Engage AudienceNet Brings Users the Ads They Want To See

  • Written By: D. Geller
  • Published: December 1 1999

Product Background

AudienceNet is Engage Technologies' (Nasdaq: ENGA) first significant application of their core personalization technology. Developing this technology has been the focus of the company ever since its founding by David Wetherell, CEO of CMGI (Nasdaq: CMGI) and Engage's CTO Daniel J. Jaye. Their purpose was to apply what they knew about direct marketing to the Internet. In the process they developed data mining technology which was sold to RedBrick (now Informix) in 1997, although Engage retained the rights to use the technology.

AudienceNet is a product aimed at advertisers and advertising agencies. Advertisers would like their ads presented to people who are likely to respond to them. Response is measured primarily by the clickthrough rate - the ratio of the number of people who click on an ad to the number of people who see it. Clickthrough rates very depending on the type of site and the types of ads, but a recent Nielson/NetRatings report of major sites gave an average clickthrough rate of 0.51 percent. Sites with specialized content may see clickthrough rates around two percent. Advertisers would like to present their ads to people who are more likely than average to click through to get more information, and will pay for higher clickthrough rates. For example, as reported by InternetWeek in December 1998, Infoseek raised its clickthrough rate from 1.5 percent to 10 percent by using information about visitor history. Advertisers using this service were willing to pay three times the usual advertising rate.

AudienceNet offers exactly that kind of targeted ad serving. It has built up detailed profiles of 35 million web surfers. The profiles were built up through data collected from an initial base of four hundred sites. Since the data are based on user behavior and are stored in cookies, they are do not actually identify users. However, they can be the basis of ad serving to these users. Some studies have already shown that a person who has clicked through on a particular advertiser's ad once is more likely to do so again. The equivalent effect for ads on different sites that share a common interest may not yet be verified by academic studies, but is a natural hypothesis.

If a website chooses to participate in AudienceNet, then the machines of any of its visitors can be interrogated by Engage to see if they contain the AudienceNet cookie. If so, their profiles can be looked up based on the identification stored in the cookie. Each site participating in the network has a cache containing the user identification numbers and some other data. The site (or its ad network) uses this file to decide whether to serve its own ad (for those that do this) or to redirect to AudienceNet for an ad. This is at the discretion of the site, which in most cases will make the decision based on deriving the most revenue or meeting its local impression contracts.

A user profile is based on more than 800 interest categories, as well as demographics and geographics. For example, there is a Sports category. Sports has subcategories for various classes of sports, for individual sports, and for other factors that, for example, distinguish participants from spectators. Any member of the network can add additional categories to the mix; so, if a site wanted to serve ads to people based on their favorite ice cream flavors, it could do so. Once these ice cream flavor categories have been added to the profile, other sites could also use them.

The process of choosing an ad uses a proprietary algorithm that takes into account recency of each data point, the frequency with which the user has shown interest in the variety of categories, and the duration of the users' relevant web page visits. Some of these data are mined from web logs of the participating websites, although sharing of logs with Engage is not a requirement of belonging to the network. While the total package of categories and other information allows sites a great deal of discrimination in serving ads, Engage expects that many will rely on predefined clusters of characteristics; these are general demographic groupings like Active and Athletic Adults, Affluent Individuals, Business Professionals, College Students, and Heads of Households.

Engage hired an independent advertising research company to evaluate the effectiveness of using profiles for serving ads. In this study profiling was compared with not profiling. Overall, it was found that serving ads to prequalified surfers can raise the clickthrough rate by 28%.

Security is an important concern for AudienceNet. Engage claims that a good deal of effort is taken to make sure that individuals cannot be identified by their profiles. As part of this effort, the company has a "Knowledge Ontology Review Board" that studies the effect of adding new categories; ontology is a branch of metaphysics dealing with the nature of reality.

AudienceNet can be marketed to anyone who serves ads. An advertising agency or an ad network could use it to improve the clickthrough rate for their advertisers. Or, an individual site that does its own ad serving can use it to enhance the response rates for its advertisers.

The market for Internet advertising services has been shaping up as a conflict between Engage's parent company, CMGI, and its archrival DoubleClick (See TEC Technology Research Note: "'Ads are us', boasts CMGI" October 7th, 1999). A distantly related DoubleClick product, DoubleClick Dart, was released a year ago and other targeting products will certainly follow. With ad sales continually rising (See TEC News Analysis article: "Advertising Revenues Grow and Grow but Slower and Slower" October 6th, 1999 ) there will be more and different products introduced for delivering targeted users to advertisers.

The potential power of AudienceNet for delivering targeted product content to users may outweigh its strength in the advertising arena. Where Amazon can now deliver to its return visitors a list of recommendations based on their previous purchases, Engage will be able to do this across otherwise unrelated sites in its network. So, a person who searches for information on Ireland in CMGI's AltaVista engine might see ads for vacations to Ireland at a travel site, a "special" on Irish linen at a home shopping site, and information about soda breads or scones at a bakery site.

A site using AudienceNet is not restricted to advertising. The same profile data could be issue for targeting page content to a user. So, different users coming to the home page of a site would see content tailored to their interests as deduced by AudienceNet.

AudienceNet is sold to advertising agencies and advertisers on a CPM (cost per thousand impressions) basis. The rates range from $30 to $75 depending on the target audience. To deliver these ads, Engage purchases inventory (impressions) from websites and advertising networks, also at CPM rates. Some of Engage's largest partners are websites and advertising networks that are also owned by parent CMGI.

Joining AudienceNet as a website is not difficult, but does take time. The company estimates that this will normally be a six-week process. The website pages have to be mapped to specific interest areas, which involves analysis by Engage's ontologists, and then the ad tags (See TEC Technology Research Note:"How to Serve an Ad: Better to Buy Than to Build" October 23rd, 1999) on each page must be modified. AudienceNet launched on October 10, but as of the end of November had not yet "launched" any sites beyond the original 400. The company explains that the first launch involves enabling all of the websites belonging to CMGI's AdSmart and AdForce networks, and this combined launch was always understood to require more elapsed time than would a single website.

Product Strategy and Trajectory

AudienceNet is Engage's first application of its targeting technology. The second will be in E-commerce space, although the company is not releasing details at this time. However, it is a sure thing that the same database will be used, and added to, for that and future applications. This will yield what Engage calls "convergence."

AudienceNet is primarily a business-to-consumer product, and Engage has limited plans to develop vertical markets. The company has been flirting with a specialized version of the product called AutoKnowledge, which would be completely focused on automotive sites, with the end goal of catching surfers during the window of time that they are planning to buy a car. They have begin to work with AutoSite to explore the concept further, and are wooing major car manufacturers.

While vertical markets are generally seen as excellent business opportunities, Engage can largely reap the benefits of vertical marketing without doing much work. If a site that has the potential of anchoring a vertical business joins AudienceNet and adds to the profile database characteristics specific to that vertical market, it will be an easy sale for Engage to attract other companies in the same market. Vertical markets can thus grow without any planning on Engage's part, purely because other ventures see opportunities to target those niches. Both Adam Smith and Charles Darwin would be very pleased.

Another recent announcement by Engage gives additional clues to the strategy for AudienceNet. Engage owns a piece of a startup called InsightExpress, which will provide customized market research to small and mid-sized sites. The company will start with the ability to collect data on about 25 demographic groups, but will easily be able to provide much narrower (and more valuable) targeting once it integrates with AudienceNet.

AudienceNet itself is expected to grow to about 1000 sites by the end of the year 2000. We estimate that the sites in this network will be serving 25 billion ad impressions each month by that time. While there are no estimates for the percentage of impressions that will be served through AudienceNet, a simple calculation is instructive. If 1 percent of those 25 billion ads go through AudienceNet at an average CMP of $50 (i.e., five cents per impression) Engage would gross $12.5 million per month. Since a large percentage of their inventory payments would go back to other CMGI companies, and since their largest website is CMGI's AltaVista, the net result would be that Engage does very well, and CMGI, in technical terms, makes out like a bandit.

Product Strengths

This is a product with promise. We believe that the future of Internet advertising and E-commerce is in targeting. Engage has a very sophisticated and powerful product with applications far beyond advertising. It is a particularly strong product because it plays with any other product. Even users of competitive ad servers can take advantage of AudienceNet. This puts considerable pressure on DoubleClick and its new subsidiary NetGravity to develop a competitive product.

Product Challenges

With no sites beyond the initial 400 used to develop the product having been launched, there is as yet no user experience to draw on. We do believe that the effort of changing ad tags will be a deterrent to some sites. We also project that beyond general consumer sites, the effort of doing page analysis and working new categories through the Knowledge Ontology Review Board will be a diversion that many websites with targeted audiences outside of AudienceNet's established characteristics will be slow to undertake. To be fair, handling sites of that nature is not within Engage's strategy for AudienceNet at present, but as the product's share of the consumer sites reaches its maximum penetration in, we estimate, two to three years, these barriers will be more important.

Vendor Recommendations

Engage must pay attention to the speed of ad serving. A site using AudienceNet can expect to see increases in the time to serve an ad. In theory, with no bandwidth limitations either on the Internet or at the various servers involved, any slowdowns would be negligible. But the Internet itself is beginning to move a little slower (See Tec News Analysis article: "Is There a Magic Pill for Web Performance Problems?" October 15th, 1999), using AudienceNet takes more processing on a site's own servers than not using it, and Engage's own servers are likely to see heavy usage as AudienceNet becomes popular. Speed of service and uptime are crucial for websites, and Engage must stay ahead of the growth curve.

We believe that extending its database to better serve narrow markets will be key for this product by the end of 2000. The current work on AutoKnowledge indicates that the company is already working in that direction.

We believe that there is a real question as to what Engage should do with its AdManager, purchased from Accipiter over a year ago. The product is not highly thought of, and to the extent that Engage pays attention to it as a product at all, would be better dropped, sold or spun off. Engage has a good chance to take and maintain a commanding position in targeted advertising and to parlay that into a similar position in e-commerce. We believe that playing around with a second rate ad server is an unnecessary diversion.

User Recommendations

For a website interested in increasing its advertising revenues AudienceNet is definitely a product to explore. Users working with an ad network should enquire whether the network is planning to enable AudienceNet. Users should negotiate specific level of service clauses into any agreement; such clauses specify average and maximum turnaround times for serving ads and the required uptime for the AudienceNet ad servers. Naturally, methods of measuring these need to be determined as well. We also recommend that a clear method for measuring ROI be determined, so that the site can evaluate the value of AudienceNet. In order to do this, the site must determine how it will track the uses of AudienceNet; this may require access of some form to Engage's records, which the company seems anxious to assist with.


comments powered by Disqus