With the pandemic waning, sneaker lovers are once again stepping out to show off their coveted kicks, and sneaker reselling is on the rise. In fact, the sneaker resale market in the U.S. is estimated to be around $2 billion and by 2030 is expected to increase to $30 billion. Some researchers are calling sneakers their own alternative asset class, analogous to NFTs.
As a result of this rise in reselling, sneakers are also emerging as the ecommerce category most susceptible to online fraud, according to proprietary data from Riskified’s ecommerce risk intelligence platform.
This article will discuss the sneaker resale trend, major indicators of fraud for retailers to be aware of, and how merchants can combat sneaker fraud using machine learning.
Heeling and Dealing
The sneaker resale market is a lucrative industry. With top companies releasing limited “drops” of desired products, limited-edition sneakers can sell for hundreds or thousands of dollars over their original retail price. The market is driven by “sneakerheads” and resellers, often using bots to buy up large quantities of in-demand sneakers, leading to shortages for regular consumers. The pandemic has further fueled the market’s growth, with online sales becoming more prevalent.
The resale value of sneakers has made them an attractive target for fraudsters. Riskified’s data found that sneakers are 162% riskier than the average industry risk level across all industries. This places sneakers ahead of high-end fashion clothing (69% riskier than average), making these two categories the only ones with an above-average riskiness score (when compared to electronics, jewelry and watches, cosmetics, low- to mid-range fashion, home and other ecommerce goods).
The data also revealed that orders between $100 and $300 are the most challenging transactions for retailers. This range has more than half of the legitimate orders and about half of the fraudulent volume, making it difficult for retailers to distinguish between good and bad orders in this dollar range. This results in higher operational efforts and costs to combat fraud. The data suggests that merchants need to improve their fraud detection strategies in this segment in particular.
In general, the highest dollar value sneaker purchases (those above $500) are the riskiest ones. They make up more than half of the fraudulent transactions in terms of dollar value but are significantly less popular among the legitimate population, and their contribution to legitimate dollar revenue is relatively small.
Interestingly, the lowest dollar amount transactions (under $100) had a similar contribution to the legitimate dollar sales revenue as the $500-$1,000 category, meaning that lower-priced transactions are more likely to be legitimate. The riskiness gap between these two purchase ranges is considerable, and the $500-1,000 transactions have significantly more potential to negatively impact sneaker merchants’ fraud costs.
When it comes to sneaker fraud, it was found that “new customers” are consistently riskier than “returning customers.”
Returning customers make up the majority of legitimate sneaker shoppers, while both new and repeat customers are evenly represented in fraudulent transactions.
In addition, new customers are more likely to be fraudsters. They account for about one-third of legitimate transactions and around half of fraudulent transactions.
Customers using proxy IP addresses are also vastly more likely to be committing fraud. The yearly averages show that the proxy users represent a relatively small piece of total legitimate transactions (< 7%), while their share of fraudulent transactions is significantly higher (just above 20%). Proxy transactions in sneakers also became riskier over the course of 2022. While both proxy and non-proxy IPs experienced an increase in risk levels throughout last year, the risk gap between these two segments widened.
However, simply declining all proxy transactions is not a recommended solution for retailers, as there are also many legitimate reasons for using proxy servers, such as anonymous browsing and access to geo-locked websites. Basing a decision purely on proxy indication can lead to falsely declining good customers.
Stomping out Sneaker Fraud with Tech
Online fraud is a significant challenge for retailers, and they must take proactive steps to mitigate its impact. Technology has a significant role to play, with machine learning and AI solutions able to detect patterns to fight popular fraud types.
While effective fraud detection and prevention is essential, it must be done without adding unnecessary friction for customers or declining legitimate transactions. To achieve this, ecommerce fraud data must be contextualized by experts. Fraud analysts must be strategists and storytellers to both gain a holistic picture of the fraud landscape and also tell its story effectively. This human element is what differentiates rules-based fraud prevention from machine learning.
Experienced fraud analysts support higher approval rates by redeeming transactions that machine learning models pinpoint as higher risk. Analysts who understand fraud can help identify blind spots in the machine learning models and account for biases and real-world changes that would impact fraudster behavior. Their experience and insights are continuously fed back into the models, creating a cycle of improvement.
In conclusion, retailers need to improve their fraud detection strategies to identify fraudulent transactions and reduce operational efforts and costs. It’s essential to strike a balance between identifying fraudulent transactions and not over-declining good customers. By prioritizing fraud prevention, retailers can protect themselves and their customers and maintain the integrity of the ecommerce ecosystem.
Lev Gal has been a Data Analyst with Riskified, an ecommerce risk intelligence platform, for nearly five years, focusing on merchant insights. She holds a degree in Economics and Business Management from the Hebrew University of Jerusalem and completed an MBA internship in Finance and Marketing at Tel Aviv University.