Most collectors stare at floors.
It's not the worst metric. It's just the laziest one. By the time a song token has a high floor, the market has already figured it out and you're paying retail.
The interesting signals are the ones that don't show up on the OpenSea card.
What floor actually tells you
The floor is the lowest current ask. That's it. It's not the last sale price, it's not the average, it's not a value estimate. Someone listed at $X and nobody beat them down.
A low floor can mean a few different things. Nobody wants it. Nobody knows about it. One impatient seller dropped a single listing well under the rest. Reading which one applies is the whole game.
Signals that beat the floor
- Stem burn ratio for the song. If 80% of the song's stems have been burned and only six song tokens exist, the next forge is a low-probability event. Scarcity is already locked in.
- Recent transfer count on the song token. Zero transfers in three months while stems are still being burned means new song tokens are sitting in wallets, not flipping. Holders are choosing to keep, which is usually a buy signal.
- Artist activity around the song. New social posts, mixer demos, mentions in interviews. Attention tends to come before price moves, not after.
- Distance from album completion. If a song is one of the last few needed to complete an album, song token demand from album-chasers will spike as soon as the rest of the album fills out.
How to actually check each signal
Most of this is free if you know where to look.
Stem burn ratio: go to the contract on Etherscan, find the stem token IDs for the song, check totalSupply on each, compare to the original mint count. Burned percentage equals (mint count minus current supply) divided by mint count.
Recent transfer count: the Token tab on the stems.fm contract on Etherscan lists transfers in chronological order. Filter by the song token ID and count what's happened in the last 30 days.
Artist activity: follow Kyler on X and watch the stems.fm Discord. Most of the relevant signal lives in those two places, not on OpenSea.
Album completion distance: count the songs on the album, count how many have at least one song token forged. The closer to all songs being forged, the closer the album burn becomes possible.
The trap most collectors fall into
Buying songs they personally like and then telling themselves it's a value play.
It can be both, but you have to be honest about which one is driving the call. A song you love at any floor is fine, that's a personal-collection move. Calling it undervalued just because you like it is how people lose money.
The check: would you buy this song token if it were by an artist you didn't recognize? If no, you're buying with your ears, which is a different decision.
What real undervaluation looks like
All three at once is the strong signal:
- Floor is below the cost basis of forging from scratch (sum of stem floors plus gas)
- Stem burn ratio is high (over 50% of stems consumed)
- Recent transfer history shows holders staying put, not flipping
When those three line up, you're looking at a song token that's structurally cheap by the chain's own numbers. Doesn't guarantee a price move, but it puts the math on your side.
When to walk away
Songs with low stem burn ratios? Early, not undervalued. Could go either way. The market hasn't decided yet.
Songs with falling floor and falling transfer count? Abandoned. Different signal entirely.
Songs with high floor and high transfer count? Correctly priced and active. Buy them if you want the song token, just don't pretend you're getting a discount.
One more thing
Spend 30 minutes a week running this check across the catalog instead of three hours staring at one floor. The math is the same, the coverage is wider, and you'll find one or two real opportunities a month.
Bookmarks worth keeping
- The stems.fm contract on Etherscan (0xA62f445A1171fCfBf76eb44a866c7b48F1a6f226) for burn counts and transfers
- OpenSea collection page for floors
- Stems.fm /collection page to compare stem counts per song
- Whitepaper: "Stems: A New Layer for Valuing Music" by Kyler Simzer for the framing
Look at the chain numbers, not the marketplace card.
