How Not to Market on the Web

Image courtesy of Flickr user bejealousofme.

Marketers have increasingly been using the Web as a vehicle to reach customers, but companies have sometimes struggled to figure out exactly how to transmit information about a product relevant to a consumer without breaching that individual’s personal sense of privacy. It’s a fine line, and some businesses have successfully straddled it — witness the effectiveness of the sponsored text ads that appear on Google’s search results — while numerous others have failed. Some recent research provides some guidance.

The study was conducted by Avi Goldfarb, associate professor of marketing at the Rotman School of Management at the University of Toronto, and Catherine Tucker, an assistant professor of marketing at the MIT Sloan School of Management. They described their findings in an October 2009 working paper entitled “Online Display Advertising: Targeting and Intrusiveness.” Goldfarb and Tucker had access to a database that tracked almost 2,900 different online ad campaigns for a wide range of products. The data, collected by a media marketing company from 2001 to 2008, contained the survey responses of a total of more than 2 million consumers — an average of about 850 for each ad campaign. For any particular product, half the survey participants, selected at random, were exposed to the ad and the other half were not but did visit the Web site where the ad had appeared. The study participants were asked a number of questions, including their likely intent to purchase the specific product in question.

The researchers’ findings? “Advertisements that match both Web site content and are highly visible do worse at increasing purchase intent than ads that do only one or the other.”

In particular, Goldfarb and Tucker investigated the effectiveness of two types of ads: those that complement the content of a Web page (such as an ad for a hotel chain that is displayed on a travel Web site) and those that are highly visible (including pop-ups and full-screen ads, video and audio ads, and ads that move across a Web page). The results show that, with respect to a consumer’s stated intent to purchase a product being marketed, complementary context-based ads are three times as effective as regular online ads, and high-visibility ads are two times as effective. Of course, people do not always do what they say they’ll do, but Goldfarb and Tucker were interested mainly in consumers’ relative intent to purchase and not necessarily their actual purchasing behavior.

Those findings should come as no great surprise to online marketers, but here’s where things start to get interesting. Goldfarb and Tucker also looked at the combined use of the contextual and high-visibility approaches and found that it was actually counterproductive. “Advertisements that match both Web site content and are highly visible do worse at increasing purchase intent than ads that do only one or the other,” they observe. The lesson is clear, concludes Tucker: “Don’t try to both complement and compete with a Web site’s content.”

But is combining the two approaches simply a case of “gilding the lily,” or is some other factor at play? By analyzing the data further, Goldfarb and Tucker uncovered a possible link to privacy issues. Consider, for example, someone who is searching a financial Web site for information on personal bankruptcy. She might find it helpful to see small, plain ads (similar in spirit to the sponsored text links shown on Google searches) for credit cards touting special offers for debt consolidation. But what if she were greeted instead with large pop-ups and videos for those same products and services? Instead of being receptive to them, she could easily be annoyed — if not offended — by the perceived intrusiveness.

Two important observations support Goldfarb and Tucker’s contention that privacy issues are a factor behind their results. First, survey participants who refused to divulge any information about their income — and were thus likely to be especially concerned about their privacy — tended to have the most adverse reactions to ads that were both contextual and highly visible. Second, the negative effects of combining context and high visibility were generally strongest for products for which privacy was of greater concern, such as for pharmaceuticals and other medical treatments. In other words, a marketing campaign for a detergent might be able to successfully deploy a contextual ad that is highly visible, whereas an ad campaign for a contraceptive might not. Thus, the larger lesson for online marketing is essentially the same as for marketing in any medium: You’ve got to understand your customer as well as your product.