Recently, we presented the results of the latest Media Quality Report for H1 2017. We were impressed by the number of insightful questions asked around the topic of ad fraud. Unfortunately, we were unable to respond to every inquiry during the broadcast. In this article, we address some of those questions.
Q: What accounts for differences in fraud rates across display and video? What about mobile?
Ad fraud represents a malicious attempt by individuals to earn revenue by generating impressions through non-human audience sources such as bots. Fraudsters are looking to earn as much money as possible; therefore, fraud tends to follow traffic sources that have the highest potential to drive revenue.
Industry average CMPs differ by ad format and by device. For example, video ads tend to generate higher CPMs for publishers when compared to standard display. This results in an increased likelihood that ad fraudsters will target video.
A similar pattern has emerged when comparing desktop and mobile display units. The mobile advertising gold rush over the last few years has pushed mobile display CPMs higher than desktop, resulting in an increased flow of fraudulent activity into mobile platforms.
Q: Why do you use present both optimized and non-optimized fraud benchmarks? Wouldn’t it make more sense to have one average?
We use two different benchmarks because they each tell us something different. The two benchmarks reflect whether or not an advertiser has developed a strategy to proactively combat ad fraud. Through this comparison, we prove to advertisers, platforms, and publishers that efforts to reduce fraud generate value and save marketers money.
Optimized fraud levels reflect impressions within campaigns that have integrated some sort of fraud verification and optimization strategy. In simple terms, that means these numbers are reflective of advertisers, publishers, and platforms that are actively taking measures to minimize fraud.
Non-optimized fraud provides a window into what the state of the industry looks like with no efforts to reduce it. These impressions are derived from campaigns where there are no optimization efforts in place. Non-optimized fraud numbers show what a typical campaign would most likely look like if an advertiser or client did not leverage an anti-fraud strategy.
Additionally, the data is presented as a range because an average will not illustrate how high and low ad fraud levels can be, particularly for programmatic.
Q: What makes SIVT so essential to my fraud mitigation strategy?
General invalid traffic (GIVT) can be detected through routine checks and filtration methods, including using lists of offending IP addresses. Key examples of GIVT include datacenter traffic; bots and spiders or other crawlers masquerading as legitimate users; non-browser user-agent headers; and more.
Sophisticated invalid traffic (SIVT) requires significant and sophisticated technical implementation, as well as complex statistical analysis for a robust detection solution. Over 90% of fraud comes from SIVT and can include hijacked devices, hijacked tags, adware, malware, among others. If your fraud mitigation strategy doesn’t address SIVT directly, then it fails to address the majority of value erosion through fraud.
Q: What are the types of patterns or behaviors bots exhibit that differentiate them from human users?
Through our extensive fraud research, we’ve identified a few key behaviors that indicate a site visitor is a bot and not a human.
Bots tend to have higher rates of viewability. This makes sense as bots exist to generate revenue for fraudsters. Higher viewability means higher revenue from fraudulent sources. When a bot is on a page, 99% of the time there is a viewable ad versus 55% when a human visitor is present.
Bots can steal valuable cookies. Not only does this result in flawed audience data for publishers, but it leads to wasted digital advertising investment for brands. Advertisers can be tricked into targeting bots with stolen cookies unaware that the intended recipient of the ad is not human. Ultimately, this leads to marketers paying for inventory that is less valuable as it’s reduced by bot fraud.
Bots tend to be very efficient and consistent in their browsing patterns. They are designed to “view” as many ads as possible and therefore will remain on a page for as long as needed for views to be registered. Humans tend to have more of a varied browsing patterns, either leaving a page immediately or leaving a tab open and inactive for hours.