Business Models Based on Advertising
My presentation will focus on the business models for advertising and commerce on the Internet, still viable despite the general pessimism about the way things are going on the Internet these days. All of the big players have had problems. But there will be a workable business model; the question is how to figure out what it will look like, and how those models can be put to use in protecting kids online.
16.1 COMPARISON OF ADVERTISING MODELS
Advertising will continue to be a significant part of Internet business models, despite what you may hear. There are four basic models for the sale of advertising. The most common models are cost per impression and revenue share, although cost-per-click and cost-per-acquisition deals are gaining in popularity.
Cost-per-impression (CPM) deals are usually experienced as banner ads while you surf the Web. You go to a site such as Excite, and the banner ad is presented to you as part of the page. This is still the bread and butter of the industry, the way most sites generate their major revenues, but it is in serious trouble. Every major Internet portal has seen a serious decline in revenue coming from advertising, and offline businesses dependent on advertising revenues have seen similar thinning.
When banner ads first came out on the Internet, people clicked on them 15 to 20 percent of the time, because nobody knew what they were and everyone was trying to bounce around and figure out this exciting new medium. Things have stabilized now to below half a percent in terms
of click rates for a basic banner ad. This has been a disaster in terms of convincing offline advertisers to move some of their budgets online—an effect that everyone has seen on the Nasdaq. In talking about these low clickthrough rates, I am referring to run-of-the-mill ads; I will discuss targeting later.
Because of this lower perceived effectiveness, a few other models are gaining greater prominence, such as “cost per click.” Instead of paying for the presentation of your product in a banner advertisement, you pay for the actual clickthrough on the ad. This is less popular and more difficult to negotiate, because Internet networks are reluctant to accept these deals. They say, “If you pay us only on a conversion, on a move, on a redirection to your site, then we cannot forecast what the revenue from this deal is going to be.” Advertisers (i.e., ad space owners) are looking for guaranteed payments—generally targeted banner ads.
Cost per lead is a slightly different model. A lead is a conversion so that someone agrees to provide a service or to accept to further direct mail or e-mail—roughly analogous to the response card in a magazine that says, “Circle here for more information.”
The revenue share, as I mentioned earlier, is also a popular type of deal. The problem with revenue share deals is that you are depending on actual commerce to pay the bill. If there is no transaction at the end of the day, then revenue does not flow back to the advertising presenter, who is thus not happy about the way the ad space has been used.
16.2 PORTALS, ADVERTISING NETWORKS, AND TARGETING
In discussing advertising-based business models, it’s important to note that the big players—America Online, Excite@Home, Yahoo—sell many of their own ads but not all of them, which is important. We have an ad sales force that spends a lot of time going to large advertisers and saying, “For x million dollars, you can get this many impressions on our network. They will be on these particular channels on the network.” Smaller players and some of the big ones outsource that type of ad sales to ad networks. The biggest one is Double Click. Other large ones are MatchLogic, a wholly owned subsidiary of our company; Engage; and 24/7 Media. These are third-party networks that operate on a variety of sites across the Internet. Double Click has 2,500 to 3,000 sites from which it serves ads across the Internet. Match Logic has about 1,000 sites. A big concern is the placing of cookies on user’s browsers and computers, to track behavior across those different sites.
Targeting is, in many ways, the Holy Grail of the industry. Most ad targeters use profiles based on your behavior across a number of sites within an hour. If you visit 10 or 20 of the 2,500 sites within a Double
Click network, then you get scores associated with each site indicating male or female, likely age, presence of children in the household, and other things like that. Once that profile is established, when you visit a site where ads are served by Double Click, it will read the cookie on your browser and say, “This person is probably between 24 and 35, probably has kids in the household, is probably female, and may have an interest in X.” Then you get served an ad that Double Click has sold to an advertiser that matches this demographic profile.
These are usually anonymous, which is an important point. This is one of the biggest sources of confusion and discussion in the privacy arena. The Federal Trade Commission (FTC) took action against Double Click because the company had plans to start personally identifying without user permission. As it turned out, they never did that and the FTC inquiry was properly stopped. They had planned something that probably would have violated the law and it would have been a false incentive advertising practice. But they did not do it.
All of this happens because of the need to drive the click rates up, to actually reach the people that you are trying to target. To the extent that these things are done anonymously, they are, arguably, wonderfully beneficial—and one of the business models that will work. If you can get to the types of people that you want, then it is much easier to present to an advertiser who has x number of dollars to spend to reach this audience, and say, “You should pay this rate, this CPM or whatever, to get these people. Because we know, based on the technology that we’ve set up, that we can get to people who meet these characteristics.”
A number of companies have tried to generate revenues this way. I am sure that a number of the big networks are very involved in ad targeting. This is similar to what grocery stores have been doing by giving out discount cards. The major difference is that the grocery discount cards have personally identifiable data, so that they can send you coupons in the mail.
16.3 CHOICE OF MODELS
Different types of Internet content providers favor different ad models. The quintessential example of the lengths to which some companies will go to drive traffic is that, if you end up accidentally on a porn site, you cannot even close your browser—the site just keeps showing up. Mainstream advertisers are starting to use these technologies, too; if you try to close a window, then ads pop up on a number of different sites. Without having done a full economic study of the porn industry, I cannot say this definitively, but my guess is that they will get hit with some of the same advertising doldrums that everyone else has. The ones making
money are probably the ones with subscription models. Porn seems to be one of the few things that people will pay for. The problem in avoiding the content is probably related to promo pages, which are designed to draw people in to pay for a subscription. Filters definitely need to catch those pages.
Most nonporn sites are not trying to show pictures or video, just animations and banner ads, so there is less concern about bandwidth cost in the presentation of screens. One reason why the ad networks have managed to prosper is precisely because their costs are so low.1 There is a high cost to build servers to push things out and to negotiate the first arrangements with Web sites to build them into the network. But once that happens, you can just serve it out. You added potential customer leads and lowered your customer acquisition cost by expanding your network, because you can send a cookie when a new browser visits a site that has, for example, a Double Click ad. That unique identifier will be carried across every site in the Double Click network and be registered in Double Click. High start-up cost and low marginal cost make a big difference in terms of overall advertising cost.
16.4 ADVERTISING, REGULATION, AND KIDS
There are many questions to be asked about advertising as a model for paying for software or services that would protect kids. The biggest player in filtering in the schools has now abandoned advertising despite the potential for real benefits in terms of a business model and potentially modifiable ad space that could pay for technology that would help to avoid indecent material. What drives these choices are worries about privacy. The Children’s Online Privacy Protection Act (COPPA) requires parental permission for any personally identifiable information collected
about children under 13 and thus severely limits business models that would target kids.
A number of other potential privacy laws and regulations also are coming that could affect the choice of advertising-based models for online safety efforts. One is self-regulation by the industry through the Network Advertising Initiative (NAI), part of a response to the Double Click ruling. A number of industry players, including Match Logic, Double Click, 24/7 Media, and Engage, got together to find a fair way to give people notice if we want to merge personally identifiable data with ad information. The group came up with strict permission and self-regulatory standards. They worked and negotiated with the FTC to establish these standards, which were unanimously approved by the FTC and sent to the Congress and are now in force.
In discussing the data models that advertisers use and particularly the potential effect on a childrens’ market, the meaning of “personally identifiable” is a huge issue. The question is how far you can move back up the chain to make data personally identifiable. According to the NAI, there will not be a move to make data nonanonymous without permission. If a hacker took the information and could match it geographically, then perhaps this could be done without permission, but it is difficult to get all the crumbs together and link them back to an actual person. Personally identifiable information usually is defined as information to be used to contact an individual directly—such as full name and physical address. E-mail address generally is defined as personally identifiable as well. Some interesting discussions are going on in the European Union about whether Internet Protocol addresses should be considered personally identifiable information. It is always difficult to figure out what will happen in the EU and which body is acting on which day.
Senators John McCain and John Kerry have proposed privacy legislation that would require Web site notice, which would affect potential children’s advertisers along with everyone else, in terms of fully disclosing the facts and the privacy laws. There are also a number of other possibilities. Some in the industry favor a weakening of COPPA because of its effects in cutting off under-13s from a socially beneficial communication source. Our network does not favor a weakening of COPPA. But it has a real effect on our site. We have completely cut off under-13s from e-mail and chat, because these mechanisms can be used to spread personally identifiable information, and the costs of getting parental permission and maintaining verifiable parental permission were not justified by the revenue. Kids on our network can get to the personalization features and use them, but we keep only the first name and birth date—everything else is deleted.
On privacy, including kids’ privacy, the corporate position that we have taken is that we are comfortable with further enforceable regulations saying what companies can and cannot do, as long as they are done carefully and do not forbid legitimate consumer-serving uses of data. Self-regulation, in which companies talk about their practices and expose themselves to both public scrutiny and government scrutiny for false and deceptive trade practices, will also be a major part of coming up with a privacy solution. There also will be new technology, which is the x factor. Some technologies will allow complete masking of information and covering of footsteps. This is difficult to implement. A number of advertisers will rely on the fact that people will find it difficult to use. Furthermore, not everyone wants to be anonymous at the end of the day. For instance, you want toothpaste if you run out. It is okay for most people that Webvan knows that fact because you want it to bring the toothpaste so that you do not have to leave home or worry about it. You want your refrigerator company to know when your compressor isn’t operating properly so that it can come out and service it.