Why using exclusion and inclusion lists is not enough

11/27 By Dale Older

For digital marketers, fighting fraud is the closest you can get to an arms race. Fraudsters are constantly developing sophisticated new tactics to steal digital marketing dollars. For marketers to keep up, it’s critical that fraud prevention techniques keep pace with fraudsters’ constant evolution. That’s why it’s surprising that some of the tools marketers have traditionally used to fight fraud have failed to evolve, and in some cases could be leaving digital ad budgets exposed to fraud while simultaneously limiting campaign reach. How the industry is fighting fraud could actually allow fraud to proliferate.

Exclusion Lists

One of the most commonly used tools for fighting back against fraud are exclusion lists. As the name suggests, exclusion lists are used to track specific websites, domains, and mobile apps that marketers specifically want to avoid. These lists often include sites that are known to be fraudulent, or which are likely to employ undesirable traffic generating practices. Exclusion lists help advertisers prevent their campaigns from running in known fraudulent environments. That peace of mind, however,  is likely to be short lived.

Adding to exclusion lists can only protect your campaign from sites that are known to be fraudulent, but the web is expanding at an exponential rate with millions of new sites appearing every year. Exclusion lists, maintained by human analysts, can’t possibly keep up with this rate of growth, leaving advertisers exposed to an ever increasing stock of new fraudulent inventory. What’s more, exclusion lists tend to be fairly static, updated by hand on a monthly or even weekly basis, far too slow to keep up with the rapidly evolving web.

Inclusion Lists

On the opposite side of the coin are inclusion lists. Rather than painstakingly track and exclusion lists every site that demonstrates fraudulent activity some advertisers prefer to confine their ad buys to a set of pre-approved sites that are known (or expected) to be fraud free. This does effectively combat the problem of unaddressed fraudulent inventory that isn’t addressed by adding to exclusion lists. Inclusion lists, however, also come with what might ultimately be a heavier cost for advertisers.

Inclusion lists, like exclusion lists, are often often static. Because they are updated too infrequently to account for new inventory, relying on inclusion lists effectively stifles scale by cutting advertisers off from valuable new inventory and limiting audience reach. These lists often don’t account for the possibility of previously high-quality sites experiencing an upswell in fraudulent activity. Without employing a verification solution, advertisers relying on a inclusion lists would have no way of knowing if the sites they’re advertising on remain fraud-free throughout the life of their campaign, giving them little recourse in requesting refunds, or even better, preventing their ads from running on fraudulent traffic in the first place.

The right lists

Ad fraud is a rapidly evolving challenge that requires an equally adaptive solution. Fortunately, the rise of machine learning powered by big data provides the opportunity to identify fraud dynamically rather than rely on static lists. IAS uses machine learning to power our fraud verification capabilities, and in parallel, we are introducing AI automation to enhance scale for our fraud detection solutions. IAS sees up to 10 billion impressions per day–more than any other verification provider–a unique data set allowing us to detect fraudulent behavior patterns that others might miss.

IAS machine learning technology is able to distinguish real user behavior from bot behavior by looking at anomalies within site visitation patterns and analyzing other relationships between site visitors and the sites they visit. Machine learning can detect these patterns even when they are too complex for human analysts to perceive, allowing us to surface bot activity and detect instances of fraud quickly and at scale. We track these patterns and add this fraudulent behavior to our models to identify fraud without limiting the scale of digital campaigns.