▲ Korea Fair Trade Commission (KFTC)
More than half of consumers using platform-based services such as online shopping, food delivery, and accommodation bookings make purchases from the top five search results provided by the platforms, according to a recent study.
Most consumers were found to complete their purchases within the first page of results, indicating that the influence of platform algorithms that favor their own products is highly significant.
The Korea Fair Trade Commission (KFTC) announced on June 28 that it has published a report titled "A Study on Consumer Behavior Regarding Platform Algorithm-Based Self-Preferencing."
This study was conducted to verify how consumer choices change when digital platforms use search, recommendation, and ranking algorithms to expose their own products at the top of the list.
To this end, the KFTC built a virtual online shopping mall that replicates the interface of actual online shopping sites and had 3,072 consumers perform two shopping tasks.
In the first round, all groups shopped in an environment without self-preferencing manipulation. In the second round, consumers were randomly assigned to four groups based on whether disclosures or labels were attached to identify self-preferenced products while such manipulation was applied.
Furthermore, each consumer was tasked with purchasing two out of three product categories: Bluetooth speakers, vitamin C, and toilet paper.
The experimental results showed that consumers rely heavily on the rankings provided by algorithms.
A total of 51.7% of all purchases were concentrated on the top five search results.
94.6% of consumers completed their purchases within the first page.
Only 25.2% of consumers changed the default sorting order, and 83.8% did not use any filter functions such as product features or price ranges.
Manipulating algorithms also led to cases where consumers purchased the same products at higher prices.
When a product that was in the mid-to-low range in the first round was placed at the top of the search results in the second round at a 10% higher price, its purchase rate surged by 34 percentage points, from 1% to 35%.
Conversely, competing products that were originally in the top tier saw their purchase rates drop by 32 percentage points, from 52% to 20%, as they were pushed down in the search rankings.
This means that self-preferencing can increase sales of specific products and deprive consumers of the opportunity to choose competing products.
The KFTC analyzed that this is because consumers easily mistake the rankings provided by platforms for indicators of high product quality.
It suggests that simple manipulation of search rankings alone can significantly distort final purchase decisions according to the platform's intentions.
The effects of labels and the disclosure of sorting criteria to mitigate such choice distortion were found to be limited.
Labels attached to self-preferenced products were analyzed to have actually increased the purchase rate of those products by 4.5 percentage points.
This suggests that such labels can be perceived as a guarantee of product quality.
As for the disclosure of sorting criteria, only 10.7% of consumers actually checked it, indicating that it was not sufficiently conveyed to the majority of consumers.
However, among the small group of consumers who did check the disclosure, a trend of an 18.4 percentage point decrease in the purchase rate of self-preferenced products was observed.
Even after purchasing more expensive self-preferenced products, consumers showed higher levels of purchase satisfaction and trust in the rankings.
This implies a structure where it is difficult for consumers to recognize the distortion of their choices caused by algorithms on their own.
The KFTC stated that this study is the first experimental research by the commission to identify the effects of platform algorithm-based self-preferencing on consumer choices, adding, "Policy discussions are needed regarding how to keep the potential for consumer choice bias in check without undermining the efficiency of algorithms."
(Photo: Yonhap News TV, Yonhap News)
※
Copying, redistribution, and unauthorized use in AI training are strictly prohibited.