Web Analytics Tutorial

Appendix B – Technical Details of Metrics Accuracy

Advanced Solutions

Summary uses an advanced heuristic algorithm to correlate all requests from each visit for analysis. This algorithm is an excellent estimate, but not exact. Especially when dealing with proxy servers (and other servers with proxy effect), the heuristics can confuse distinct users who appear to be using the same computer. Fortunately, there are a few techniques that can be employed to improve the accuracy of visit tracking.

Cookies

In Lesson 1 - What is ‘Web Analytics?’, we mentioned that one of the most popular techniques for improving visit metrics is to configure your web server (or site) to manage ‘session cookies.’ A session cookie is assigned to each visitor who does not already have one. The cookie lasts until the visitor closes his browser. So each cookie is associated with all of his requests for a single session. While each visitor’s browser may still cache some of the requests, two different visitors accessing your site through a web cache or proxy server will make distinct requests to your server. Summary SP can be configured to use session cookies for visit identification if you have configured your web server to log them.

Session Keys in URLs

Some web servers and content servers allow you to create session identifiers that are put in the URL of each request. This usually takes the form of a CGI variable, but could sometimes be an extension of the request name. In order for this to work, every link on every page needs to be dynamically changed to include the session ID. Then, as a visitor browses through your site, each request is tagged with the session identifier, making it simple to combine the proper session codes into a single visit. Summary SP can use these session keys for visitor tracking on web servers (such as IIS) that log then session ID to a separate field.

A major advantage of session identifiers in URLs is that web and browser caches will not cache the item between sessions because the request is distinct. Your server will still not get requests on repeat visits to a given page in the same session that have been cached by the visitor’s browser. However, the session identifiers have some disadvantages too. First, they do not exist on the first request of a visit because the file requested did not contain an ID. So using URL-based sessions keys will not properly identify entry pages of visits. Second, if visitors bookmark a page or publish a link to your site, it could contain a session ID which will confuse the session tracking algorithm.

Client-side Applets

While HTTP protocol is stateless – that is, there is no connection between subsequent requests from a given visitor – it may be possible to have your visitors run some software that does preserve state. For example, you could develop a custom client-side program that interfaces to the data your server provides or you could include a link to a Java applet in each page of your site that collects and returns tracking information about where your visitors are going around your site. With this technique, you can gather details about the pages visitors accessed, even if they are pulled from the browser cache. A custom solution like this could send information back to a server application on your web server and, if logged in a format similar to web logs, can be analyzed as web logs and provide the most accurate information about visitor behavior. Unfortunately, such client-side applets tend to increase the download time of your web pages. In addition, you will always get some users who have turned off Java support in their browsers and won't get your applet. Finally, some savvy customers may be upset about your information collection practice and view it as an invasion of privacy. Installing such a tool could provide deeper insight to customer behavior, but at the loss of some customers.