The follower gap: +744 listeners, +9 followers

August 21, 2026 · sauvachi

Two numbers I have no obligation to publish: 139,602 monthly listeners and 1,755 followers. The second is the one that belongs to me; the first is rented. This is a reading of the gap between them. Nothing about publishing the first one changed the outcome, and I am not going to pretend otherwise.

The two readings

  • Entry one: 139,602 monthly listeners · 1,755 followers
  • Entry two, one month later: 140,346 monthly listeners · 1,764 followers
  • Change: +744 monthly listeners · +9 followers

The conversion rate — followers as a share of monthly listeners — was 1,755 ÷ 139,602 = 1.257%. It is now 1,764 ÷ 140,346 = 1.257%. To three decimal places, in a month, on a catalogue that gained three releases this summer, it did not move.

The more interesting figure is the marginal one. Of the 744 listeners the reach grew by, 9 became followers. That is 1.21% — within rounding distance of the 1.26% baseline. The new audience is converting at exactly the same rate as the old audience. Whatever is producing follows is a fixed property of the listening experience, not something that responds to a journal post about it.

Why these two numbers are not the same kind of number

Worth being precise, because people compare them as if they were. Followers is a cumulative stock: it only goes up unless somebody actively unfollows. Monthly listeners is a rolling 28-day window — a flow measurement that resets its own denominator every day. The +744 is not 744 new human beings. It is the net difference between everyone who entered the window and everyone who aged out of it. If a playlist placement lands and then decays, that number can fall by thousands without a single person deciding they dislike the music.

That asymmetry is the whole argument. Reach can be taken back by an editorial decision made by somebody who has never heard of me. A follow cannot. Followers are the population that gets the release into Release Radar and gets the new-music notification on release day. Everything else is weather.

Publishing the gap did not close the gap

I want this stated flatly because the temptation with a public ledger is to only publish it when it flatters you. Entry one was read, it was shared, and it moved the follower count by an amount statistically indistinguishable from doing nothing. Nine follows in thirty days is the background drift of a catalogue that is being discovered passively. It is what happens when nobody is asked.

That is the actual diagnosis and it took a second data point to see it. The problem is not that the ask is badly worded or badly placed. There has been no ask. Listeners arrive through algorithmic surfaces — radio, autoplay, an editorial slot — hear a track, and leave, because at no point in that path does anything suggest a next step. A 1.26% conversion rate is roughly what you get from the fraction of people who go looking for the follow button unprompted.

What gets tested before entry three

One change, isolated as far as it can be isolated, so the next reading means something:

  • Follow before play on the release link. The default action on a smart link is listen. For the next release cycle the follow action goes above it, on the link and on sauvachi.com, on the theory that the moment of highest intent is the second somebody has just decided to open a new song and has not yet heard it.
  • Ask on the site, not only on the platform. The site already hosts the catalogue and the player. Anyone reading this page is further down the funnel than anyone hearing a track on autoplay, and there is currently nothing here that converts that.
  • No change to release cadence. Cadence is the variable I would most like to change and the one I have to hold still, or entry three tells me nothing.

Here is the number that makes it falsifiable. Baseline drift is 9 follows a month. If entry three reports anything under about 30, the test failed and I will say the test failed. If it reports 100, the mechanism is real and the previous year was a self-inflicted wound. Anything between is noise and I will call it noise rather than a trend.

The rules of the ledger

Naming it so it has somewhere to live. The Follower Gap Ledger publishes both figures roughly monthly, whichever direction they went. The rules I am binding myself to:

  • Only numbers off the dashboard on the day of writing. No estimates, no projections, no per-track figures dressed up to look like totals.
  • Both numbers every time, and the change since the previous entry, including when the change is embarrassing.
  • The named test and its result. If a mechanism does nothing, that gets published too — a negative result is more useful to somebody reading this than another post about consistency.
  • No revenue claims attached to listener counts. Payout varies by country, by service and by deal, and anyone quoting a single per-stream rate as universal is guessing. Streams are streams; money is a separate conversation.

Almost nobody at this scale shows the working. The numbers get rounded up in a bio, or they appear once in a pitch deck and never again, which makes them impossible to check and therefore worth nothing. Two dated readings a month apart are worth something, because the second one can contradict the first. This one nearly did nothing at all, and that is the finding.

Entry three in roughly thirty days.