How to read your streaming data: The six numbers that tell you what to do next

You open Spotify for Artists on the bus. The big number says 2,314 streams this week. You feel something — good or bad depending on what last week said — and you close the app. That is the entire data practice of most independent rappers, and it's about as useful as checking the weather by reading the thermometer in your fridge.

The stream count is the least useful number on the page. It's in the biggest font because it's the one that makes you check again, and it tells you almost nothing about what to do next. The dashboard behind it does. Spotify for Artists and Apple Music for Artists expose, for free, the numbers labels used to pay analysts to assemble: who's listening, how they found you, whether they came back, where they live, and whether the growth is real.

This is a plain-English tour of six of those numbers and the one decision each one drives. Tab names and layouts change with every app update, so we'll describe what each number is rather than where the button sits this month. Labels vary by platform and version; the mechanics don't.

Streams are the least useful number on the page

A stream count is the product of two things: how many people heard the song, and how many times each of them played it. Streams = listeners × streams per listener. When the big number moves, it can't tell you which side moved — and the two sides mean opposite things. Look at these six instead:

1. Listeners. Unique people who played you in a window, usually the trailing 28 days. This is reach.
2. Streams per listener. The repeat rate. This is whether they came back.
3. Saves per listener. How many of the people who heard it kept it. This is conversion.
4. Source of streams. Where the plays came from. This is how fragile the growth is.
5. Follower delta. Net new followers this month. This is the asset that compounds into the next release.
6. Top cities. Where the listeners physically are. This is booking data.

Everything below is one of those six, read properly.

Source of streams, decoded

Both dashboards break your streams down by where they started. The categories wear different labels depending on the version, but they describe five kinds of listening:

• Your profile and catalog. Someone searched your name or opened your page and pressed play. The purest signal of intent.
• Listeners' own playlists and library. They saved the song or dropped it into a playlist they built. They now own a copy of your song in their world, and they'll hear it again without any platform deciding to show it to them.
• Algorithmic playlists and radio. Release Radar, Discover Weekly, song radio, autoplay — the platform's machine chose to put you in front of someone. Powerful, and rented.
• Editorial playlists. A human at the platform added you. A gift, and usually a temporary one; songs rotate out.
• Other listeners' playlists. Playlists built by strangers — independent curators, fans, that one guy with the 4,000-follower workout mix.

What a healthy mix looks like depends on size, and none of these are official thresholds — they're what we'd want to see. At around a thousand monthly listeners, most of your streams should come from your own profile and from listeners' libraries and playlists: your people, found through your posts and shows. Algorithmic share will be small and that's fine. At around ten thousand, algorithmic share grows, because the save-and-playlist behavior of the first thousand taught the machine who to show you to; what you watch is whether the owned share — library, listener playlists, profile — keeps growing alongside it. At a hundred thousand, you want no single source carrying more than half the song.

The failure mode is "mostly algorithmic." A song whose streams are, say, four-fifths Release Radar and radio is living on a recommendation that can be withdrawn next week, by a machine that never met your fans. Nobody chose that song. The data looks great until the feed moves on, and then there's nothing underneath it. The rollout guide explains why listener playlist adds and saves are the strongest organic signals; this is the dashboard view of the same idea.

Repeat rate and save rate, with the arithmetic

Two songs, same stream count, completely different futures.

Song A: 4,000 streams from 3,900 listeners. Streams per listener: 4,000 ÷ 3,900 ≈ 1.03. Almost nobody played it twice. That's the shape of a song that ran once inside a big playlist or an algorithmic feed and was never sought out again.

Song B: 4,000 streams from 1,200 listeners. Streams per listener: 4,000 ÷ 1,200 ≈ 3.33. The people who found it played it three times on average. That's a song with hooks in people.

Now layer in saves, which both dashboards report. Song A got 40 saves: 40 ÷ 3,900 ≈ 1% of listeners kept it. Song B got 84 saves: 84 ÷ 1,200 = 7%. Seven times the conversion rate from under a third of the audience.

The rule this produces is the most useful sentence in the guide: promote the song that converts, not the song you love most. Ad money, curator pitches, the second video, the SongBattle entry — all of it goes to Song B, because every new listener you push to it has a seven-in-a-hundred chance of becoming a fan instead of a one-in-a-hundred chance. Your favorite song is a feeling. Your converting song is a fact.

Nobody outside the platforms knows what save rate or repeat rate the recommendation systems actually reward, and anyone quoting a magic number is guessing. Compare your songs to each other, not to a benchmark someone made up on a forum.

The spike-and-cliff diagnostic

The daily streams graph tells you what kind of growth you're looking at before you read a single number.

A playlist add looks like a wall: a vertical jump the day the song was added, a plateau while it sits in the list, and a cliff the day it rotates out. Impressive on screen, borrowed in reality.

A fan base looks like a hill: lower, bumpier, rising slowly, with small spikes on the days you posted, played a show, or got shared. Less exciting, entirely yours.

To tell them apart on a new single, run the 28-day decay check. Average your daily streams over days 1–7, then over days 22–28, and divide.

• Single X: week one averaged 400 streams a day; week four averaged 40. Retention: 40 ÷ 400 = 10%. Ninety percent of the launch was borrowed — a playlist, a feed, a burst of friends. Stop spending on it, and build the next one.
• Single Y: week one averaged 200 a day; week four averaged 120. Retention: 120 ÷ 200 = 60%. Most of the launch audience stuck around. Keep spending: more curator pitches, the second visual, the live clip, the open-verse challenge.

Half the artists who read this will discover their "disappointing" single is the healthier one. A song that peaks lower and holds is worth more than a song that peaks high and vanishes, because the first one is building something the next release can stand on.

What Apple Music for Artists adds

Apple's dashboard covers the same ground — plays, listeners, plays per listener, city and country breakdowns — with one addition that deserves its own habit: Shazam counts by city.

A Shazam is a different kind of signal from a stream. Someone heard your song out in the world — in a car, a barbershop, a DJ set, under somebody's video — and reached for their phone to find out what it was. It's intent captured at the moment of discovery. A Shazam spike in a city you've never played means someone there is playing you, and that city just moved up your booking list.

Read Apple's numbers as a trend line rather than a total. As of this writing Apple Music has no free ad-supported tier, so its audience is almost entirely paying subscribers, which tends to make the counts smaller and the behavior different from Spotify's. Don't compare Apple totals to Spotify totals; compare each platform to itself last month.

City data is booking data, demographics are ad data

The most money-shaped numbers on either dashboard are the ones most artists never scroll down to: top cities.

Each month, write down (or export, where the dashboard allows it) your top ten cities by listeners. Then run the listeners-per-city vs. draw sanity check, because monthly listeners are not ticket buyers. An illustrative rule of thumb: a small single-digit percentage of a city's monthly listeners will show up to a show. Two thousand monthly listeners in a city, times 1–2%, is 20 to 40 people — which happens to be exactly the kind of honest draw number that the booking guide's promoter email is built around. Now the sentence "draws 30–40 in [City]" has data behind it, and you can say where that number came from if a promoter asks.

Use the cities to route. Your top three within driving distance become the regional run; the city with the Shazam spike becomes the support slot you pitch for; your hometown number tells you whether you've exhausted it or barely touched it.

Age and country splits are for a different decision. An age range doesn't fill a room, but it does tell you which platform to post on and when, and a country breakdown tells you where it's worth spending ad money and where it isn't. Demographics decide where your attention and your ad budget go. Cities decide where your car goes.

Once you have a top-cities list, the natural next move is to start collecting those people directly — the email and text list guide picks up there.

Spotting fake streams in your own account

Bought streams don't only happen to people who buy them. A "promo service" a friend paid for, a playlist that turned out to be botted, or occasionally someone trying to sabotage a rival can land artificial plays on your song. Your dashboard will tell you, if you look:

• Listeners in countries you've never touched, appearing overnight, with no post, show or playlist add that explains them.
• Streams per listener pinned at almost exactly 1.0, with near-zero saves. Real people repeat and save at some rate; bots play once and leave.
• A playlist with thousands of followers and no visible engagement showing up as a top source — no likes, no activity, a cover made in five minutes.
• A spike with no cause. Real spikes line up with something you did. Fake ones line up with nothing.

What to do is counterintuitive: don't celebrate, and don't wait. Screenshot everything, then contact your distributor before the platform does. Platforms strip artificial streams from royalty reports and can pull tracks or penalize the account, and distributors can and do take down releases with suspicious activity. A distributor that hears from you first, with receipts, treats you as a victim; one that hears from the platform first treats you as a suspect. The scams guide covers the whole botted-playlist economy and what it does to a catalog — the short version is that the vanity number isn't worth the risk to everything you've released.

The profile tools, briefly

Both dashboards double as a storefront. The free tools are worth ten minutes a month: an artist pick pinned to whatever you're pushing, a Canvas loop on your current single, a fresh bio and photos, merch links where available, and concert listings pulled from the ticketing partners your shows are sold through. A profile that looks active gets a second click from the listener your data just earned.

Then there are the paid and promotional tools, which come in two shapes. One is a budget-based campaign — you pay to have your release shown to listeners who've engaged with you before. The other trades a reduced royalty rate on the tracks you opt in for extra algorithmic consideration. Neither is a scam and neither is free; one costs cash and the other costs a slice of the royalty on every stream it touches.

We're not going to tell you to use them or not. We'll say this: read the current terms before you opt into anything, and only test either one on a song that already converts by the math above. Promotion multiplies what a song does on its own. Multiplying a song with a 1% save rate buys you a bigger number of people who didn't care.

For the royalty side of all this — what a stream actually pays and the floor a track has to clear before it earns anything — the income breakdown already has the numbers, so we'll leave them there.

The monthly dashboard

Six rows, updated on the first of the month, five minutes once the habit sets. One line each, with the decision it drives. The cutoffs are personal benchmarks you set against your own history, not platform rules.

1. Listeners (28-day): this month ___ · last month ___ · Decision: down two months running means release or post; up means find the source and do more of it.

2. Streams per listener: ___ · Decision: near 1.0 on a new single means it's background music somewhere; comfortably above 2 means it's sticking — push it.

3. Saves per listener: song A ___% · song B ___% · Decision: promotion money goes to the highest; spending on the lowest stops.

4. Source mix: profile/catalog ___% · library and listener playlists ___% · algorithmic ___% · editorial ___% · Decision: if algorithmic is over half, spend the month building the owned share (saves, your own playlists, the text list).

5. Follower delta: +___ · Decision: flat during a campaign means the content entertained but didn't attach to you — more face, more story; every follower is guaranteed reach on the next release.

6. Top three cities: ___, ___, ___ · Decision: update the draw number in the promoter email; pick the next run.

What "good" looks like isn't every row going up. Listeners rising with streams per listener steady means reach is growing. Listeners flat with saves and repeats rising means the song is deepening with the fans you have. Both are wins. The month to worry about is the one where listeners spike, saves don't move, and the source is a playlist you've never heard of.

Pair the dashboard with a number nobody can buy

Every number in this guide can be gamed by someone with a bot farm and a credit card, which is why the smart reading of streaming data is skeptical by default. There's one metric in your toolkit that a stream count can't fake: real people, verified one by one, choosing your song over the others on the ballot. A public competition result sits next to your dashboard as the receipt — proof that the saves and the repeat rate were humans, not a feed.

So take the song your data just told you converts, and put it where people vote on it. Enter the competition with the record that earns its saves, and let the leaderboard be the seventh number.

—

Related:

• TikTok for rappers: The platform playbook nobody explains straight
• How to read the SongBattle leaderboard: What the numbers actually mean (and what they don't)
• Your private stats page: What the four numbers mean and how to use them during a competition run