In the 1985 sci-fi comedy Real Genius, we meet Lazlo Hollyfeld, a former engineering student who spends his days trying to game a Frito-Lay prize giveaway that advertises: “enter as many times as you want, no purchase necessary.” Ultimately, he sends 1,650,000 entries, calculating that this will guarantee him “32.6% of the prizes.” At the end of the film, the main characters share a good laugh watching Lazlo drive off in a brand new RV packed with winnings.
It’s meant as a joke, but people have been gaming giveaways for as long as they’ve existed. In fact, Lazlo’s exploits in the film were based on a very real “prank” pulled by CalTech students in 1975. They used a computer – in the days before ubiquitous computing – to enter a McDonald’s promotion at huge scale, noting that the contest rules didn’t limit entries. In all, they netted themselves more than $10,000 in cash and prizes, including a station wagon. Admirably, they donated most of the winnings in the end, but McDonalds publicly criticized the group, saying their actions were “in complete contradiction of the American standards of fair play and sportsmanship.” McDonald’s also changed the entry language in future promotions.
A half-century later, what were once the subversive side projects of computer science majors are now the modus operandi of highly skilled, technologically advanced fraudsters stationed all over the world. And these are not “victimless crimes” either. The data generated from sweepstakes and promotions like these are the lifeblood of marketing campaigns that keep companies growing and teams of people gainfully employed. The fraudster vs. marketer cat-and-mouse game has been escalating for decades, with hackers continually increasing their level of sophistication. Thankfully, there are new weapons against this kind of fraud: Machine learning and AI.
Fraud has certainly gotten more sophisticated since the 1970s. While the CalTech students were admittedly advanced for their era, much of their entry scheme still involved lots of time-consuming human activity: Printing. Mailing. Deliveries to mailboxes. Today, comparable scale can now be achieved in mere minutes from anywhere in the world, using just off-the-shelf software, a VPN and a few basic algorithms.
Advertisement
Here’s an example of the scale marketers face today: a recent promotion run by our team ultimately drew more than 239,000 submissions. Of those, we found that about 220,500 were fraudulent, meaning that they were submitted in huge bulk by scammers who’d created single-use email addresses for the sole purpose of trying to rake in winnings. Of the total 239,000, only about 18,600 entries were found to be viable — meaning that they came from real people who were truly engaged with the brand and wanted an honest chance to win prizes.
While that may sound disheartening for the marketers on the other side of campaigns like this, those numbers represent a huge win for the team. We were able to turn over more than 18,000 verifiable contacts, representative of real consumers who were aware of the brand if not already loyal customers. If the client converts just a fraction of those over time, the campaign will have driven impressive revenue.
When a sweepstakes attracts upwards of 92% fraudulent attempts, it’s critical to develop an airtight defense. Historically, data from the majority of online promotions and giveaways have been scraped using only one or two metrics — IP addresses being the main factor. In contrast, by using advanced machine learning, we are now able to scour and verify data using 24 separate factors. This is done without slowing down the process or overloading the team — humans are always in the loop to ensure accuracy. Using ML reduces the machine time needed by at least 50 hours per campaign.
As fraudsters grow more sophisticated, the system continues to spot new patterns and root out falsified data coming in through the new tactics they try. Over the next few years, the learnings derived from ML systems will feed increasingly powerful AI and proprietary LLMs, which will then be able to spot even more patterns and improve results at more granular levels.
The overarching goal of this technological arms race, of course, is to ensure the accuracy of incoming customer data and the efficacy of sweepstakes and promotions campaigns. After all, we’re talking about real people: hardworking marketers who need to make the most of their budgets to see their companies (and careers) grow, along with real customers and fans who deserve to have a real shot at prizes and winnings. Advanced ML and AI offers real help to these very real people. Incorporating this technology is a stroke of real genius.
Dan Jahn is President and CEO of Probability, a leader in promotional marketing (sweepstakes, contests, and games) that serves some of the world’s biggest brands in retail, entertainment and consumer products.
