Skip to content
Clark Hayashi

Side Quests

Listening

Four years of my own Spotify history, counted honestly

The record so far

Streams
79,764

plays that cleared 30 seconds

Hours
4,199

175 days of audio across 1,414 elapsed

Artists
3,360

across 11,092 distinct tracks

Skipped
30.8%

all time, and rising: 25.4% in 2023, 50.9% so far in 2026

2022-09-17 to 2026-08-01. Audio adds up to 12.4% of every hour that passed in that window, sleep included.

Four years, by month

2023202420252026 196h
Hours listened Listening from outside the US
Hours per month. The peak is 196 hours. The two partial months at either end of the export are left out so the edges do not read as a collapse. The strip beneath marks months with real listening from another country, which is why the next chart needs timezone handling at all.

When I actually listen

How I listen, not just what

skipped shuffle 2023202420252026 100% 50%
Two things move here, in opposite directions. Shuffle falls from 92% of plays in 2023 to 59% in 2026. Skipping climbs the other way, from 25.4% of starts to 50.9%. Less shuffle and more skipping together read as hunting for something specific rather than letting a playlist run. Neither is visible in the flat all time figure of 30.8%, which is the reason this chart exists.

Finding new music

2023202420252026 225
Artists heard for the very first time, by month. This is the number I most wanted to see, because it is the one a year end summary never shows you.

The big six, month by month

One panel each rather than six lines on shared axes. Six overlaid series would need six colours that stay separable for red-green colourblind readers, and this palette is deliberately muted, so the colours would fail before the chart did. All six panels share a scale, so the heights compare directly.

Drake

6,527 streams · peak 318 in Aug 2023

Jack Harlow

1,544 streams · peak 160 in Mar 2026

TWICE

1,515 streams · peak 195 in Jul 2023

LE SSERAFIM

1,430 streams · peak 130 in May 2023

NewJeans

1,422 streams · peak 171 in Jul 2023

SZA

1,182 streams · peak 123 in Dec 2022

Most played, all time

Tracks

  1. 01

    Pipe Down

    Drake

    294
  2. 02

    Passionfruit

    Drake

    266
  3. 03

    Spin Bout U

    Drake

    258
  4. 04

    Feel No Ways

    Drake

    242
  5. 05

    WAIT FOR U (feat. Drake & Tems)

    Future

    209
  6. 06

    Super Shy

    NewJeans

    189
  7. 07

    Out of Time

    The Weeknd

    188
  8. 08

    Luv Is Dro (feat. Static Major & Bryson Tiller)

    Jack Harlow

    186
  9. 09

    Hold On, We're Going Home

    Drake

    177
  10. 10

    Ditto

    NewJeans

    175

Artists

  1. 01

    Drake

    328 tracks · 396 hours

    6,531
  2. 02

    Jack Harlow

    76 tracks · 68 hours

    1,544
  3. 03

    TWICE

    119 tracks · 80 hours

    1,515
  4. 04

    LE SSERAFIM

    74 tracks · 67 hours

    1,430
  5. 05

    NewJeans

    27 tracks · 63 hours

    1,423
  6. 06

    SZA

    68 tracks · 63 hours

    1,182
  7. 07

    The Weeknd

    71 tracks · 68 hours

    1,104
  8. 08

    beabadoobee

    45 tracks · 47 hours

    970
  9. 09

    Bryson Tiller

    57 tracks · 36 hours

    755
  10. 10

    keshi

    63 tracks · 35 hours

    731

Counted by plays, from the raw export. Ranking by hours instead reorders this list, which is the honest problem with any single top ten: a three minute song played twice beats an hour long mix played once.

Right now

Checking Spotify

 

 

Recently played

  1. Loading

Current rotation

Tracks

  1. Loading

Artists

  1. Loading

This block is live from the API and is a different measurement from the all time list above: Spotify returns its own affinity ranking here, not a raw play count. The three windows shown are the only ones the API offers.

How this is counted

A play counts at 30 seconds. That is Spotify's own bar for paying a rights holder, and it keeps skips and mis-taps out of the totals. It also means 30.8% of starts are excluded, which is a large enough share to report rather than quietly drop, and a moving enough one to chart.

Times are local, and that took work. The export stamps every play in UTC. Over these four years I listened from the United States, Japan, Italy, Canada and Singapore, so neither raw UTC nor one fixed offset is correct. Each play is converted using the country the export recorded. The one case that field cannot settle is the United States, which covers both Seattle and Honolulu, three hours apart, so the summer 2025 stretch in Honolulu is mapped separately. An hour-of-day chart built without this is quietly wrong.

Rankings need tiebreakers. Ordering by play count alone left tied rows in arbitrary order, so two runs over the same export produced different top tens. Every ranking now breaks ties on artist and title, and the output is identical run to run.

Podcasts and audiobooks are excluded. Across the whole export they come to 195 rows, next to 115,286 music starts, so this is a music history rather than a listening history.

Everything except Right now and Current rotation is computed from the raw export with DuckDB. Only the aggregates are published. The raw file stays on my machine, since it records where I was and what I was doing, minute by minute, for four years.

Listening data from Spotify.