If an Advertisement Runs Online and No One Sees It, Is It Still an Ad? Empirical Generalizations in Digital Advertising
- Measurement errors caused by Internet users deleting their cookies are compounded by other cookie-related problems: the same person using multiple devices (e.g., a person using both a work and a home computer and, therefore, having more than one cookie per person for a given Web site or advertising campaign) and different people using the same computer (resulting in more than one person per cookie).
- Solely using cookies to target digital advertisements to specific demographic and behavioral segments typically results in inaccurate advertisement-delivery information due to targeting errors caused by cookie deletion; multiple devices per user (which leads to multiple cookies per individual); and multiple users on a given computer (which makes it difficult for an ad server to know which person is using a "cookied" machine at any point in time). Additionally, targeting accuracy declines with an increase in the number of demographic variables used to describe the target segment.
- Approximately one in three delivered digital-advertising impressions never have the opportunity to be seen (i.e., are never visible to the end user), with this viewability rate varying dramatically by site.
- The number of advertisements delivered next to brand-unsafe content is not substantial, but the authors of the current study believe that the absolute number of consumers impacted is significant.
- On average, geo-targeting of digital campaigns at a country level using IP addresses is quite accurate, with only 4 percent of advertisements in the United States and 7 percent in Europe falling outside of the intended geography. On an individual campaign basis, however, error rates can be high, reaching 27 percent in Europe and 15 percent in the United States.
- Non-human traffic, including fraud, is a significant challenge for accurate digital-advertising delivery, and it is not adequately eliminated by industry blacklists of known robots or fraudulent operators.