The real data for, say, an artist would be the actual, physical artist. What we have is data about the artist, we don’t actually have a digital copy of any living (or dead) artists in our system(!), but we do have data about these artists. That is, we have metadata about the artists.
The real data for some audio/music would be the actual audio/music—we don’t store music in our system, so we can’t have the actual audio (this would be a copyright/licensing hell), but we can store data about the audio/music. This would be metadata about the audio/music.
So our data is metadata about other data.