Site icon IT World Canada

Merging analytics and intelligence

Web analytics – software that tracks user behaviour on a Web site – is, in theory, not that far removed from the BI (business intelligence) products that have been around for much longer. Both aggregate data and perform statistical analyses to produce reports that tell how well or poorly things are going.

But in practice, Web analytics and BI are still worlds apart. This, according to Kevin Scott, an analyst at AMR Research Inc. in Boston, is primarily due to a cultural divide that still separates all things dot-com from the mainstream side of business.

“Who runs these tools?” Scott asks. “In the case of business intelligence it is the statisticians, whereas a Web site analyst will likely be turning the knobs on a Web analytics implementation.”

Nearly two years ago Jon Zimmerman, director of IT at travel site Expedia.com in Bellevue, Wash., started eyeing the massive amounts of data streaming in from the Web site and piling up in his SQL Server data warehouse. He knew he had gold buried in all that information, and the trick was to find a way to get it out.

“I wanted to see how our customers were using the site, the purchase paths they took, where they were getting confused [and] leaving the site, that sort of thing. These are complex problems,” Zimmerman explains.

Complex because, according to Scott, the raw data from a Web site does not readily yield its secrets: “Typically it is very unstructured strings of text that are not suitable for input to most business intelligence tools.”

And there are lots of those strings. “The amount of data a Web site can gather is huge,” says Scott, adding that this spurred the rise of Web analytics.

Companies such as NetGenesis, Accrue, WebSideStory, and Sane Solutions all sprang up in the last few years to sell customers, such as Zimmerman, products that claim to grab the unruly data stacking up in Web log files and make sense of it.

But they are the new kids on the block. Before Web analytics – even before the Web – was BI. Vendors including Business Objects, Cognos, MicroStrategy, and SPSS were selling data analysis tools. It is possible, therefore, to attack Web analytics with existing BI tools.

This is the first big decision in implementing Web analytics: Do you go with a traditional BI vendor or one of the new-breed companies? You might expect that a company such as Expedia.com, born in 1996 at Microsoft and spun off in 1999, would naturally gravitate to the latter.

Not so. Zimmerman worked with his staff to evaluate several Web analytics tools and found them wanting. “These tools didn’t really address our needs,” he explains.

So Zimmerman went with SPSS, a Chicago firm that got its start in 1968 when Fortran was still the language of choice for crunching numbers. “I came to the conclusion,” explains Zimmerman, “that just because we are a Web site it doesn’t mean that our data analysis needs are all that different from a brick-and-mortar business.”

Zimmerman says the SPSS installation involved following a set-up wizard program and took only days to begin running on his Microsoft SQL Server and IIS (Internet Information Server) platform. The biggest challenge was getting fluent with the tool itself. “It took a few months for us to learn to use some of the scripting languages,” he says.

A word of caution is in order, however: Zimmerman has a master’s degree in operations research, the very discipline that packages such as SPSS’ are designed to support. He also has three people on staff with similar backgrounds.

Zimmerman maintains that IT managers do not need this kind of training to effectively use a tool such as SPSS’ for Web analytics, but there is no denying that it helps.

SPSS is apparently aware of the user gulf; in October the company announced plans to buy NetGenesis in Cambridge, Mass. “[SPSS] wants to get more established in Web analytics,” says Scott. “It makes sense on paper to integrate the two worlds, but I don’t think we will really see strong solutions that provide value across all channels, including the Web, for at least a year.”

If and when that happens, these integrated tools, which meld the functions of Web analytics with BI software, won’t be cheap. The average sale for NetGenesis is already US$150,000 or more, says Scott.

And companies will probably pay at least that much for software from Accrue, in Fremont, Calif., a Web analytics vendor. Gary Beberman, director of technical research at Macys.com, installed Accrue’s software in the summer of 2000.

“When I joined Macys.com in September 1999, we had no sense of why customers were responding to our products,” Beberman says. “I looked at a lot of Web analysis tools and I liked Accrue since they could tell me how customers, in bulk, move through the site.”

He adds it was a quick install. “It is not a perfect solution, but that is partly due to the fact that we are still in start-up mode. We don’t yet have the staff to get as much out of the tool as we would like.”

Beberman’s experience also supports Scott’s claim that Web analytics are still a world apart from traditional BI because, “we don’t yet merge any of our data with Macy’s store data,” explains Beberman.

But there are more economical alternatives to these pricey integrated tools. Web analytics software from Sane Solutions in North Kingstown, R.I., for example, is an order of magnitude cheaper: A typical sale comes in around US$25,000, says CEO Jim Rose.

Rose says Sane has kept to a well-defined strategy since the company’s 1996 inception. “We never planned to compete with the business intelligence vendors,” he says, crediting this focus with protecting the firm from the current economic storm.

Jeff Anderson, senior manager of e-business at Otis Elevator in Farmington, Conn., chose Sane to get a better handle on Otis.com. “We are in over 53 countries and 27 languages,” says Anderson. “We use Sane to track clickstreams and see the demographics of our Web customers.”

Anderson also uses BI software from Cognos. “We use Sane to interpret our log files and, for example, to see who clicks on what piece of content in the [United States],” Anderson explains. “Then we can roll that information up into Cognos to do more traditional multidimensional data analysis.”

“We have been able to show individual countries what new customers are doing on the Web,” he adds.

Hosted services are another option. Brian Ficek, manager of e-commerce at Northwest Airlines in Eagan, Minn., went through a two-month evaluation of five different offerings before selecting WebSideStory’s Hitbox Enterprise.

“We needed a better way to track the success of our WorldPerks Mall site, which we launched in November 2000,” says Ficek. He particularly likes Hitbox because the analysis is available in near real time. “We went from nothing to being able to see things like traffic patterns, URL paths, and time spent on the site.”

Real-time response in a hosted application that may be thousands of miles away may sound unrealistic, but Michael Christian, senior vice-president of WebSideStory in San Diego, says the key is the software itself. “It is proprietary, and we built it from the ground up to handle high volume. We do about 20 billion page views per month on thousands of Web sites.”

WebSideStory prices start at about US$2,000 per month and can go to several times that depending on usage. “We still think it is much cheaper than installing a solution in-house,” says Christian.

The wide array of choices indicates that the age of Web analytics has arrived, but none of this software can do the most important thing. “Before you go shopping for tools, make sure you understand your business, and what you expect to get from any new data analysis,” says Macys.com’s Beberman.

Expedia’s Zimmerman puts it even more strongly. “If you really understand the information you want from your data then you can do a lot with some basic tools. And, if you don’t understand the business problem in the first place, don’t assume spending megabucks on new Web analytic software will be the solution.”

Analytics Challenge Data Warehousing

Guy Abramo, chief strategy and information officer at Ingram Micro, the giant technology distribution firm in Santa Ana, Calif., bucked conventional wisdom when he implemented Web analytics a little more than two years ago.

“We had completely redesigned our Web site,” explains Abramo. “The old one just couldn’t handle the volume. In addition, we wanted to start analysing customer behaviour on the site.”

Abramo looked at some of the newer Web analytics tools, but decided to go with software from Business Objects, an established business intelligence vendor in San Jose, Calif.

The reason was simple. The last thing Abramo wanted was a silo of information separate from the data collected offline. “It makes no sense for us to just analyse Web behaviour,” he says. “We also need to look at telephone, EDI (electronic data interchange), and XML-generated information.”

Business Objects may have been a good choice for analysis, but it meant Abramo still had find a way to get the data into the firm’s data warehouse.

“Once the data is in a data warehouse, you can analyse it with our tools,” says Dave Kellogg, senior vice-president of marketing at Business Objects. “The main challenge with Web data is getting it into the warehouse. That’s essentially an ETL (extraction, transformation, and loading) problem.”

Kellogg goes on to say that Web log files are “incredibly arcane,” and that Business Objects is happy to leave the ETL work to others. Abramo could have used a Web analytics tool to do this, but in order to gain control and flexibility, he chose a different route.

“We built our own system,” he says. “It took about six months in the second half of 2000.”

Abramo is pleased with the results, but it was a substantial effort, and well beyond the means of smaller companies. Ingram, for example, brought in outside consulting help from KPMG.

The technical issues were minor compared to the cultural ones, adds Abramo. “Getting the data warehouse up and running was a technical challenge, but the hard thing was answering questions like, ‘What data do we want to analyse, and what is the insight we hope to get?'”

Exit mobile version