Accuracy of Inferences

Commerce Server 2002 uses the concept of user visits to understand Web site traffic. A visit can be defined as a set of requests from the same user within a specified time span. However, the identification of users is not completely accurate because the Web client computers send only a limited amount of data to the Web server computer. Commerce Server attempts to identify users based on the properties in the Web log file. Inferences refer to the process of identifying unique users based on the user and visit data in the Web log file.

The accuracy of inferences depends on the data that is available. To maximize accuracy, Commerce Server uses a user inferencing algorithm on each hit in a priority order:

  • If the hits include Commerce Server or customer-defined cookies, or user names, the accuracy is quite good because the users are being identified explicitly: they have enabled cookies, they are within a domain, or your site is using AuthManager, which supports cookieless shopping.

  • If the hits do not include cookies or user names, Commerce Server uses a combination of the client IP address and client browser agents to identify the users. In this scenario, the accuracy cannot be measured.

  • If client IP addresses are translated to a single IP address, use either cookies or user names in the Web server log import for inferring users.

    Two DTS tasks are affected by IP translations:

    • Web server log import task. When inferring users, if the IP address of some users is translated to a single IP, and if the users all use the same browser, then they will be treated as same user.

      For example, if there are ten users whose IP address is translated to single IP address, and five of those users use browserA and remaining five use browserB, then all ten users will be treated as two users. This is the result if the IP address and browser type (instead of cookies or user names) are used for inferring users.

    • IP resolution task. When some IP addresses are translated to a single IP address, you may not get accurate domain information on all of the users visiting the site because the IP resolution task performs a look-up on only one IP address.

    To use either cookies or user names to infer users, see Setting User Inferences.

You can set user and visit inferences in the Data Warehouse to customize the assumptions made during import about users and visits to best meet the analysis needs.

Ee796976.note(en-US,CS.20).gifNote

  • After you perform these procedures, you must synchronize your site configuration with the Data Warehouse before importing data into the Data Warehouse.

For information about analyzing data from cookies, see Analyzing Data from Cookies.

See Also

Running the Configuration Synchronization DTS Task

Configuring the Web Server Log Import DTS Task

Importing Data into the Data Warehouse

Extending the Data Warehouse

Data Warehouse Schema

Troubleshooting the Data Warehouse Import Process

Copyright © 2005 Microsoft Corporation.
All rights reserved.