Advertising fraud is typically done by creating fake ad traffic using content-scraping websites or other environments, launching ads outside of a user’s view, or creating other fictitious mechanisms for delivering ads that are not seen by consumers.
The Sizmek engine identifies fraudulent activity in several categories in verification reporting:
Important: Ad Fraud metrics and categories are available in varying degrees for pre-bid targeting, creative targeting, and post-buy verification. See REFERENCE: Media Relevance Metrics and Categories Grid and REFERENCE: Media Relevance Metrics and Categories Glossary for more information.
- Ghost Sites: The Ghost Sites category identifies suspected fraudulent (bot) traffic from sites that are entirely non-human. It is derived from a statistical algorithm that analyses traffic patterns of sites in real-time bidding (RTB) and flags suspiciously high traffic sites as bots, while combining information about the site volume, page structure, and content quality. Due to the nature of bot sites and the high frequency at which they appear, the algorithm is dynamic and is updated on the fly to help detect new Ghost Sites and maintain good coverage.
- Ad Stacking: Ad Stacking is a common fraud technique in which the publisher places multiple ads on top on each other in a single ad placement; only the top-most ad is viewable. Sizmek defines Ad Stacking when 100% of an ad is covered for at least one millisecond and is not 100% transparent. The Sizmek engine detects Ad Stacking for display ads and In-stream ads (VPAID and Enhanced VAST); Ad Stacking information appears in verification reports.
Note: We support Ad stacking for In-stream ads (VPAID and Enhanced VAST). Only users with Enhanced Viewability or Video Viewability enabled will see Ad Stacking information.
- Zero Ads: Zero Ads identifies sites that do not contain ad placements. Ad impressions that are served to these URLs are generated by ad injection, malware, or another fraudulent method. Examples of sites with Zero Ads may include government websites or subscription-based websites.
The Zero Ads classification is attributed to the entire domain and not to specific pages. This classification includes "child" pages in the site; if all of these pages have no ads, the platform tags it as Zero Ads.
- Fraudulent Impressions: This metric represents the de-duplicated count of fraudulent impressions. This metric identifies impressions that were classified with at least one of the fraud categories (ghost sites, zero ads, ad stacking) and excludes overlap between impressions that were classified with multiple fraud categories.
For example, when one impression is classified as both ghost sites and ad stacking, the fraud counts are calculated as follows:
- Ghost Sites: 1
- Ad Stacking: 1
- Fraudulent Impressions: 1
- Suspicious Activity: Suspicious activity impressions are impressions that were redirected from a URL that Sizmek has classified as a ghost site or malware.
To minimize or prevent ad fraud, you can do the following:
- For direct buys, use the Ghost Sites and Ad Stacking in verification reporting to remediate with the publisher directly.
- Use Peer39 anti-fraud targeting solutions in pre-bid environments to eliminate many of these fraudulent impressions:
- Bot Site - Non-human traffic
- Zero Ads
How Does It Work?
The Sizmek engine filters known, IAB-defined bots on every campaign served, irrespective of the use of Sizmek Verification on every campaign. Additionally, the Peer39 statistical algorithm analyses traffic patterns of sites in RTB and flags suspiciously high traffic sites as bots while combining information on the site volume, the structure of its pages, and the quality of its content. Due to the nature of bot sites and the high frequency at which they appear, the algorithm is dynamic and is updated on the fly to help detect new sites and maintain good coverage. The algorithm also identifies attempts to mask domains in RTB and surface the data for targeting purposes.
Note: Sizmek follows IAB guidelines in filtering out suspicious impression or click activity likely caused by bot traffic.