Self-Preferencing
Self-preferencing means that a large internet company makes its own offerings more visible on its platform than those of competitors. Competition authorities often see this as an unfair advantage, because the company is simultaneously referee and player.
Many large internet companies operate a marketplace and at the same time sell there themselves. Amazon is one such case: other merchants offer goods there, but so does Amazon with its own products. Google displays search results from external websites, but also shows its own services like Shopping or flight search alongside them. When such a company pushes its own offerings up in the ranking or places them more prominently, this is called self-preferencing, or in German 'Selbstbevorzugung'. The accusation then is: the company sets the rules of the game and plays along at the same time. Whether this is permitted has occupied courts and authorities in Europe and the US for years.
Why a ranking decides revenues
Visibility is almost everything on the internet. Studies on search engines have long shown that the first results get the majority of all clicks. Whoever ends up on page two practically doesn’t exist. A merchant who doesn’t appear in Amazon’s so-called Buy Box thereby loses the most convenient purchasing path for customers. A small shift in ranking can therefore mean millions in revenue for a company.
The problem is the dual role. A platform has data about all providers: which products are selling well, which prices work, what users are currently searching for. It can use this information to develop its own competing products and then place them more favorably. Competition watchdogs fear that good providers may disappear not because of poor quality, but because of poor placement.
However, there is a counter-position. Platforms argue that their own services are often faster, more reliable, or better integrated. A restaurant suggestion directly in the map saves the user a click. The difficult question, therefore, is not whether a platform is allowed to show its own offerings, but at what point this preference systematically damages competition.
How preference hides in the code
Self-preferencing becomes visible through sorting. Almost every platform uses an algorithm, that is, a fixed calculation rule, which decides which result appears on top. This algorithm evaluates factors such as price, ratings, delivery time, or previous clicks. If a rule is built in that additionally boosts the company’s own offerings, the preference is created directly in the software. From the outside, the result looks like a neutral recommendation.
But the preference can also occur more subtly. Own offerings get a separate field at the very top, while competitors appear in a normal list below. Or competing products are checked according to stricter criteria and approved more slowly. Sometimes it’s enough to restrict an interface for external providers slightly. This interface, technically called an API, is the connection through which other companies feed their data into the platform.
Proving this is therefore difficult. The source code of ranking systems is secret, and modern systems partly learn their weightings themselves from user data. Authorities therefore resort to measurements from the outside: they submit thousands of search queries and statistically compare where own and external offerings land. Precisely such data series were key evidence in the EU proceedings against Google Shopping, which led to a fine of 2.42 billion euros in 2017.
The term in proceedings and in everyday life
In the news, self-preferencing usually appears in connection with antitrust law. The EU has created a law with the Digital Markets Act that explicitly prohibits self-preferencing for particularly large platforms, the so-called gatekeepers. Affected companies include Google, Apple, Amazon, Meta, and Microsoft, among others. Violations can be punished with up to ten percent of worldwide annual revenue.
You encounter this topic more often in everyday life than you might think. A new smartphone has the manufacturer’s apps preinstalled. A streaming service suggests its own productions first. An operating system opens links by default in its own browser. Not every one of these cases is prohibited, but all follow the same pattern.
The term is being newly discussed with regard to chatbots and AI assistants. When an assistant, upon request, always recommends the in-house map service or shop, this is a form of self-preferencing that hardly anyone notices. Because unlike with a results list, the user no longer sees alternatives, only a single answer. Experts consider this the harder-to-control variant of the problem.
