We can estimate that at least £23 million of UK spend was potentially intercepted by fraudsters last year – that’s based on fraud currently sitting around the 1% mark and the £2.3 billion spend on display ads during H1 2018, according to the IAB UK. Digital advertising has never before appeared as a bigger, more enticing ‘X marks the spot’ for fraudsters – and where the money is, the fraudsters follow.
With high levels of spending and loss, it’s understandable the industry is eager to flag any potentially fraudulent occurrences. But, in reality, this enthusiastic approach can make it hard to pin down genuine cases of fraud.
To effectively defend against fraud, it’s time to answer a crucial question: where do we draw the line — and how can we better tell friend from foe?
Fraudsters of disguise
Ad fraud is complex. As the industry develops advanced technologies to detect hazardous activity, fraudsters continue to learn how to better replicate user behaviour. So, it’s no surprise advertisers, agencies, and tech partners, frustrated by the difficulty of fraud identification, are increasingly inclined towards knee-jerk blocking.
Attempts to tackle the issue have made progress, particularly the Trustworthy Accountability Group’s (TAG) Certified Against Fraud initiative. The guidelines strongly recommend publishers to adopt the ads.txt specification and independent validation requirements. This increases transparency and the ability to reduce ad fraud. In fact, when all components within the ad purchasing cycle are TAG compliant, invalid traffic (IVT) rates for display are 94% lower than the industry average. While positive, such programmes don’t resolve the challenge of establishing whether entities are reputable or not, or prevent the continued growth of ad fraud operations: now large enough to gain the FBI’s attention.
What’s needed is a greater understanding of how best to measure ad fraud across multiple media platforms so that resources can be directed to address the real problem.
Knowledge is the best defence
Stage one of improving ad fraud protection is defining where current methods fall short. Take exclusion listing. On the surface, this seems like an efficient approach: sites with fraudulent history are added to a list of no-go areas, and advertisers stay safe through avoidance. But it has limitations. Not only are lists updated infrequently — for example, rogue publishers can reappear under a new URL — but they also restrict scale. Even premium sites can fall victim to ad fraud, and removing them as a supply channel can mean cutting off valuable inventory, as well as impact funding for quality content. Moreover, exclusion lists don’t actually determine if ad fraud is present: they only show sites it has affected in the past.
Stage two is recognising which techniques should be applied. At a basic level there is robust vetting — checking that trading partners have the TAG or media ratings council (MRC) anti-fraud seal of approval — and adjusting key performance indicators (KPIs) to minimise risk. For example, replacing metrics that are easy to fake, such as high clicks at low cost, for KPIs linked to specific business goals, including sales and possibly other types of conversions.
At a more advanced level, buyers and sellers must implement granular measurement that enables them to quickly isolate and prevent ad fraud by covering three core analytical pillars.
- Behaviour and network analysis: Using AI/ML techniques for pattern analysis in order to find the differences between human and bot activity, and thus pinpoint signs of ad fraud. Detection models powered by this insight can then measure impressions against varied criteria: such as delivery channel and source, typical user activity patterns, page interaction, and traffic location.
- Environment verification: This is about mapping out the attributes of browsers, devices, and web pages to discover what lies beneath. With respect to bots this includes unique characteristics specific to browsers and devices. When inspecting web pages and the ad serving chain, it may cover the hallmarks of domain spoofing or hidden ads. Both require varied investigation methods, such as ‘honeypots’ used to trick bots into giving themselves away, or watching for inconsistent viewability measurements.
- Malware and targeted reconnaissance: Working with cyber security experts to uncover the details of how the ecosystem is being exploited by fraudsters. This involves taking malware apart to see how it works, assessing invalid traffic, or collaborating with white hat hackers.
As fraudsters become more adept at hiding their true identity, accurately evaluating the legitimacy of impressions can feel like a near impossible task. But industry players must resist the temptation to aim for blanket protection by arbitrarily blocking sites, because in this way they risk jeopardizing scale, missing valuable opportunities to reach customers. Successfully preventing ad fraud starts with an appreciation of the true scale of the issue, and critical to obtaining that is in-depth measurement. Only with proactive and comprehensive analysis can marketers, agencies, and tech partners spot and stop genuine hazards – and ensure ad spend flows towards ads that are served by real publishers that reach real people.