Tracking Fraudulent and Low-Quality Display Impressions

Abstract

Display advertising is traded in a complex market with multiple sides and intermediaries, where advertisers are exposed to several forms of potentially fraudulent behavior. Intermediaries often claim to implement measures to detect fraud but provide limited information about those measures. Advertisers are required to trust that self-regulation efforts effectively filter out low-quality ad impressions. In this article, we propose an approach for tracking key display impression metrics by embedding a light JavaScript code in the ad to collect the necessary information to help detect fraudulent activities. We explain these metrics using the campaign cost per thousand (CPT) and the number of impressions per publisher. We test the approach through six display ad campaigns. Our results provide a counterargument against the industry claim that it is effectively filtering out display fraud and show the utility of our approach for advertisers.

Notes

1 For example, the International IAB/ABC Spiders & Bots List and the Trustworthy Accountability Group (TAG) list made available by Google.

2 See https://iabuk.net/blog/brand-campaigns-benefit-from-contextually-relevant-placement.

3 See https://www.bannerconnect.net/exposure-time-a-new-standard-for-measuring-digital-effectiveness/.

4 New York Times. A version of this article appeared in print on March 30, 2017, on p. B1 of the New York edition with the headline: “A Bank Had Ads on 400,000 Sites. Then Just 5,000. Same Results.” See also https://www.nytimes.com/2017/03/29/business/chase-ads-youtube-fake-news-offensive-videos.html?smprod=nytcore-iphone&smid=nytcore-iphone-share (accessed December 8, 2017).

Additional information

Funding

This work is partly supported by the European Union through SMOOTH (786741) and PIMCITY (871370); the European Social Fund through Ramón y Cajal (RYC-2015-17732); and the Spanish Ministry of Economy and Competitiveness through ECO2015-67763-R and PGC2018-096083-B-I00 projects.

Notes on contributors

Patricia Callejo

Patricia Callejo (MSc, Universidad Carlos III Madrid) is a doctoral student in telematics engineering, IMDEA Networks and Universidad Carlos III Madrid.

Ángel Cuevas

Ángel Cuevas (PhD, Universidad Carlos III Madrid) is a Ramón y Cajal fellow and assistant professor in the Department of Telematic Engineering, Universidad Carlos III Madrid; an adjunct professor at Institut Mines-Telecom SudParis; and a fellow at UC3M–Santander Big Data Institute.

Rubén Cuevas

Rubén Cuevas (PhD, Universidad Carlos III Madrid) is an associate professor in the Department of Telematic Engineering, Universidad Carlos III Madrid and deputy director of the UC3M–Santander Big Data Institute.

Mercedes Esteban-Bravo

Mercedes Esteban-Bravo (PhD, Universidad Carlos III Madrid) is a professor of marketing and markets research, Department of Business Administration, Universidad Carlos III Madrid.

Jose M. Vidal-Sanz

Jose M. Vidal-Sanz (PhD, Universidad Carlos III Madrid) is a professor of marketing and markets research, Department of Business Administration, Universidad Carlos III Madrid.