Xenakis on Technology

Web Analytics Can Improve Your Marketing

Written by John J. Xenakis for CFO.com, Feb 21, 2001.

The Web can provide marketing executives with more information than they ever dreamed of. Making sense of it is another matter.

Perhaps one of the best business innovations brought on by the Internet is its ability to provide nearly instantaneous feedback on marketing campaigns. For example, once a campaign is launched, a company can measure the impact on its Web site by doing further research on the products being marketed and the people buying them. Gift catalog merchant ABC Distributing Inc. sends out millions of catalogs each month, and it uses the Internet to get fast feedback from test mailings.

"One of the biggest benefits for ABC is detecting as quickly as possible the results of sending out test catalogs," says Daniel Gudema, Web site manager for the North Miami Beach firm. "We can monitor the browsing level of the online catalog on an hour-by-hour basis, and then use that to determine within two or three days whether to send the test catalog out for a major run."

But getting detailed feedback of this sort requires accurate measurements of a Web site's traffic, an area known as "Web analytics."

"The Web analytics market is relatively crowded with 36 major suppliers, and the number is growing," says Guy Creese, analyst for the Boston based Aberdeen Group. He adds there are a number of dirty little secrets that these companies don't tell you, which means you should be careful when selecting products.

Click here to read Aberdeen's dirty little secrets of Web analytics.

ABC Distributing tracks the activities in real time on its Web site (abcdistributing.com) by means of an ASP service called HitBox, provided by WebSideStory Inc. (Websidestory.com).

In order to make this work, the company puts JavaScript commands into the HTML for each of the site's pages. When a user visits one, the JavaScript sends information about the visit and the section of the page the mouse has been clicked on (or been dragged) over back to WebSideStory's database.

"Originally we were quoted a price of $4,000 to $5,000 dollars a month for the service," says Gudema. "But now we've gone up to 20 million to 30 million page views per month on the Web site, and at that level it's $7,000 to $9,000 per month."

Even at this price, the service is well worth it, according to Gudema. "Anyone in the company can log in from the office or from home and know exactly what's going on," he says. "It's brought a lot of accountability."

GMAC Insurance Online notes that Web analytics is important for other reasons. The company has just signed a contract with Coremetrics Inc. (http://www.coremetrics.com), an ASP whose base service costs $250,000 per year, which includes some consulting.

"Strategically, we believe that this kind of tool gives our customers a better online experience, by enabling us to provide them with the services and products they want," says president Mitch White. "For example, if 50% of our customers were going to the billing screen, and it took them three clicks, we could serve that screen up in just one click."

Another reason to use analytics is to get quick marketing feedback. "A real life example is that a competitor of ours did a giveaway of a $10 gift certificate to Amazon.com," says White. "The campaign was a big success based on the click-through, because they gave away a lot of gift certificates, but unfortunately they sold only one policy."

Good real time analytics, even though expensive, can nip disasters like that in the bud. In fact, since Web-site development and maintenance can easily cost millions of dollars per year, a strong case can be made for spending an extra $250,000 to $300,000 per year for Web analytics.

Companies put off by the price tag can opt for some simple, homegrown ways to analyze their Web site's traffic.

Web servers generate "Web log files" that record visits to most sites. A company can obtain generalized analytics figures in terms of "Web site hits" or "page views" just from the size of these log files.

Several companies, such as WebTrends Corp. (http://www.webtrends.com) and Accrue Software Inc. (http://www.accrue.com), have become market leaders in Web analytics by providing software products that analyze these log files.

Server Data or Browser Data The analysis of server-side data has been around a little longer than the analysis of browser-generated JavaScript files. For this reason, most of the industry-leading Web analytics firms have grown up using server data. The server method is effective in providing data about how the Web site's servers are being used and for monitoring the growth of the Web site.

Browser-side data analysis is a newer method, and appears to be more effective in providing marketing or business data. As a further wrinkle, one company, Elytics Corp. (http://www.elytics.com), is launching a product that processes and merges both server-side and browser-side data.

But at the end of the day, a Web analytics system has to do more than just count page hits.

For example, if a user visits your site on Monday and Tuesday to do the Web equivalent of window-shopping, and then makes a purchase on Thursday, the analytics software has to encapsulate all three visits into a single experience leading to a sale. Determining this from either server-side or browser-side alone is extremely difficult.

In fact, Gudema has analyzed a number of Web analytics products, and found none that provide functionality in all four of the areas that he requires: "One [area] is the volume of traffic through the Web site; one is a system that tracks and monitors the health of the servers, and how heavily they're loaded; one is the detailed behavior analysis of people who make purchases; and the fourth is E-mail management."

Gudema requires multiple products to handle all these areas.

Privacy Issues If there's one issue that's a potential showstopper in the area of Web marketing, it has to be privacy. That presents an extra challenge to marketing executives. Analytics services have to provide enough data to satisfy the marketing aims of their clients, but not so much that they invade the privacy of their clients' customers.

"With this technology you shouldn't be collecting any personally identifying information on the user," says Michael Christian, WebSideStory senior vice president. "And sometimes you have to do some special things [in the JavaScript] to strip information out." Some of his competitors have been hit with lawsuits from users with privacy concerns.

The problem is that there's a lot of data floating around -- the user's behavior at the Web site, credit card data, purchases, demographics, and, depending on the technology, information about other Web sites a visitor has been to. A malicious vendor could combine all this information and get a detailed personal profile of each Web site visitor.

It doesn't help that a lot of Web analytics data is stored on the servers at an ASP. That means that businesses are relying upon the security of the ASP's firewalls to protect their customers' data.

Even with literally dozens of Web analytics on the market, and more being introduced every week, the technological questions are still at an early stage of being solved. But Web site developers and providers will have to make sure that the privacy issues don't get overlooked as they race to improve the technology.

Sidebar -- Web Analytics's Dirty Little Secrets

By John Xenakis CFO.com Feb 21, 2001

If only Web-analytics software worked as advertised. Before you buy a service, here are five things you should consider.

During the sales process, Web-analytics suppliers rarely mention some common problems and pitfalls, according to Guy Creese, an analyst with The Aberdeen Group. Here are some of them:

Data collection perfection is unattainable. The Internet was not designed with auditing of end-user behavior in mind. Different technologies gather different views of the user's behavior. Enterprises that depend on only one of these technologies are seeing no more than a partial view of Web-site behavior, yet getting a "total" view can be expensive or impossible.

Page views and Web-site visits must be defined. Individual Web pages can be very complex, using multiple frames, tables, and graphics, some of which are shared between pages. Collecting raw data about the elements a user has downloaded to her browser is one thing, but the enterprise's business analysts have to decide which of the elements have to be combined to form a true page view or Web-site visit.

Understanding the visitor's journey is key. Enterprises have a functional view of their Web site  One group of pages goes under "products and services," while another set belongs under "partners." However, this view clashes with the process orientation of visitors who are browsing, buying or complaining, involving several functional areas. The key to understanding visitor behavior is to study visitors' journeys across the Web site rather than simply count the waypoints.

Affiliated business processes are not always easy to implement. Enterprises typically underestimate the amount of work required to integrate Web analytics into business processes, just as they initially underestimated the work it takes to maintain a Web site. For example, an enterprise must agree on the amount of information it will gather. Enterprises that track too much risk antagonizing privacy watchdog groups. If they track too little, the corporation won't understand its customers.

Beware of corporate politics. Web analytics systems initially tracked activities of particular interest to IT managers concerned with predicting Web-site growth. But with the shift toward business metrics, marketing managers who are currently excluded from seeing the data should now have access to it. Click here to return to main story on Web analytics.

(This is a modified version of an article that originally appeared on Feb 21, 2001 on CFO.com at this location. )