You're thinking of going to a movie tonight, but which one? If you're like a lot of Web users, you might check out a Web site that will give you reviews, previews, and even clips of current films to help you decide. And if you're one of the thousands every day who visit Hollywood Online, one such site, you'll probably do it on your lunch hour, whether you live in New York, Chicago, Los Angeles or anywhere in between. If the World Wide Web is going to become a profitable medium, site developers, Webmasters and advertisers need to know just that sort of information.
In the past 18 months of explosive growth, the Web increasingly has been seen as a business tool. With its flashy graphics, sounds and video capability, it has rapidly become more than a way for computer geeks to download pictures of Pamela Anderson Lee or songs by AC/DC or They Might Be Giants. Top sites receive thousands of visitors a day, and market forecasters predict that visitor counts at today's Top 10 sites will be average in another 12 to 18 months. The Web already is a factor in the worldwide electronic economy. Website developers are looking for advertisers to support their sites, and marketers and advertisers are looking for top sites that will net them exposure and customers. But the bottom line in business is à well, the bottom line.
Other media ù newspapers, radio, television, billboards ù all have various ways to measure their audiences, to tell advertisers what they're getting for their investment. Businesses advertising on the Web need to know not just the number of ôhitsö a site receives, but demographic information about the people behind those hits: who they are, where they live, what kinds of products they're interested in, when and where they are most likely to access the Web. And how to keep them at a site long enough to see and interact with their advertisements.
Andromedia of San Francisco is one of a number of companies developing software systems to provide just that kind of information. AndromediaÆs Aria software system, released commercially in December 1996, is a powerful tool for companies and users to measure the success of their Web presence. Among the Web sites already using the Aria system are Hollywood Online, one of the top commercial sites, and New Media's Hyperstand. Andromedia is a Sun Microsystems developer, and in the fall of 1996 inked a deal with Sun that includes beta-testing Aria at www.sun.com, assistance through the Sun Catalyst program, packaging an Aria demo on Sun CD releases, and price breaks for other Sun developers. Aria also is working with Sun's Enterprise Network Services Group, to evaluate Aria on Sun's intranets.
Hollywood Online is a site that might be visited by the hypothetical moviegoer at the beginning of this article. The site gets about 700,000 hits per day, approximately 3,000 visits per hour. Gregg Hiscott, the Website manager, expects traffic to at least double, and possibly triple by the end of 1997. Currently, Hollywood Online consists mostly of a collection of multimedia files from current or recent films for movie fans to view. A number of different site enhancements are in the works, including a database of local theaters across the country, so people can find out what movies are showing nearby.
In mid-January, Hollywood Online had about seven advertisers at any one time, including Ford, AT&T and Microsoft. ôThe three pieces of information that are most interesting to our advertisers are traffic levels ù visits and hits; browser types used by people that are visiting the site; and how people are navigating through the site,ö Hiscott says. ThatÆs the sort of information the Aria Aria Web Recording and Reporting system is designed to capture.
The Aria.recorder is a multi-threaded, multi-process daemon. It accepts information from the monitor, buffers it, and turns the information into objects that it passes to the Aria.reporter. The reporter gives information, both real-time and historic, to the site administrator in the forms of graphic charts and tables.
Two of the most important features of the Aria system are its flexibility
and scalability, says Scott Capdevielle, AndromediaÆs chief technology
officer and one of two co-founders. The system is flexible, so that it can
be tailored to capture and present the information thatÆs important to each
client; and scaleable, so that it can handle peak loads that fluctuate
widely on a daily basis. The key is its object orientation, Capdevielle
says. Information about each Web transaction is stored as an individual
object, not as part of a relational database.
ôWe can show you the state of this object at any point in time,ö he explains, ôso you can request an object and we can feed that to you very quickly. If you have a relational data base and want a report on a set of information, you have to run a query, and that can take a long time, particularly on high-end sites with up to a million hits a day. We process the data ahead of time, so you just grab that object and get the same information in a fraction of the time.ö The system reads all kinds of information from the Web surferÆs computer: the visitorÆs IP Number, where he surfed in from, and factoids from cookies, such as whether heÆs visited the site before. It also keeps track of how each visitor navigates around each site: does he or she click on the ads? Which graphics seem most interesting? Which articles draw the most readers?
The system can monitor individual visitors to a site, and if theyÆve visited before, could even choose from among different ads available, depending on which ones the visitor is likely to respond to, Capdevielle says.
That brings up some ethical issues that Capdevielle says are up to each Webmaster to grapple with. ôI think itÆll definitely get tougher to distinguish information from advertising,ö he admits. ôMaybe they wonÆt consider it an advertisement ... If they click on a site and see, hey, this is the type of stuff IÆm interested in ... you can feed them an ad and they wonÆt perceive it as such.ö Should Websites use disclaimers that say ôpaid advertisement,ö the way newspapers do when an ad looks too much like a news story? ôThatÆs going to be more of a Website decision,ö he says. ôItÆs up to the siteÆs editors to determine what is ethical and what is not. WeÆre just trying to give people tools so that they can provide the user with a more tailored experience.ö Gregg Hiscott has learned some interesting, if not titillating, facts about Hollywood OnlineÆs visitors. ôForty percent of our traffic is from overseas, which is very interesting to us,ö Hiscott says. ôAnd on our site, 30 percent of the traffic is using
Internet Explorer, which is more than at most sites.ö Peak use at Hollywood Online is at noon in each time zone, and traffic falls sharply during TV prime time, 7 to 10 p.m. Where visitors enter his site from ù Yahoo!, AOL, or some other link ù and what they do once they arrive is important to Michael Lynch, assistant editor and Webmaster for New Media magazineÆs Hyperstand site. Hyperstand bills itself as ôthe magazine for creators of the digital future,ö people who develop Web sites, CD-ROMs, kiosks and other multimedia vehicles. ôOur site is unique in that itÆs got a lot of new content daily,ö Lynch says. ôWe post three or four stories, five days a week, and we have about 18 advertisers who use the site.ö Hyperstand beta-tested the Aria software and provided many suggestions about usable features. Aria allowed Hyperstand to get rid of its sign-in page, Lynch says. ôThe whole idea of signing in was to get demographics that we could show to potential advertisers,ö Lynch explains. ôBut signing in wasnÆt very efficient and didnÆt work for us. WeÆre discovering that people on the Web donÆt want to give information about themselves without an immediate promise of something in return ù they want to see content up front. Signing-in was scaring away too many people; when we got rid of the sign-in page, our traffic increased.ö And Aria gives much more information about users than is captured by sign-ins. ôWe can go in at any time and pull an ad up, and it will tell us a bunch of statistics in a graphical format that pleases upper management,ö Lynch says.
The Aria.recorder generates HTML-based reports that can be called up by anybody on the server, not just the Webmaster. ôIf I want to know right away how many hits the site has had this morning, I can find out immediately through the browser,ö Lynch says. ôAnd I can have editors look at it to see how their stuff is doing, and the CO of my company can look at it. They donÆt have to come and bug me for the information.ö Hyperstand currently is running on a SPARC 5 with 64 mg RAM, and handles an average of 60,000 hits daily with ease. ôThe code is exceptionally well written,ö Lynch says. ôItÆs real fast, and there was barely even a glitch when we installed it. That was a big thing for us. We now have a bigger, faster machine, but at the time we needed it to be efficient, and it sure was.ö The systemÆs ability to grow with the Web site is important, too, Lynch says. ôWeÆre hoping and expecting by mid-summer to be getting about 150,000 hits a day.ö
Among the information Aria has given Lynch is that Hyperstand is busiest on Mondays ù about 20% busier than any other day of the week. Most visitors ù about 75% ù use Netscape Navigator, with another 20% on Internet Explorer. Activity on the site matches business hours across the time zones, but it remains fairly active until 2 a.m. Pacific Time, then picks back up at 6 a.m., which is 9 in the morning on the East Coast. During one recent week, international visitors from the site included 275 from Canada, followed by Australia, Japan and Sweden, with The Netherlands and Israel close behind.
ôWe have a lot more diverse audience than I thought we did,ö Lynch says. But mostly, Hyperstand uses Aria to decide which features at its Web site are successful. ôWhen we go to post new articles every weekday, we look at these reports and decide whatÆs working and whatÆs not,ö he explains. ôIt drives our content; if itÆs succeeding, it stays up and if itÆs not, we bump it down to lower page.ö
Hollywood Online started out with a single Sun SPARC 10 server, but in mid-1996 because of volume went to three clustered Integra Sun-clone servers using Netscape server software. They first hooked up the Aria software to only one of the three, but in mid-January added the other two. This scalability was an important feature for Hiscott. ôI didnÆt want to put it on all the machines at once in case something went wrong,ö Hiscott explains. ôWeÆre comfortable with it now, so we're moving forward. The way itÆs set up, we could go to a four or five-unit cluster and distribute the load that way.ö Hiscott cautions that site administrators may not want to count exclusively on installed net tracking software. ôIÆm not sure how much credence an advertiser should put in these numbers, because theyÆre self-generated,ö he says. ôItÆs important to have some outside source verify the traffic that you're telling people that you're getting.ö To that end, Hollywood Online also contracts with another firm to independently monitor the siteÆs traffic. ôThatÆs important, because people are paying to get the exposure.ö
One Web site where advertisers can get a lot of exposure is Pathfinder, the gateway to all of Time WarnerÆs on-line products, including People, Time, Sports Illustrated, Fortune, Money and other magazines, a total of more than 80 ôcontent partners.ö With millions of hits per month, Pathfinder was a perfect place to test AriaÆs capabilities. Michael Coble, head of marketing analysis for Pathfinder, met with Andromedia representatives in mid-1996 and beta-tested the system. ôI've worked with other members of our marketing team to write a number of programs that generate weekly reports,ö Coble says. ôWeÆve been developing this stuff in-house for over a year now, and we realize that somebody is going to overtake our ability to produce new tools. I think we'll probably buy a third-party tool at some point.ö
Scale is the main concern at Pathfinder, which contains 180,000 distinct URLs and generates a huge number of pages every week. ôWith the breadth and width of content and usage that we have, most of the products out there are not geared to a site our size. Some products can handle the volume of the data, but fall down with the breadth; others can handle a range of different areas, a whole bunch of different compartments, but the amount of data bogs down their system. ItÆs a very challenging problem for software companies. I'm confident that some company out there is going to be able to do it, but I'm not sure if todayÆs products will do it.ö
Andromedia's Aria has ôthe right mix of ideas. It's a fairly robust product and it probably can handle the load,ö but Pathfinder has not yet decided which tracking software it will eventually use. Michael Lynch of New Media is a believer. ôThis is an essential business tool for anyone who wants to do business on the Web,ö Lynch says. ôSo many sites just sort of blindly look at raw hits per day, and they donÆt get down to any kind of meaningful details. This (Aria) is really powerful stuff. ItÆs been a big help for us.ö