in General

Leveraging AI to Ensure Brand Safety and Combat Fraud

By Quantcast

Marketers have anxiety dreams about inventory control. Evil bots lurking to torpedo ROAS.  Mud wrestling videos waiting to drag brands through the dirt. Fraud and brand safety throw shade on the promise of programmatic, but tools exist to come out on top. Clearly understand the obstacles and apply the right strategies, and you can meaningfully and cost effectively reach the real audiences you want–on your terms.

Fraud vs. Brand Safety

Although fraud and brand safety may appear to blend together in a haze of digital angst, they are not the same thing.  

Both fraud and brand safety are inventory control issues.  But the use of deception to misrepresent and miss-sell inventory constitutes fraud.  The industry does not define types of fraud in any one consistent way, but advertisers typically worry about 1) fake/bot activity sold as human activity–including click fraud, 2) domain spoofing where a cloaked URL appears as a genuine website on an exchange, and 3) hidden ads and viewability issues — such as a page loading ads behind other content. Since only people have credit cards, fraud occurs around click-through-rate and other incentives that don’t involve purchases/conversion.

Online advertising offers fraudsters an alluring opportunity: according to the International Advertising Bureau, these shadowy opportunists netted $7.2 billion in 2016 alone. A big contributor? The complex ecosystem of exchange-based layers creates a lack of transparency. This can make the process of buying and selling inventory hard to audit and creates the chance for fraud to happen.

Brand safety, on the other hand, consists of real human traffic on sites a brand wants to avoid (usually because brands perceive the content on those sites as in some way conflicting with their core values).  Bots do not create brand safety issues–human decision-making errors and flaws in machine training do. Brand safety guidelines vary from client to client. (Typical sites brands strive to avoid might include gaming, sensational news, and pornography.)  

Take Control

Typically, the industry manages fraud and brand safety through manually maintained blacklists– as well as whitelists, agreed deals, exchange and vendor-provided URL classifications, and content validation vendors. A few problems arise with traditional methodologies however, especially with static blacklists.  These quickly become outdated–both in terms of anti-fraud and brand safety applications.  In addition, the majority of programmatic partners rely on third parties to place bids on ad exchanges, building another layer into an already-cluttered ecosystem, and creating an additional chance for both dishonesty and human error.

A Three-Pronged Approach

At Quantcast, we rely on our live data set and machine learning capabilities to combat fraud and ensure brand safety. In addition, we sit directly on the ad exchange which gives us the opportunity to evaluate the safety and quality of opportunities autonomously, rather than relying on third parties to filter.

Proactive: Quantcast engineers are in the business of being picky.  Our systems carefully weigh opportunities against potential fraud and brand safety criteria before bidding.  Here’s how:

1) The team leverages our real-time insight into live audience behavior across the entire internet to discern the differences between human and non-human traffic. We model fraud using these signals and use these models to detect and reject suspicious opportunities–including spoof sites.  

2) Our systems and machine-learning algorithms stay one step ahead through the dynamic maintenance and continual optimization of blacklists, which draw on a combination of cookie behavior, known lists of suspicious data center IPs, dubious publishers and exchanges to identify untrusted publishers and suspicious traffic. This enables us to block human-driven click, form fill and impression fraud–as well as non-human traffic (bots, toolbars, and malware)–during our pre-bid filtering.

3) Our dynamic and custom-tailored blacklists–together with category blocking, keyword blocking, and publisher blocking–weed out inappropriate opportunities and reduce the impression universe, on a case-by-case basis, to only what is safe for each client’s brand. 

4) In the US, we also offer the option of high impact ad formats in brand safe whitelists–with viewability guarantees and contextual relevance,  as well as keyword blocking globally.

Reactive:   We work fast to get it right.   Despite our thorough safeguards, not all fraud and brand safety issues can be detected beforehand.  We respond rapidly and thoroughly to put corrections in place when a customer or partner alerts us of an issue. In cases of fraudulent placements or placements that violate brand safety specifications, we add to blacklists, and thoroughly investigate to avoid any repetition in future.

We also offer guarantees.  Specifically in terms of viewability, all brand campaigns come with a 55% guarantee, with upgrades available up to 70%.  vCPM (guaranteed viewable cost per mille) is also available.

Collaborative: We ally with the best in the industry to fight harder, together.  We work with companies across the Real Time Bidding ecosystem, like SSPs/ Exchanges and trade bodies such as IAB’s Trustworthy Accountability Group (TAG), to identify and help create industry-wide blocks of any inappropriate or miscategorized inventory. We run Moat verification on all of our video and display creatives, as well as Integral Ad Science and DoubleVerify on a sample of these creatives, to measure ourselves and constantly improve our own blocking algorithms.

For an additional level of control over both fraud and brand safety, we can integrate client-chosen content verification vendors (CVVs) for specific campaigns. This collaboration enables a tracking tag to be placed into creative, giving visibility into exactly where each ad is served, and even allowing for the placement of blocking tags–so ads are NOT served if environments do not stringently meet all brand safety and anti-fraud criteria.

Quantcast engineers work closely with CVVs to align on each client’s safety criteria (e.g., some clients might classify a music site containing graphic lyrics inappropriate, while others would not), and direct efforts to strike the perfect filtering balance between mitigating risk and maximizing opportunity.

At Quantcast, the security and peace of mind of our clients come first.  We take fraud and brand safety very seriously.  And because of our dedication, visibility into live behavior across the internet, and superior AI–we get serious results.

Quantcast