[Inside Chodong]The Invisible Hand Exploiting Algorithm Blind Spots

Self-Trading Detected in Coupang Rocket Delivery
Fake Purchases and Reviews Manipulate Customer Ratings
Artificially Boosting Sales Volume for Higher Profits
A Critical Blow to Platforms Scaling Up Through Auto-Recommendation

The power of algorithms on online platforms is formidable. According to the 2019 report "YouTube Recommendation Algorithm and Journalism" published by the Korea Press Foundation, an algorithm is defined as a set of procedures or rules established to solve a particular problem. Through these algorithms, it is possible to process and reflect real-time results from an enormous volume of data that would be impossible for a person to review in a lifetime.

Online video services (OTT), including YouTube, utilize algorithms to recommend videos based on users’ preferences and usage patterns. The more videos users watch on these platforms, the greater the revenue generated. YouTube CEO Neal Mohan stated in an interview with The New York Times that "70% of YouTube watch time is generated by recommendation algorithms."

Algorithms are also a powerful weapon in e-commerce, which dominates the distribution market. They highlight products or discount information tailored to specific users on homepages or mobile applications, thereby encouraging purchases. Sometimes, these platforms come under scrutiny for so-called "manipulation," where recommendation methods are artificially arranged. The market defines such practices as unfair conduct and imposes strict controls.

If it is not the platform but the seller who exploits loopholes in the algorithm, this presents a different dimension of the problem. A recent account from a seller on Coupang was particularly candid. Company A, a hair accessory manufacturer that supplies products for Coupang Rocket Delivery, has been recruiting members via a KakaoTalk open chat room to place orders for its products and write online reviews. Coupang Rocket Delivery operates by purchasing manufacturer products directly. For products with low recognition or newly launched items, Coupang initially makes small purchases of two to three units. If actual sales occur, the algorithm recognizes these as genuine purchases and places additional orders.

Company A adopted a "self-trading" method, generating fake purchases to increase exposure for poorly selling or newly introduced products on Coupang. Every day, the company introduced at least three to four products to about 500 participants gathered in the open chat room, requested them to make purchases, and paid a nominal reward for leaving reviews. To avoid suspicion from Coupang regarding abnormal transactions, they specifically targeted products with low ratings to boost customer satisfaction scores and focused on items with few reviews. They manipulated the algorithm by instructing participants on how long to stay on the platform and how many photos to include in their reviews.

This poses problems in three respects. First, Rocket Delivery suppliers can expose a variety of products to consumers with less cost and time compared to general sellers, thereby enticing Coupang to purchase larger quantities directly. General sellers, who must bear sales commissions, marketing expenses, and shipping costs, are at a competitive disadvantage. For Coupang, failure to filter out fake purchases results in unnecessary additional orders and increased inventory burden. Consumers, who trust Rocket Delivery suppliers as verified companies, may be deceived by someone’s fake purchases and reviews into buying lower-quality products.

Coupang stated that to prevent practices that disrupt platform order, such as self-trading or fake reviews, it is strengthening its seller monitoring system. Upon detection, it suspends orders and requests explanations, and if a satisfactory explanation is not provided, it ultimately terminates transactions. However, Company A’s manipulative activities were only recognized belatedly through reports from other sellers, and sanctions were imposed afterward. Although only a few cases have surfaced for now, the so-called "invisible hand" exploiting algorithmic blind spots will become increasingly subtle and sophisticated. If these are not properly filtered out, not only consumers but also platforms that have built their influence through algorithms will become victims.