underlying database
and has a profound impact on its size

3.Identify the dimensions

- Choose the dimensions that will apply to each fact table
- For each dimension describe all its attributes

4.Identify the facts

- Choose the measures that will populate each fact table record
25

25

Modelling Grocery Store Business

1. Business process to model
 Sales

2. Granularity level (level of detail)

Options:
 Sales of products by store by promotion and by individual
customer ticket transaction
In this grocery store chain, there is no effective way of identifying
individual customers at the cash register
 Sales of products by store by promotion and by day (or by week or
by month)
Weekly or monthly storage item movement would miss too many
important analysis, such as difference in sales between Mondays and
Saturdays

Best grain for this grocery store chain DW is considered to

be the product (or SKU) sales, by store, by promotion
and by day
26

26

7
 Modelling Grocery Store Business
3. Dimensions involved
- Date
- Product
- Store
- Promotion

4. Facts/Measures of interest
- Value sold
- Units sold
- Sales cost
- Sales profit
- Sales margin

27

27

Date Dimension
 Date dimension is present in every DW, because every DW is a
time series
Date Dimension
date-key
full-date
day-week Unlike almost all the other
day-number-month
day-number-year dimensions, date
week-number dimension can be built
month-name in advance – five or ten
month-number
semester year of history records can
quarter be loaded, as well the next
year few years
last-day-month-flag
season
…

 Surrogate key assigned to the date dimension should be

assigned consecutively in the order of date

28

28

8
 Product Dimension
Product dimension describes every SKU with as many descriptive
attributes as possible, including the existing hierarchies
Product Dimension
product-key
SKU-description
SKU-number
package-size
brand
subcategory
category
department
package-type
diet-type
weight
weight-unit
…

It is possible to browse among dimension attributes whether or

not they belong to a hierarchy and it is possible to roll up and
drill down using the attributes that belong to a hierarchy
29

29

Store Dimension
Store dimension describes every store in the grocery chain –
geographic dimension
Store Dimension
store-key
store-name
store-number
store-address
store-zip
store-city
store-district
store-region
store-manager
open-date
last-remodel-date
store-sqft Numeric attributes,
grocery-sqft
…
however they are clearly a
constant attribute of store

30

30

9
 Promotion Dimension
 Promotion dimension – describes each promotion condition
under which a product is sold in the grocery chain
 Causal dimension – describes factors that cause a change in
product sales
 Needs a special register ”N/A” to join sales in fact table
without promotion
Promotion Dimension
promotion-key
promotion-name
price-reduction-type
ad-type
display-type
coupon-type
ad-media-name
display-provider
promo-cost
promo-begin-date
promo-end-date
…

31

31

Fact Table

date-key … value-sold units-sold sales-costs sales-profit sales- margin

Sales Fact 101 … 780 78 263 517 0,66
date-key 102 … 1044 18 580 464 0,44
103 … 213 10 140 73 0,34
product-key 104 … 95 19 39 56 0,59
store-key Total 2132 125 1022 1110 0,52
promotion-key
value-sold
units-sold
Business Measures / Facts
sales-cost Is not
sales-profit stored
in the
Fact
 First three facts are additive
Table
 sales-profit = value-sold  sales-cost  additive

 sales-margin = sales-profit / value-sold

No-additive calculation – can be calculated for any slice of fact
table by calculating the sales profit and value sold before dividing

32

32

10
 Grocery Store Business Schema

Sales Fact
date-key Product Dimension
Date Dimension
product-key product-key
date-key store-key …
… promotion-key
value-sold
units-sold
sales-cost
Store Dimension sales-profit
Promotion Dimension
store-key
promotion-key
…
…

Advantages:
- Easy to understand
- Better performance
- Easy extensible: new dimensions and new facts

33

33

11