How AI will put click farm criminals out to pasture

As marketers invest more in programmatic marketing, criminal elements are increasingly setting their sights on stealing our precious media dollars.   Could Artificial Intelligence help thwart them…?

In my January blog Dark Secrets Lurking in the Ad Supply Chain, I explained how advertisers are becoming increasingly distanced from the media they’re purchasing. Target audiences are being assembled and on-sold through a deep, complex supply chain of agencies and third-party networks.

The deeper the supply chain, the greater the risk of phantom audiences being conjured in the darkest corners of the internet.  Bundled up and and on-sold as genuine prospects, they are to unsuspecting advertisers what sub-prime mortgages were to investors, pre-GFC.

From Risk to Racket

Criminals are already raiding advertisers’ media budgets. A month ago, three Chinese citizens were arrested in Thailand after a police raid uncovered a massive click-fraud operation.  About 500 smartphones and 350,000 SIM cards were being used to artificially inflate the number of clicks on ads in WeChat, China’s largest messaging app.

By inflating the clickthrough rates, the fraudsters were able to trick companies into believing their ads were over-performing, inflating bid prices. A 2015 study estimated that this kind fraud potentially duped advertisers out of billions.

With programmatically-purchased ad spend now estimated to be in excess of 55% of the digital advertising market in the US, the incentive for criminals to click-farm is already huge.  That incentive will only grow as advertisers chase the average 22% increase in ROI being reported by those investing thoughtfully in programmatic buying.

AI to the Rescue?

IBM’s Watson AI is already helping optimise programmatic buying, achieving impressive results.  In a 2015 trial, Watson reduced IBM’s cost per click an average of 35% – up to 71% in some cases.  IBM is now making that technology available to agencies and exchanges so they too can maximize ROI through “cognitive bid optimization.”

Watson achieves these advances by detecting previously unseen patterns in data.  Far beyond the capabilities of humans or algorithms, Watson digests almost inconceivable quantities of unstructured data to find hidden relationships, constructing and testing its own hypotheses to continuously optimise results.

In the case of cognitive bid optimisation, Watson plugs in directly to the media bidding platform via an API.  It then leverages historical performance data to recommend the optimal bidding strategy for each campaign, using data to determine budget allocation and deciding on the best ad format, size and optimal time of day to target users.

With Watson’s ability to continuously learn at scale, advertisers and agencies can add new data dimensions over time.  Watson will refine its models with each new insight, better predicting optimal bids in real time.

Dusting for Digital Fingerprints

The same capacity to learn will enable AIs to fight click fraud in the advertising supply chain.  Fraudulent clicks and phantom audiences will have subtle, unique characteristics that AIs will be able to detect and flag.

In much the same way that The Matrix‘s Neo was able to discern glitches in his simulated world, AIs will be able to detect illegitimate audiences from their distinct digital fingerprints.  They will reveal themselves through patterns of access, IP addresses, click-rates that stand out as abnormal versus real human audiences.

Advertisers won’t even need to intervene to protect their investments.   AI’s will quarantine the illegitimate sites and audiences, blacklisting them from bids… maybe even flagging them to law enforcement agencies.

Appropriately-equipped advertisers and agencies will be able to share their intelligence on fraudulent fingerprints, in much the same way that computer security companies come together to share share virus signatures.

The Current Competitive Edge

The more advertisers and agencies who employ AI’s the stronger all marketers’ “herd protection” will be.  But until then, savvy marketers – those employing AI’s like IBM’s Watson – will enjoy a competitive edge.

In the programmatic buying arms race, those without AIs will be at increasing risk of being duped out of their precious media budgets.


Leave a Reply

Your email address will not be published. Required fields are marked *