Ways to Use Data
There are three general ways that enterprise data is used:
-
To process transactions.
-
To discover strategic business information.
-
To maintain history.
Much of the same enterprise data is used in all three ways. This data often needs to be moved and transformed as it is being used for different purposes.
Transaction Processing
Companies gather a great deal of data during the ongoing operation of their business. This data includes information about the following:
-
Orders
-
Purchases
-
Production
-
Sales
-
Shipping
-
Inventory
-
Employees
-
Expense and revenue
-
Customers
-
Customer contacts
The primary purpose of On Line Transaction Processing (OLTP) data is to keep track of individual transactions so that each one can be handled properly. A company needs to be able to answer these questions:
-
What did the customer order?
-
Are the ordered products available?
-
How much was the customer charged?
-
To what address was the product shipped?
-
When was the order shipped?
-
When was the payment received?
Business Analysis
Data is needed for business analysis. A company needs to know more than the individual details of its business transactions. This information has to be assembled into a meaningful format so that it can be used for business decisions.
The business analyst asks these kinds of questions:
-
Which stores had the greatest increase in sales this past month?
-
Which types of products are increasing in sales? Which are decreasing?
-
Who are our most profitable customers? The least profitable?
-
What are our most profitable products? Our least profitable?
-
Did we make a profit?
-
How can we make a greater profit in the future?
There are three broad categories of business analysis—reporting, OLAP, and data mining.
Reporting
Business analysis has traditionally taken the form of reporting. A report is typically a collection of spreadsheets that provide information about a company's performance from a variety of perspectives.
OLAP
In recent years, a new kind of interactive reporting tool has been developed. On Line Analytical Processing (OLAP) is a software tool that presents millions of different spreadsheet views to an analyst. An OLAP tool allows a user to easily move between these different spreadsheets:
-
Drilling down to a level of greater detail.
-
Drilling up to see the broader picture.
-
Looking at the data from different perspectives or dimensions.
-
Slicing (filtering) on a particular factor.
-
Combining dimensions to see how different factors interact.
Microsoft included OLAP Services with SQL Server for the first time in version 7.0. In SQL Server 2000, OLAP is part of the functionality of Analysis Services. Figures 3.1 and 3.2 show the process of drilling down to a greater level of detail using the Excel PivotTable.
){kind=link}
){kind=link}
Figure
3.1
OLAP data viewed in Excel 2000.
Figure
3.2
You can drill down to a greater level of detail.
Figure 3.3 shows the data from a different perspective or dimension—customer income instead of store location.
){kind=link}
Figure
3.3
You can change perspectives (dimensions).
Figure 3.4 shows the Store Location being used to filter, or slice, the results.
){kind=link}
Figure
3.4
You can filter (slice) the data from another perspective.
Chapter 21, "The Analysis Services Tasks," describes the use of the Analysis Services Processing task for processing cubes. Chapter 31, "Creating a Custom Task in VB," shows how to create a custom task for generating local cube files.
Data Mining
Data mining is an automated form of business intelligence. A data mining tool is programmed to analyze a set of data to find significant patterns.
Microsoft Analysis Services in SQL Server 2000 includes data mining functionality. There are two data mining algorithms included with Analysis Services:
-
Decision Trees—Discovering the most likely patterns for predicting data values.
-
Clustering—Finding the natural ways that records can be assembled into groups.
Figure 3.5 shows the Customer Pattern Discovery mining model from the Foodmart 2000 sample database.
){kind=link}
Figure
3.5
The data mining model browser in Analysis Services.
Chapter 21 explains how to use the Analysis Services Processing task to process cubes and data mining models, and how to use the Data Mining Prediction Query task to use a mining model to predict unknown data values.
Note - I think "data mining" is a wonderfully descriptive term for the automated analysis of business data. Tons of ore have to be processed to obtain a few ounces of gold. In the same way, we have a mountain of data—much more than we could ever understand. The data as a whole is worthless, but hidden in that data are nuggets of significant business information. Our data mining tools sift through the massive quantity of data to find the facts that have significance for our business processes.
Maintaining History
The third primary need of an enterprise data structure is to maintain history. A company needs to have an accurate record of what has happened in its past—for historical business analysis and because these transactional records may be needed to resolve business disputes.
Data used for maintaining history usually does not have to be as readily available as the current transaction or business analysis data. It is often archived to a different database so that the volume of data in the current OLTP and OLAP databases can be maintained at a manageable size. Archived data is often kept online in a summarized form for business analysis.