Q2 2026 Streaming Intelligence Report

Campaigns create
the attention.
Listener behaviour decides what happens next.

We analysed 28.1 million streams, 34 releases and 9,246 logged predictions between 1 April and 30 June 2026. This report shares what separated the releases that kept growing from the ones that faded, including a campaign that lifted an artist's daily streams by half until the day it stopped.

28.1M
Streams Analysed
9,246
Predictions Logged, 4,833 Graded
<1%
Of Algorithmic Streams From Discover Weekly

All artist, song and label identities are anonymised. Every figure is computed directly from observed daily Spotify for Artists data. No projections, no estimates, no fabricated benchmarks.

01 The Dataset

Q2 in numbers.

Based Audio ingests daily Spotify for Artists data (streams, listeners, saves, playlist activity and algorithmic surface breakdowns) and scores every artist state daily, across pop, R&B, dance and house, Afro and amapiano, Latin, folk, reggae, and rock and metal rosters.

28,125,725
streams analysed across the roster
25.2M
listener-days observed
778,924
saves recorded (2.8% of streams)
18
artists tracked daily
34
new releases in the quarter
179
songs with daily-level data
9
campaigns active during Q2
9,246
model predictions logged
31.5%
of May–June streams were algorithmic
02 Key Findings

Five findings from the quarter.

Finding 01

Campaigns create attention. Behaviour decides what happens next.

Two paid campaigns ran across the roster in Q2. Both created a clear lift in daily streams once spend began, but one release held momentum and kept growing while the other faded as soon as the paid exposure stopped. Within days of Campaign 1's spend finishing, Discover Weekly picked the catalogue up entirely unpaid. The paid exposure was temporary; the algorithmic momentum it helped create wasn't.

+48%
avg. daily streams while spend was active (Campaign 1)
−83%
below during-campaign average within 10 days of spend ending (Campaign 2)
−91%
from peak before spend had even finished (Campaign 2)
Finding 02

Two releases, two save rates, two very different quarters.

Two tracked releases sat at opposite ends of the engagement spectrum. Release D peaked above 2,000 streams a day but, saved by fewer than one in a hundred listeners, fell to single digits within weeks. Release J, saved by one in five listeners, never had a dramatic peak and never needed one: its base held steady all quarter at a level Spotify kept feeding. Exposure opens the door; saves keep it open.

0.7% vs 20.7%
save rate across Q2, Release D vs Release J
~5 vs 380
streams per day by June
Finding 03

Launch size predicted nothing.

We compared the first 30 days of every Q2 release with at least four weeks of observed data. Release B had by far the biggest launch of the quarter and had lost two-thirds of that pace by week 4. Release C logged five streams in its entire first week and was doing 7,500 a week a month later. Across all eight releases, week-1 volume simply did not predict which group a release would land in.

59,726
week-1 streams for the biggest launch, down two-thirds by week 4
5 → 7,500
week-1 streams vs weekly pace a month later (Release C)
Finding 04

Where algorithmic streams actually come from.

From May, when surface-level tracking began, 31.5% of all streams we observed were algorithmic, 7.0M streams in total. Algorithmic Mixes (2,257,323) and Radio (1,641,277) did the heavy lifting; Release Radar delivered 249,894 and Discover Weekly just 64,750. The biggest gains consistently arrived after Spotify gained confidence in a track, not on release day.

12x
Radio + Mixes delivered 12 times the streams of Discover Weekly and Release Radar combined
<1%
share of algorithmic volume from Discover Weekly, the surface artists obsess over
2.5:1
algorithmic streams out-delivered editorial-playlist streams over the quarter
Finding 05

What was predictable: measured, not guessed.

Based Audio logs a prediction for every artist, every day, then grades it against what actually happened: 4,833 graded in Q2. Weekly trend direction is learnable: our model scored 49.2% on three-state calls against a 40.8% persistence benchmark. Next-day calls are the opposite story: even our model only hit 15.1%, and naive "same as yesterday" beat it. We publish that number because it is the point: single-day stream moves are dominated by noise.

49.2%
weekly trend-state accuracy vs 40.8% naive baseline (n = 1,619)
15.1%
exact next-day accuracy. Daily moves are noise, nobody should decide on them alone
03 Discussion

Three myths the data disproved.

Myth 1

"The biggest first day wins"

The quarter's biggest launch faded to a third of its launch pace by week 4, while a release with five first-week streams was doing 7,500 a week a month later. Launch-week size did not predict which releases were still growing a month on.

Myth 2

"Discover Weekly is the goal"

Discover Weekly delivered 64,750 streams across the entire roster in Q2, under 1% of the 7.0M algorithmic streams we observed. Radio and algorithmic Mixes did the heavy lifting, and algorithmic surfaces out-delivered editorial playlists roughly 2.5-to-1.

Myth 3

"Daily numbers tell you how you're doing"

Weekly trend direction is learnable (49.2% correct against a 40.8% persistence benchmark). Exact next-day calls scored worse than assuming nothing changes. Reading meaning into a single day's rise or fall is, statistically, reading noise.

04 Conclusion

The biggest Q2 takeaway.

"Campaign spend generated exposure. Listener behaviour decided whether Spotify kept recommending."

Every paid campaign we measured lifted daily streams while spend was active, and in every case the paid streams themselves decayed rather than compounding. What differed was the handover: one campaign gave the gains back entirely, while the other seeded an unpaid Discover Weekly pickup just past quarter-end.

Read the full report

Nine pages of charts, methodology and anonymised release-level data · PDF · 2.3 MB

Download Q2 2026 Report
About this report. Based Audio analyses daily Spotify for Artists data to understand how releases build or lose momentum, when Spotify increases algorithmic exposure, and when artists should push, hold or stop spending. All figures were computed from observed data for 1 April – 30 June 2026. Identities are anonymised; days without captured data are treated as gaps, never estimated. Where a metric was only tracked for part of the quarter, we say so: surface-level algorithmic tracking switched on at the start of May, so surface figures cover May and June.