Response rates to banner ads have decreased dramatically, paving the way for two strategies for online advertising. One is ads incorporating audio and video to make them more appealing, and the other is Google developing Adsense; displaying simple text ads that are targeted to content. This is the summary of an article written by Avi Goldfarb and Catherine Tucker, which investigates both strategies and their effectiveness for online advertising, and how consumer perceptions of privacy and intrusiveness affect them in such. You can get the PDF of the behavioral targeting article here: Online display advertising..
Data gathering
This study incorporates a large randomized field data from 2,892 distinct web ad campaigns of various products, shown on many different websites. On average, there 852 survey takers per campaign. Furthermore, consumers are randomly exposed to the ads or not, and campaign lasted for 55 days. Furthermore, respondents answer online questionnaires with five-point scale as to what extent a respondent is likely to buy a certain product, and whether one recalls the ad. Income, age and number of hours online are also asked for regression purposes.
Contextually targeted and High visibility advertising
There are 3 kinds of Banner ads compared in this study: 1. “targeted” and page relevant where product category matches website category, 2. “obtrusive”, using video, pop-ups, etc., 3. and both targeted and obtrusive. Comparing groups exposed to the one among the three ads versus groups not exposed at all (control), the 3 types of banner ads are evaluated to see which is most effective.
Methods
Effectiveness (purchase intent) per individual is measured as a function of contextual targeting and high visibility. Estimation is done by converting effectiveness function into linear model of visibility, which features advertisement-web and viewer characteristics. Furthermore, the random nature of data allows for estimation of ad exposure effect through a difference-in-difference equation, which features effectiveness and a control for possible demographic disparities among ad exposed group and control group. A direct estimation is used because of the random nature, experiment-like setting for data collection. A groups’ purchase intent is investigated to find whether the difference between control and exposed groups relate to ad visibility and ad-website content relations.
Combination nullifies individual success
Results show that increase in purchase content is shown for targeted and obtrusive ads, where contextually-targeted ads are slightly more effective. Surprisingly, however, combining both techniques nullifies the success each has separately; it is significantly negative. This result is more pronounced for categorically “private” products, and private-minded people.
Highly visible ads have higher recall than contextually-targeted ads. Combining two techniques are shown to not have significant effects on ad recall. Thus, the negation effect must be due to resulting influence in customer behavior. Results suggest that by combining both techniques, consumers feel that advertisers are putting a more conscious effort to manipulate them. Targeted ads are perceived as useful, but combined with obtrusive techniques, they become perceived as manipulative; a negative effect on how consumers perceive the advertised products.
Privacy as the primary factor
Negation among consumers is increased when both are implemented because of privacy concerns. Targeted ads are informative, but they also raise privacy issues. Privacy entails prevention focus among consumers, such that they become more sensitive to presence or absence of negative outcomes. Furthermore, prevention focus makes consumers more wary of manipulative intent. Ultimately, it is privacy concerns that primarily cause negative impressions on consumers when both techniques are combined. In this study, privacy is stratified when users check the box “I prefer not to answer that question” for income, segregating privacy-minded people from respondents. Privacy-minded people affirm highly visible ads, as long as they are not contextually targeted.
Other results
This study also clarifies online advertising through its extensive data gathering which allows for more generalized conclusions on the success of online advertising. One conclusion, for example, states that there is 3 to 4 percent boost in stated purchase intent through online advertising, although banner ads are shown to have small effect on consumer’s likeliness to make purchase and ad recall.
Implications
The results explain the unexpected success of Google Adsense, with ads that are not visually appealing but wouldn’t be as effective if made into such. These results also have relevant economic implications. $664 billion is spent each year for ads that are both obtrusive and targeted. Advertisers could spend 5.3% less and still maintain performance of their ads, which are no more as effective as standard banners. Another implication is that government policy regulators, on use of browsing behavior data for advertising, should weigh the two strategies carefully in light of viewers’ demand for privacy and reduction of manipulation.