“I’m not who you think I am,” said the searcher to the engine

Last week, Yahoo introduced a product called SmartAds that’s being touted as its great leap forward in targeted advertising. Some commentators are even suggesting that SmartAds may be Yahoo’s best hope of catching up to Google.

Like all targeting technologies, Yahoo’s SmartAds isn’t as smart as its boosters claim. The decisions it makes when it serves up an ad are based on historical search behavior, which may or may not be indicative of future purchasing intent. Yahoo’s SmartAds demo contains the example of a guy in Los Angeles who likes to visit online poker sites. On the basis of this past behavior, Yahoo will serve up a travel ad touting the latest airfares from LA to Las Vegas.

There are many obvious problems with this scenario. The “guy in Los Angeles” may surf poker sites because he likes to play poker with his friends, or play online poker, or perhaps isn’t even in Los Angeles although his Internet service provider may be. He may be someone researching poker sites for a term paper or a graduate thesis on gambling addictions. He may not even be a “he,” but a “she” who is sharing a machine with someone else. But none of these uncertainties will stop Yahoo’s profiling system from erroneously concluding that it’s got a viable target to serve targeted ads against.

The exponents of behaviorally targeted advertising argue that it theoretically improves the relevance of advertising, but in practice such advertising is often more annoying than untargeted advertising. The more tightly the audience is segmented, the fewer advertisers vie for each segment, which results in the user being exposed to the same ad repeatedly. Frequency caps that turn off a given ad after a given number of impressions result in a house or public service announcement ad being displayed, neither of which are at all relevant.

Another problem with behavioral and other targeting technologies is that they surround the user with a blanket of ostensibly “relevant” advertising that screens out any chance of their being exposed to messages that don’t make it through the behaviorally targeted filter, which is subject to the flaws mentioned above. Human beings have interests which are not always expressed in search behavior, and while it’s impossible to quantify the degree to which people respond to advertising messages they did not deliberately seek out, this phenomenon clearly exists. Behaviorally targeted means that fewer and fewer chances of these serendipitous exposures will occur. Essentially, the fact that I’ve never searched for Disney World will mean that Disney World will cease to exist in my personalized universe.

Finally, users are not stupid, and while cookie deletion rates have fallen in the past several years, behaviorally targeting technologies run the risk of causing a significant percentage of users to rebel once they come to the realization that they are being followed around by ads. The only reason this hasn’t happened yet is that behaviorally targeted ads still constitute a fractional component of the ads actually served. But it’s inevitable that over-aggressive advertisers will seek to display personalized ads that cross the line between relevance and active surveillance. At this point, users will no longer passively accept the idea that their personal online behavior can be mined by profit-making companies and will use any means necessary to escape from those they believe are stalking them.

To put it bluntly, the first time I see a gaudy, flashing banner ad following me around on the Internet will be the last time I visit whichever search engine or portal put it there, and I’m quite sure that I’m not the only person on the planet who feels this way.

Related Posts