Depending on its type and reach, fraud could result in exceptionally costly problems for businesses and even individuals. However, companies depend on big data to make fraud less likely to happen. Here are some of the ways technological advancements are gaining ground in the effort to stop fraud in various industries.
Stopping so-called "serial returners" in retail
Most established companies have return policies that help create trust between the retailer and purchaser. For example, if a person has a proof-of-purchase and offers an explanation for why they want to send something back, those pieces of information may be sufficient for processing a return.
Retailers, however, are learning to detect the instances when people might be returning items to participate in return fraud, and they often do that with big data. For example, a store might have compiled information about what constitutes "normal" behavior for customers who buy a certain number of items per year. Perhaps the majority of people who buy at least 40 products from a merchant per year return no more than three of them.
If a person behaves in a way that does not fit with what is typical, brands may create blacklists. A survey published in 2018 found 42% of US retailers saw an increase in serial return activity over the past year. Some also revealed a willingness to ban customers who displayed too many suspicious tendencies.
At the very least, tapping into customer data before processing a return allows a retailer to assess whether a person may be falsifying the reasons for trying to bring back a purchase, instead of merely accepting the shopper's description of events.
Cutting down on tax evasion
Paying taxes makes most people grumble, and some individuals think they can defy the taxation organizations and commit tax fraud. However, the Internal Revenue Service (IRS) is among the US agencies relying on big data software to pinpoint potential fraud cases. One company uses neural network models with 600 variables to detect outliers.
If a person declares business losses for multiple consecutive years or got audited before, those are two things that could translate into red flags for the software. Evidence also suggests the IRS is interested in a data-mining tool that looks through social media accounts for information that supports previously identified tax compliance cases associated with particular individuals.
Such applications of big data do not overpower the need for human involvement. But, big data platforms can often process substantial amounts of information much faster than people could without help.
Highlighting the telltale signs of spam emails
The broad reach of a spam email means criminals can send out messages targeting thousands of users or more in seconds. Email-based fraud attempts come in various forms, with the goal of fooling even tech-savvy users. Some emphasize a sense of urgency, such as by telling people they need to act fast to claim lottery winnings or prevent themselves from losing access to accounts.
Others ask for sensitive information while posing as legitimate organizations. Or, the emails might appear to come from people recipients know and include attached files masquerading as vacation photos or work documents. Big data platforms and machine-learning algorithms can learn the characteristics that often show up in spam emails, then warn email users of potential problems.
Some companies also dig into data to determine the most common kinds of spam messages during particular times. One company did a study and found 46% of spam emails sent in spring 2018 were for dating-based scams, according to the collected samples. The research also warned emails with malicious attachments or links were common.
When people use big data platforms to analyze emails, they can then warn the public about things that could cause them to fall for scams. Then, it could become a more straightforward task to stay ahead of the cybercriminals who send messages to trick recipients.
Curbing Medicare fraud
Medicare is the primary way Americans who are 65 or older receive health coverage. However, some Medicare users game the system by submitting claims for treatments they never received, or physicians will illegitimately prescribe medications. Data indicates Medicare-related fraud cases, or associated waste or abuse, cost between $19bn and $65bn annually.
In one study, researchers examined millions of cases and came up with ways to create an algorithm that was better than average at picking out positive instances of Medicare fraud.
The team emphasized other issues, such as clerical errors, could initiate false positives. But, the aim was to use the algorithm to aid human auditors, thereby drawing attention to possible cases of actual fraud and reducing time-consuming processes.
Reducing bank fraud cases
Banks often use technology similar to some of the methods used above that learn what authentic user behaviors look like and flag potential differences that could indicate wrongdoings. There is also a push toward empowering users to conquer fraud, especially if the apps associated with banks and credit cards send real-time notifications of purchases and let people track purchases.
Big data is frequently at the heart of those alerts. Algorithms recognize trends, and after spotting something out of the ordinary, might cause a text message to go out that asks "Did you make a large transaction at 3:30pm. on January 12?" or something similar. A person could respond to the message to confirm they did, or realize it is time to contact their financial institution.
Also, if individuals could disable their cards as soon as they become aware of unauthorized activities, it would be possible to stop criminals from doing as much damage as they otherwise might once a card falls into the wrong hands.
Fraud is rampant, but technology fights back
People will never live in a world without fraud. Criminals perpetually evolve and come up with new methods.
Fortunately, technologies like the ones described here could make fraud cases less prominent and limit their effects.