(Aug. 09. 2001) Like a shopper wandering a huge new mall, IT professionals trying to understand Web analytics find a maze of clashing claims, competing technology and uncertain needs. To sort out this bevy online measurement tools the place to begin is terminology and the technology.
Web analytics lets you use a variety of diagnostic tools to learn about Web site usage. The goal of course is to somehow coalesce the data with off-line information as well as other knowledge gleaned from internal IT sources such as customer relationship management (CRM) systems, inventory systems or even telemarketing operations.
And while the business benefits are compelling, as with many overreaching theories calling for aggregation of data, the implementation can be complex.
The technology of Web analytics can be divided between software run by a particular enterprise and a service offered by an application service provider (ASP).
According to Guy Creese, research director of Internet analytics at Boston-based Aberdeen Group Inc., the historical basis for this technology difference is routed in how data about a Web site is collected. Web logs and network sniffers are the technologies used for Web site data collection by many of the enterprise software vendors. A competing technology that each of these technologies contains advantages and disadvantages, says Creese, author of a December 2000 study on Web analytics for Aberdeen.
“Several companies now offer variations of both technologies,” says Creese. For example software from Ottawa-based Buystream Inc. runs on a company’s server, but can also accommodate tagging. Unlike most software that processes back-end logs, it uses a browser tracking model similar to an ASP.
Geric Johnson, the vice president of direct marketing at show retailer Skechers USA Inc., opted for the ASP model using San Diego-based WebSideStory Inc.’s solution called HitBox Commerce.
“We are growing so fast,” says Johnson, “my staff needed something to interpret the click trails of our customers and potential customer.” A benefit for Manhattan Beach, Calif.-based Skechers is the ability to quickly understand where online visitors go and what they want, “In the past we attempted to get the log files off the Web, but they are unbelievably large – it took us 26hours to analyze 24 hours of logs – we are not using the activity reporting function of HitBox and can see where people are bailing out of the site.” This allows Skechers to modify the online display of merchandise. “We found people wanted to get right to the product, they didn’t want this whole community experience,” says Johnson, who has 11 year’s experience in retail marketing. “A general rule of thumb is if you can get 2% of site visitors converted to sales, that’s tremendous on the Web.”
For Skechers, the additional boost it received from Web site analysis was the ability to merge information with off-line data from its catalog operation. “We want to build a relationship with the customer in whatever manner the customer prefers.” He says. And while 70% of the catalogs go straight into the trash, the remainder presents an opportunity for Skechers to transition them to the Web. “Whenever the catalog drops, Web usage takes off,” notes Johnson.
As far as the actual technology is concerned, it was simply a matter of tagging the Web site’s pages. “We wrote a little routine so any page we create has a HitBox tag,” Johnson says.
Clearly in the retail sector, analyzing customer movement through a Web site can give managers a window into how a product is performing and can help make a determination if a particular product should be a candidate for wider production.
At Brooks Brothers Inc., Nelson Sanchez, director of e-commerce marketing at the New York-based retailer, uses HitBox Commerce to tie order and dollar information to traffic data, “to tell us which visitors are the most valuable and contribute to a higher conversion rate,” he says.
For example, Sanchez explains that even though search engine A may send three times the visitors to the Brooks Brothers site, search engines B’s visitors convert at a rate twice that of A. As a result, Sanchez says, “Perhaps we will re-evaluate which search engine we advertise one.”
One of the specific pieces of information Sanchez received form the analytics was AOL users have a higher propensity to buy, at a higher order value. “That was a surprise,” he says. In addition he found Japan business larger than previously thought. “We may cater to that market specifically,” he added.
Sanchez feels using a Web analytics tools is essential in evaluating marketing campaigns. “At a very reasonable start-up and monthly cost and a fairly painless implementation, it’s a no-brainer.” Sanchez says, and while there is no firm way for him to calculate the return on investment on using the technology, he feels that just optimizing the search engine placements more carefully according to conversion rates would be a simple ROI.
Creese says he agrees that Web analytics can help target online marketing campaigns by establishing the profitability of banner ads and other online promotions such as e-mail newsletters. But there are other business benefits as well.
“The reports generated by Web analytics can help in forecasting Web site traffic and can also lead to more compelling online and off-line offers,” Creese says. “You can segment Web site visitors this way,” In addition, Creese points out there are other aspects to consider in relation to Web analytics technology.
Some elements include:
- Visitors profiles
- Data capture
- Transformation of visitors
- Storage of data
- Report and action
“This gives people a frame of reference to think about the different systems and architectures available,” says Creese.
Part of Terry Lund’s decision for Web analytics as the director of site capabilities at Kodak.com was the vast amount of legacy data that existed inside Kodak’s IT operations. Prior to selecting enterprise software from Fremont, Calif.-based Accure Software Inc., “we were using homemade tools,” says Lund. For Kodak, Accure offered two advantages in analyzing Web traffic: network collection rather than just log files, and packet sniffing. Packet sniffing involves network devices that monitor the TCP/IP packets that stream by. To work, sniffers have to be installed on the server’s network segment, which means “they cannot be used in hosted environments, such as an ISP,” says Creese. Packet sniffers also have a drawback similar to Web logging technology, as they can’t decode encrypted packets created during an actual purchase session online.
But clearly not all Web sites need analytics linked to e-commerce functions.
As Web information manager at the Menlo Park, Calif.-based Stanford Linear Accelerator Center (SLAC), Ruth McDunn wanted Web site analysis to report activity on three different serves running three different types of software – Unix, Linus and an older VAX machine. The SLAC, which is founded by the U.S. Department of Energy but run by Stanford University, doesn’t permit the use of cookies on its Web site (about 200,000 indexed sites in all), so capturing data using log files proved more than sufficient. “I used to do researches pretty much by hand, dumping the log files into Excel,” says McDunn, who decided to use the Web analytics tool NetGenesis5 from Cambridge, Mass.-based NetGenesis Corp.
The enterprise software runs on a Unix server using an Oracle database. Each night McDunn’s IT team closes out the log files, and a script goes out from NetGenesis5 and pulls the logs onto the Unix server. McDunn has also been able to set parameters as to the kind of data transferred each night – deleting the large number of graphic files that might otherwise be logged and reported,” It is a way to get rid of a lot of noise,” McDunn says. About 50MB worth of logs are extracted each session. “We set up out Oracle tables to accommodate 40GB and, if necessary, we would just increase the amount of RAM on the Unix box if the log files became too large,” she adds. Using NetGenesis5, which can also capture data using network packet sniffers and Web server plug-ins, McDunn noted. “The information we get allows is t work on the parts of the site that are the most popular, and because we have so many sites it is important to have a tool that can pull information from a variety of server sources.”
The variety of data sources extends beyond Web servers to encompass data servers and message service as well as more traditional database sources such as DB2, SQL Server and Oracle.
San Jose-based NetIQ Corp. (formerly WebTrends) has used the ability to analyze data from multiple sources to expand beyond its WebTrends Log Analyzer software to support distributed reporting, clustered servers and multiple platforms. Indeed the company has pursued partnerships with CRM, content management and analytics suppliers for just this reason. As a result, the definition of where analytics ends and e-business begins is starting to blur. Customers for NetIQ include MLB Advanced Media LP, the interactive media and Internet company of Major League Baseball.
Another approach using a hybrid of data collection methods is Bellevue, Wash.-based digiMine Inc., which offers a managed service for Web analytics that acquires click stream data from Web logs, transaction servers and any other database source a customer might have. Billed as a data mining and warehousing service because of the ability to synthesize information from these different sources, customers such as Seattle-based Getty Images Inc. use digiMine to look at where Web visitors go and where they have trouble, as well as whether the visitor received a paper catalog. Bud Albers, chief technology officer at Getty Images, says he packs off the logs to digiMine every night and the next day he is able to get a variety of reports about Web site traffic. “They combine all the different data and we get to choose how we want to look at the site,” he says.
What becomes evident is researching different data collection and reporting mechanisms is the choice depends almost entirely on the individual company’s requirements and expertise. As Creese point out, “Data collection perfection is unattainable, the engineers who designed the Internet did not design it with the auditing of end-user behavior in mind.” He says the different tracking systems are going after “digital crumbs” to re-create a path through a Web site and that each method has pluses and minuses.
“Back-end methods such as Web logs, packet sniffers and server plug-ins gather different views,” he says. For example, Web logs capture most user clicks but might be confused by proxy servers. Packet sniffers log low-level network behavior and might leave out transactions. Server plug-ins are good at collecting application-level views. “These technologies might underreport page click,” Creese says.
ASPs counter that their browser-based collection methods are therefore better, but, notes Creese, “What happens if the ASP goes down?” Indeed, Creese says that we will begin to see a mix-and-match scenario in which data from multiple enterprise systems is pulled together to offer a greater view of overall Web site activity.
So while the market will consolidate and hybrid solutions will evolve, there is one thing all Web site participants – marketers, salespeople, technical managers and designers – agree on. You need to start somewhere. Web analytics is crucial to understanding why you have a Web site in the first place.
By: Pimm Fox
Source: Computer World