MondoSoft's BehaviorTracking

Setting Up Shop

I installed BT on a test server with dual 1,400-MHz Intel Pentium IIIs, with 1,024 MB of RAM, in our Syracuse University Real-World Labs. The product requires Windows 2000 or 2003 Server with at least a 1,000-MHz Intel Pentium Processor, 512 MB of RAM and 500 MB of disk space; Microsoft IIS (Internet Information Server) with .Net Framework 1.0 or 1.1 to drive the user interface; and Microsoft SQL Server 2000 (SP3a) to store search information. I installed Ultraseek 5.2.1 search engine and indexed our production Web site (www.nwc.com).

The BT installer identified my running copy of IIS and installed a virtual directory linking the Web server's root file system to BT's application folder. Once all the files were in place, I used the configuration wizards to finish the setup. I created a database for BT using Microsoft SQL 2000 Server, then identified the URL path to BT's log files, and defined an administrative user ID and password. I could create other users with individual settings from here as well. Unfortunately, the product does not integrate with LDAP or Active Directory.

The BT Connector installed on the same server as BT and Ultraseek. The installer placed the Connector and its support files in a subdirectory under the same file structure as BT, then mapped the subdirectory to a new virtual directory in IIS.

The Connector (BTCLog.dll) is an ISAPI program that works with IIS. It accepts log data from the Web server that runs the search-engine interface and writes the information in a format BT understands. By default, the Connector runs as the Internet Guest Account (IUSR_ ). If users authenticate to access the Web and the search engine, the Connector runs under their flag, so they'll need read and execute permissions to the Connector and full access to the directory where the logs are written.Start Seeking

I had to modify the templates to Ultraseek's user interface so it could submit user-tracking information (queries and click details) to BT log files via the Connector. To do this, I added a transparent 1x1 pixel GIF image to four of the files making up the user interface in Ultraseek. Here's a example of how to do that:

Mondosoft includes four template files for Ultraseek with embedded links. It also includes the files you need for SharePoint. If you use an unsupported search engine, you'll need to decipher the Connector arguments and construct your own transparent GIFs to add to your search-engine interface.

After replacing the template files for Ultraseek, I accessed the Web interface to BT, changed BT's log file location to point to the same location as the Connector's log files, and ran a preconfigured batch file to collect and import the log data for BT to process.

BT was blind to search-engine statistics until I ran a preconfigured batch file (MssStatImp.bat) that collects the log data and imports it into Microsoft SQL. I executed this once and, using the Windows Scheduler, I scheduled it to run daily. Once the data is imported from logs, BT makes it available through its Web browser interface. This sounds like an awful lot of work, but the results are worth it: BT provides much more than a log analyzer; it includes an analysis and diagnosis of poor search results and strategies for improvement.

BT's default home page provides an overview of the number of searches (including those with no results, and with results but without clicks), sessions and clicks (links or pages selected from search results). It also depicts search-engine performance in a Performance Map. The map graphs the percentage of searches with results over the percentage of searches with clicks.Using data from the Performance Map, Mondosoft calculates a Site Success Index. The SSI indicator ranges from 1 (low) to 100 (high) and comes with basic tips to improve the score. For example, you can add synonym and stemming strategies to the search engine to increase the number of hits per search. It also provides tips to improve the click rate on results such as applying search categories or subjects and utilizing metatags. Although BT can generate synonyms for search terms, enhancing or modifying search-engine features is up to you and your search-engine vendor.

From the Top

Detailed information on searches, available from the "Searches & Clicks" menu (see screen, below), includes a view of the top search terms, the top "not found" search terms (search terms with no results), the top "not clicked" terms and the top links selected. I could even enter a URL in a search window to see how often it was returned in response to a query, what search terms triggered its return and how often it was selected from the entire result set sent to an end user.

Drilling down, I could view "related terms" (terms used most within a single session), the most recent sessions that included those terms and the individual pages selected within those sessions. This information gave me a better view of a site user's search strategy.

Session-detail information includes host name, date, duration and number of searches in the session. It contains what a user searched, retrieved and clicked on--almost as good as looking over the shoulder of someone searching your site.Watch groups can be set up by IP address. After I defined a watch group ("doherty"), I could easily view the search data associated with my IP address, but I couldn't enter an IP domain to view all the search information coming from a certain domain. It would be a boon to see what information a competitor searched and retrieved from my Web site.

In addition to viewing this information from the Web, I could print the data in table format, as well as export it to Microsoft Excel. I set up BT to generate regular reports and e-mail them to me in Excel.

Worth the Price?

BT's installation is time-consuming, but it's justified by the rich search data you get in the end. Although the sticker price for BT is three times the cost of the Ultraseek search engine, it is competitive in this unsettled market of software and service solutions with five- and six-figure price tags. If you're searching for a new statistics package and want to find out how your search engine is behaving for end users, BT should be on your short list.

Sean Doherty is a technology editor and lawyer based at our Syracuse University Real-World Labs. Write to him at [email protected].0