Amazon, the masters of data, came up short as far as an Indiana couple that scammed the online retail giant for $1.2 million in an electronic return scheme. Big data only does so much when it becomes a substitute for smart business practices.
It's easy to get duped by data. For example, Uber reportedly dodged authorities by using big data. But it can be just as easy to hoodwink yourself through bad interpretations, or by thinking that data solves everything.
Here's the original description of the charges that faced couple Erin And Leah Finan, according to the US Attorney's Office of the Southern District of Indiana:
According to court documents, the Finans defrauded Amazon by falsely claiming that the electronics they ordered were damaged or not working, and then requesting and receiving replacements from Amazon at no charge. Amazon's customer service policy allows, under certain circumstances, customers to receive a replacement before they return a broken item. Amazon closely monitors customers' accounts and orders for possible fraudulent activity. The Finans allegedly went to great lengths to conceal their fraud, creating hundreds of false online identities to perpetrate the scheme. Eventually, however, Amazon and federal law enforcement caught up with them. In total, the Finans allegedly stole over $1.2 million in consumer electronics from Amazon, including GoPro digital cameras, Microsoft Xboxes, Samsung smartwatches, and Microsoft Surface tablets.
The Finans allegedly sold the products at a big discount to a third person, Danijel Glumac, who then supposedly transported the goods and sold them at a mark-up to "an entity in New York," when then sold the product, in turn, to the public. Glumac was also, according to the US Attorney's office, the person who told the couple how to get the returns by Amazon.
Clearly the couple knew which "certain circumstances" allow someone to receive a replacement before sending the original back. Processing returns is expensive, so anything that can increase the efficiency is money in a company's pocket. When you handle as many transactions and the inevitable portion of returns as Amazon, even a small reduction percentage matters.
The couple used multiple online identities, but goods still have to be shipped someplace. Someone has to pay for them, Perhaps there were multiple online payment accounts with debit cards or some other scheme. But, ultimately, how many physical addresses can you have? How could so many returns be tied to a small set of locations?
Even years ago, the biggest online sellers had the ability to use sophisticated statistical methods and multiple data sources to detect fraud. There were many potential triggers, including mismatches of billing and shipping addresses and other data patterns.
It seems clear that some signs finally did tell Amazon something was wrong, and it would have brought in federal investigators. But that's a pretty high trigger point. What if the couple had stopped at, say, $750,000 in merchandise? Or what if many others are using the same techniques over a longer period of time, or quitting earlier? How much might Amazon be losing a year in order to save some money by not having more people going beyond the algorithms of an AI system?
Automation can be good. But it also can take a toll. There's the clearly economic one, but also the encouragement of other would-be criminals who see that a couple could get away with it for so long and take that as encouragement to be smarter, or spread the theft around. A lot of little nips can add up quickly. When you're making a business argument, you should look at all the potential impact of decisions and recognize that big data software is supposed to help aid decisions, not replace them.