Spotify Wrapped is the big news today (which rather raises questions about the definition of “big news” but that’s another post). It’s a fun summary of what we’ve listened to over the last year. It was nostalgic to see that my favourite artist this year – Billy Joel – is the same artist I obsessed over as a kid.
There are global stats as well. The Guardian reports that Taylor Swift was the most streamed artist of 2023. But what, exactly, does “most streamed” mean? Does it mean streamed for the most minutes? Streamed the most times? Streamed by the most people? Streamed on the most days? What even IS a stream? How much of a song has to be streamed before it counts as a stream?
It turns out that both popularity AND revenue are determined by number of songs streamed. Taylor Swift apparently figured out some time ago that shorter songs would maximise her streaming ranking and her income, and many artists have followed suit, causing average song length to fall significantly over the last few years.
So now we know that “most streamed artist” means most songs streamed, but what, exactly, does “streamed” mean? It turns out that Spotify defines a stream as 30 seconds or more of a song. From Spotify’s point of view, if I play 29 seconds of a song I haven’t listened to it at all, but if I play 31 seconds I have listened to the whole song. At least, for counting purposes.
Skipping before the end of a song might impact the chances of the algorithm playing you that song again, but we don’t know that for sure, because the algorithm is a secret (and very likely changes over time, to maximise Spotify’s revenue and minimise the amount they have to pay artists).
The way we define the data we measure can completely change the results that we get. I’d be fascinated to know which artist was the most streamed on Spotify by minutes streamed, rather than number of songs. Sadly I don’t think Spotify is particularly interested in giving us real data. They just want to give us the kind of data that keeps us on their platform for longer. Sparkly, tasty data, high in clickbait and low in information.
Many of my guests on Make Me Data Literate Podcast have highlighted that definitions can dramatically change the data, so now every time I hear a statistic mentioned, I get curious about the definition underlying it. It fascinates me, because we tend to think of data as simple, clear, and well defined. How many kids sat Naplan this year? That must surely be a straightforward number! But, like most data, it turns out to be surprisingly complicated underneath.
We have an intuitive understanding of the word “most” that hides a thousand potential quirks in the definition. Most doesn’t necessarily mean what we think it means. Any plain language description of a value is incredibly unlikely to map onto a single, unambiguous definition of the data that’s actually been collected.
So next time someone throws a statistic or a graph at you, see if you can find out how the data is defined. It might change everything!