Pandora & Hindi Music — A Bumpy Ride
Disclaimer: This and the articles that follow are an attempt at understanding the music tech industry from a Product perspective. We do not claim to be experts in the field. All we claim is to be students of music tech. We hope this series will help others understand music tech and hopefully appreciate the industry.
TL;DR: If you’re not too picky and are okay with only the most common music, Pandora might cut it for your radio needs.
Coming from a data science background, I appreciate the Music Genome Project a lot. There’s something spectacular about music not bound by genres and/or languages but bound by the kind of sound that the song produces.
I got to give it to Pandora, they are remarkably committed to boiling down music to its essence and reconstructing it into a radio service that people just like. The fact that the interface of such an intricate piece of tech is just a thumbs up/thumbs down button makes it even more fascinating.
Sometimes, it falls flat.
I’ve mentioned the ‘how’ below.
However, it’s probably not their fault. It seems to me that Pandora is most exclusively trying to fulfill a market and that market is the average US music listener, someone who just listens to something while commute or listens to music in the background while working. Who won’t want to listen to something specific.
And that was cool until they started focusing on playing particular tracks too. Now it’s just confusing to me, more on that later.
I’ll use the following method to check on the music catalogue.
- A few songs from every popular label
- A few songs from every era of music
- A few songs that I love but are highly underappreciated and mostly unheard.
- A few songs from Indipop music.
So onto my findings then. These are strictly for Hindi music, mind you.
- Music categorization is really bad.
- There is no intuitive order to how music is played.
- Plugging in old songs gave me an enjoyable radio that went on without me itching to press the skip button.
- A Mikey McCleary station gives you music that’s neither English nor Hindi. Mostly some instrumental numbers from unheard artists. I can see how Mike using a lot of instruments can result in a instrument heavy sound that fools the algorithm, but hey, you got to be better than that.
- New songs are again a no-no. Piggybacking on the Spotify post, it seems to me that music labels are not actively looking for music contracts with Global giants as YouTube does the job well. I want to talk to some one from a reputed label about it. Will tweet/post about it.
- Maati baani is another new artist that I tried. A surprisingly low ~100 people have listened to their station yet.
- Even worse is when I try to start a KK radio and it’s a disaster. First, I can’t find the right KK, so I go to similar artists and find KK in their similar artists.
- You cannot put an umbrella term “influence of Indian music” and get away with recommending anything.
- One thing I believe, again I’m speculating, is that Music Genome Project does not give weightage to artist, album names, collaborators or era of music. This leads to perhaps an enhanced feel to English Music because other parameters are well defined but for Hindi music, it becomes a disaster.
- Probably found 30–40% of the tracks I looked for.
