Web Analytics and Data Warehousing

Tue, 04 Apr 2006 06:26:26 GMT

Extracted from Graves et. al., Professional SQL Server 2000 Data Warehousing with Analysis Services, Wrox Press, ISBN: 1861005407; Chapter 18:

Web Analytics is the analysis of web data. More specifically, it is business intelligence that allows analysts to gain insight into their web-based businesses (e.g. e-commerce, customer support, etc.). Web Analytics allows you to understand customer behavior and identify common sales trends allowing you to personalize services for your customers and increase customer satisfaction. It allows you to understand how well your products, content, and online processes are working.

The difficulty of Web Analytics is that it requires you to combine different sources of data and correlate the data in order to resolve these issues. With the complexity of web sites, offerings, multiple locations, and promotions, it is difficult to answer even the most basic questions concerning these online businesses. For example, in order to answer the question "How well did our web promotion work", you will need to:

Transform web log data to determine the number of users who visited the site due to the
promotion
Transform commerce data to determine the number of users who made purchases due to the
promotion itself
Associate the commerce data to the web log data to include only purchases that were made due
to those promotions
Transform banner ads and e-mail marketing campaign data to determine the number of users who
actually received the promotion
Finally, correlate this data against all of the data above

For more info, refer to the book, eh?! Meanwhile, a good article on "Storing Session State in a SQL Server database" which is key to web analytics data warehousing (i.e. handling the scale of data):

http://www.developer.com/db/article.php/3595766

Data Warehousing Planning

Wed, 15 Feb 2006 05:11:15 GMT

When planning to build an enterprise data warehouse, there are many of small details ranging from hardware allocation to error reporting. So unless you plan for this correctly, you may deliver an inadequate system.

A good article to review on the issues to resolve when planning a data warehouse can be found in Craig Utley's article within the July 2004 edition of SQL Server Magazine "Beating the Odds"

Company Politics: turf battles, lack of clear vision, bureaucracy, etc.

Lack of User Involvement: Ask the questions and figure out what is needed and what to report.

Overbuilding: Do not try to build a warehouse that has everything. A big problem with enterprise warehouses is that one tries to put all attributes into it, irrelevant of whether you need it or not.

Super cubes: related to “Overbuilding”, placing all dimensions and measures into one cube will result in poor query and processing performance.

Key Performance Indicators (KPI): related to cube design, unless you know what you are measuring – you’ve created cubes that provide numbers not metrics.

User Interface (UI): Even if you build the perfect warehouse, unless there is a good UI in front of it no one can actually access the data.

Poor Data Quality: bad data = bad reporting. What’s the point of reporting if you cannot trust the report?

Inadequate Funding: proper budgeting from the people to the resources required. Don’t forget bug fixes, training, and the fact that data warehouses are consistently evolving.

Denny Lee's Windows Live Space: Data Warehousing

Web Analytics and Data Warehousing

Data Warehousing Planning