Spotify’s Wrapped doesn’t really tell us the soundtrack of our life, but how the algorithms that accompany us every day work.
Like every year, at the beginning of December our social pages were literally invaded by the so-called “wrapped”, annual reports that platforms such as Spotify and Amazon Music offer to users to describe the past year “in numbers”. The most popular songs, favorite artists, dominant genres, total minutes of listening and, this year, other new features such as the average age that is shown based on one’s tastes, or the creation of a perfect fictitious music festival with the most loved artists.
Everything is transformed into colorful and easily shareable graphics, but behind this story of each user’s musical tastes lies a less obvious question: does this data really represent what we listen to, or is it just a simplified version of our habits?
Selected data. A careful analysis of the newspaper attempted to answer Slashgearwhich lasted over a year and was led by technology analyst Anthoni Oisin. The first critical point concerns the method with which Spotify collects and filters the data used to build its Wrapped: essentially, not everything that is listened to contributes in the same way to the final statistics.
A song, for example, must exceed a minimum threshold of use to be counted, while some audio categories are excluded upstream, such as relaxing sounds or environmental tracks. Furthermore, data collection stops early, usually in mid-November, to prevent Christmas playlists from skewing the results and to allow time for processing. This means that several weeks of actual listening never make it into the final summary, reducing the representativeness of the sample.
Ambiguous metrics. The discrepancies become even more apparent when you compare the data collected by Spotify with that obtained through independent tracking platforms like Last.fm, which record every single “play” without editorial filters. In many cases, significant differences emerge not only in the order of the most played songs, but also in the dominant albums and genres.
A central issue is the way in which Spotify defines concepts such as “album listened to”: often only the almost complete and continuous use of a disc is counted, while other systems consider the set of songs played over time. The result is that experimental albums, soundtracks or compilations with short tracks tend to be penalized or excluded, even if listened to frequently.
Active algorithms. Another key element is the role of recommendation algorithms. Spotify doesn’t just record what is chosen manually, but actively influences our music by starting automatic playlists, personalized radios and infinite playback that starts at the end of the programmed one.
These functions, designed to keep the user on the platform for as long as possible, end up driving a significant part of the annual listening. As a result, Wrapped not only measures spontaneous tastes, but also the choices suggested by the algorithm itself. In scientific terms, it is a system that observes behavior while simultaneously modeling it.
Partial portrait. The end result is an entertaining story, but often incomplete and not entirely truthful. Wrapped, in fact, end up diverging from the neutral representation of their musical tastes, transforming themselves into a synthesis built according to proprietary criteria, technical filters and product logic.
From a data point of view, what we “unpack” at the end of the year is more a narrative than a rigorous measurement of our audience, useful for entertaining and stimulating sharing, but not for providing a faithful picture of our real habits.
