Search for a product on a big e-commerce website such as Amazon, and you’ll often see many different results accompanied by a blizzard of information: paid product ads, badges and tags, ratings, reviews and more.
Most e-commerce websites offer tools – such as star ratings, written user reviews and information on how many products have been sold – to help consumers make informed decisions.
But is that information always accurate? And does it help buyers make the best decision – or might other motives influence the algorithms?
That’s the question Fei Long, a marketing professor at UNC Kenan-Flagler Business School, and her colleague Yunchan Liu of the University of Illinois Urbana-Champaign explore in “Platform Manipulation in Online Retail Marketplace with Sponsored Advertising” published in Marketing Science.
Large e-commerce sites, such as Amazon, eBay and Alibaba, no longer make money solely by selling products. Sales of advertising on their own platforms have become major sources of profit.
In 2022, for example, Amazon brought in $38 billion in ad revenue – an income stream that analysts say is likely more profitable than selling products or some of its other businesses. Ads are typically bought by sellers through an online bidding process where the top bidder is able to, in effect, purchase more visibility.
With multiple revenue streams coming from online sales, Long wondered how that might influence the behavior of these e-commerce platforms. Would they still be incentivized to show consumers the “best” deals or would they have reason to adjust listings based on whether they were maximizing ad revenue or product sale commissions?
So, Long and Liu created a model that simulated how big online retailers might change their websites to maximize profit – even if it came at the expense of transparency with consumers.
Among the questions they explored:
To work through these questions, Long and Liu developed a model of consumer behavior. Then they applied that model to a variety of scenarios to determine how e-commerce platforms might act to maximize their own profits.
To construct the model, the researchers posited there are two types of consumers: explorers and nonexplorers.
Explorers take the time to examine multiple products listed in search results, and they prioritize more attractive products, such as those with higher ratings, greater past sales or badges like Amazon’s Choice.
Nonexplorers only examine top search results, and for those consumers, product visibility is key – the top organic search result and the top advertised product, for example. Research about how people behave on e-commerce sites indicates that many consumers search very few or just one product, suggesting nonexplorers are common.
Given these circumstances, it becomes possible to model how e-commerce platforms could maximize revenue by making products more or less visible and attractive to potential consumers.
Long and Liu looked at what would happen when a consumer got search results with two potential competing products: The best option and another option. They used a variable called “match,” which roughly equates to the average popularity for each product. A higher match means, essentially, the two products are more competitive in terms of desirability.
When it is a low-competition market – one is quite popular and the other has very low popularity (low match) – the e-commerce platform has an incentive to boost the low-match competitor (such as tolerating its fake sales and fake reviews or granting to make the product more attractive).
That’s because if the product appears to match better, and hence becomes more competitive, the lesser product will be incented to spend more on advertising, bidding up ad costs and creating more revenue for the platform.
On the other hand, when it is a high-competition market – the two products both match well and are already close – the e-commerce platform is incented to boost the stronger product to create more distance between the two, such as boosting the seller’s visibility in organic results, in addition to making it more attractive. This occurs because two well-matched products might compete on price, lowering the e-commerce platform’s commission on sales.
When product match between two products is between those extremes, however, the e-commerce platform has little incentive to boost one or the other.
Sellers on e-commerce platforms should also consider how tactics they use to increase their attractiveness, such as buying fake reviews or fake sales, might stimulate changes in the e-commerce platform’s algorithms, either helping or hurting visibility and, consequently, sales.
Certain types of deceptive actions by sellers, such as buying fake reviews, have been illegal for years. But the large and diverse e-commerce landscape makes it difficult for regulators to police all sellers.
Long says her research suggests that policymakers might be more effective by focusing on e-commerce platform behavior. This could include how the platforms fail to police fake reviews and fake sales, as well as how their algorithms influence buying decisions.
For consumers and sellers, the research is a reminder that online marketplaces no matter how much we rely on them – are not neutral arbiters of value.