Spotify Discovery Mode

Spotify Discovery Mode Strategy

Before you ask, yes, that’s Christopher Columbus, who set sail across the Atlantic on a voyage to discover some fresh jams.

I previously wrote that using Spotify Discovery Mode was a no-brainer. I’ve heard it compared to standing up at the theater. Sure, the first person to stand gets a better view, but before you know it everyone is standing. No one is any better off than before, just less comfortable.

Of course, that was the plan all along. Spotify now pays out 30% less on Radio, Autoplay, and Daily Mix tracks from Discovery Mode participants. We’re less comfortable than before, yet we continue to opt in because ultimately we just want our music to be heard.

Spotify advertises Discovery Mode as a way to highlight important releases. The implication is that for maximum exposure, you should limit the number of tracks you opt in.

But is that true? What’s the optimal Discovery Mode strategy? That’s the question I hope to answer today.

It depends on whether Discovery Mode operates at the artist level or track level.

In other words, when a slot is available, does it say “give me some Color Theory”? If so, I’d rather it choose my catchiest song.

On the other hand, if Discovery Mode says “give me this track” then we might as well opt in all of them!

The ultimate test would be to only opt in a single track for a month and compare it to the total number of Discovery Mode streams from the previous month. I’m not willing to do that.

Instead, let’s look at the data I already have and see if streams go up as the number of tracks opted in goes down.

My song “Anyone Would” has done pretty well in Discovery Mode from the beginning:

Discovery Mode “Anyone Would”
“Anyone Would” Discovery Mode Streams

Just looking at August versus November, the stream count is nearly the same despite opting in twice as many songs in August.

Here’s another decent performer, “The Limit”:

Discovery Mode “The Limit”
“The Limit” Discovery Mode Streams

Looking at August versus November again, opting in twice as many songs got me more streams, not less.

That said, I kinda sorta think I spot a trend towards more streams per song as the number of songs opted in decreases. Which is exactly what I’d expect to see!

Here’s why:

In November when I opted in 20 songs, those were my 20 best eligible songs. You’d expect them to perform better than the next 20 songs, right?

To put it another way, my 1st choice is more likely to resonate with listeners than my 40th choice.

Ipso facto, fewer songs = better songs.

Based on these two examples, I’m inclined to think that Discovery Mode operates at the track level.

I’m going back to including all eligible songs, except perhaps the ones it recommends against:

Discovery Mode fewer streams

If you’re still not convinced, here’s the clincher: my total Discovery Mode streams.

Total Discovery Mode Streams

It looks like the more songs you opt in, the more streams you get.

The effect is large enough that, for me at least, streams per song is irrelevant. I want as many streams as I can get, so I’m going to opt in as many songs as I can!

As an aside, I thought my total streams kept declining as more artists were getting access to Discovery Mode, but that doesn’t explain the surge in December.

Perhaps a better explanation is that the pool of new releases is larger in September and October, which is prime album release season. Those album tracks become eligible for Discovery Mode after 30 days, which explains the high demand in November. And then December is the slowest month of the year, so there are more slots available.

Then again, those November tracks are most likely still eligible in December. Perhaps there’s another explanation. Your guess is as good as mine!

I’ve already got 6K Discovery Mode streams in January with more than a week left to go, so December doesn’t appear to be an outlier.

Are you using Spotify Discovery Mode? Have a different take on my data? Share your thoughts in the comments!

4 Comments

  1. Hey Brian,

    Thanks so much for the thought provoking post!

    I hadn’t thought too much about trying to optimize Discovery Mode. I just opted in all the tracks I’ve released on Spotify and hoped for the best. But after reading your post I think that there’s value in trying to get some understanding of what’s going on with that algorithm and even if it doesn’t matter what choices we make, see what useful information we can pull from that data.

    But in order to do so, I think a lot more data is required. And as I looked at my data in comparison to yours an idea struck me. An idea that you’re well positioned to take up, given your blog’s audience…. You could, if you wanted, send a simple survey to your Passive Promotion mailing list and collect data from other artists and see what trends, if any, emerge. You could structure the survey like: April 2023 # songs, April 2023, # streams, etc.. which would result in fairly structured data. Which is the first possibility that occurs to me, although there may very well be other, better methods. Anyhow, that could make a really interesting followup post!

    I’ll share my data below. I haven’t released my back catalog to Spotify, so I only have the tracks I’ve released over the past 2 years to work with. But what I notice is: Roughly the same number of streams per month as you reported – despite less tracks. Unlike you, I had a huge dip in December. January is gonna be a bit better, but nowhere near as good as October and November. And here’s a possible theory on that… I did a lot of successful promotion of a couple current tracks in Oct. And I turned it all off for December and am just starting to turn it back on now. My guess is that there might be a correlation between overall activity and Discovery Mode.

    The other interesting thing I notice is that I have a big percentage of the streams coming from one or two tracks. In October, half the streams came from one track. Interestingly, that was an early track I released which I never really promoted – no real Facebook ads or playlist promotions, or much of anything. But it’s not always the top track. There’s been 4 top tracks so far. (I’ve included a keyword from the title in my data below, in order to see what’s happening.)

    Cheers! – Brock

    April 2023 9 songs, 8183 streams, top track (Guy) 1598
    May 2023 9 songs, 6201 streams, top track (2nd) 1430
    June 2023 10 songs, 7051 streams, top track (2nd) 1235
    July 2023 10 songs, 7926 streams, top track (Choose) 1482
    August 2023 10 songs, 7414 streams, top track (Choose) 1365
    September 2023 11 songs, 7445 streams, top track (Guy) 2025
    October 2023 11 songs, 9384 streams, top track (Guy) 4408
    November 2023 12 songs, 9075 streams, top track (Guy), 2749
    December 2023 12 songs, 4991 streams, top track (Guy) 1095
    January 2024 12 songs, 5969 streams so far, (Get), 1175

    1. Thanks for the detailed and thought-provoking comment Brock!

      Those are some really nice numbers! So cool that your top song got 4.4K streams in October, and that you’ve got multiple “hit” songs.

      As you can see from my data, I don’t have any big winners. There are plenty of losers though, that only see a handful of streams per month.

      You saw peaks in October and November where my lulls were, so there goes my theory on prime album release season.

      As for my primary theory (more songs = more streams), your data can’t tell us much, since you’ve opted in roughly the same number of songs every month.

      So you’re right – more data required! But that sounds like a lot of work ;). Unless I see something that would persuade me not to opt in all eligible tracks, I’m willing to live with a bit of mystery.

  2. I hear ya, Brian, we’ve all got enough things on our plate! 🙂

    But I don’t think it would be that much work to run a survey and who know what we might learn…. So tell you what, I’ll offer to create the survey, write the copy for the email that explains it, and analyze the aggregated results. All you’d have to do is send the email with the survey and hopefully share any of your conclusions in a blog post so that people who participate in the survey get something back for their effort. Any interest? (If not, no worries – I appreciate ya!)

    1. That’s very generous of you Brock! I’ll definitely keep it in mind. Maybe enough readers will share their results in the comments that it won’t be necessary (hint, hint!).

Leave a Reply

Your email address will not be published. Required fields are marked *