Howard Huang
← All work

UX case study · Spotify onboarding · 2026

Reducing signup drop-off

Spotify asks for six things before it plays a song. I put the payoff first: a playlist built from your taste, then the account.

RoleSole designer: audit, flows, UI, prototype
Timeline4 weeks, Aug–Sep 2026
TypeSelf-directed, unsolicited
ToolsFigma, FigJam
6→0

asks before you get anything

3→0

error loops before any value

0→2

choices you make before the ask

Counted from the two flow maps. No access to Spotify's data, so this is a hypothesis, not a result.

01

Audit

Spotify's signup today

Drop-off isn't caused by length. It's length that gives nothing back.

Current Spotify signup flow: six asks, then artist selection after the account exists
Six asks first. Taste comes once, after the account exists. Open full size

Constraint: date of birth is legally required, so it stays.

ServiceStarts withMust collectBurden
YouTube MusicGoogle account and historyNothing
Apple MusicApple ID, no historyTaste
SpotifyNothingIdentity and taste
Tidal, DeezerNothing, plus a paywallIdentity, taste, payment

No major music app shows the payoff before the ask. The closest pattern is Duolingo: a lesson before the account.

02

Drivers

Why people leave, and the response to each

D1Effort with no reward

Six asks before any audio.

Payoff before the ask

D2Unknown length

No progress shown.

Segmented bar from screen one

D3Costly errors

Three validation loops, all before value.

Validate after there's something to protect

D4Asks feel irrelevant

Date of birth arrives unexplained.

Explain every ask

D5Decision fatigue

Artist grid, three-pick minimum.

Fewer options, no minimum

D6Commitment too early

Account before proof.

Prove the product first

William Chen, 24Master's student, just switched from iPhone to Android, 5–6 hours a day on headphones I gave it three artists and it played me the same stuff it plays everyone.
03

Bet

One hypothesis, written to be proven wrong

If

a new user gets a personal playlist before registering,

Then

more people who open the app reach first play,

Because

the account saves something they own, instead of gating something unproven.

Fails if

playlist → account converts no better than Spotify's welcome → account.

“You added three screens.” Yes, and each gives back. So the metric is end-to-end completion, not form completion.

Redesigned flow: genres, artists, curating, playlist ready, then sign up
Two choices first, the only ask last. Skipping both falls back to a starter mix, not an error. Open full size
04

Screens

Five steps, one ask

Genre selectionStep 1

Genres

D5Ten chips, multi-select said out loud

D3No minimum, so no error loop

Artist selectionStep 2

Artists

D5Faces first, names beneath

D3Back button to fix a pick

Curating your mixStep 3

Curating

D4A visible wait, so the result feels made for you

Playlist readyStep 4 · the payoff

Playlist ready

D4“24 songs” reads as computed

D6Lands right before the ask

Sign upStep 5 · the only ask

Sign up

D4Subtitle says why: save what you built

D3Inline errors, password toggle

Green means one thing: selected, enabled or done. The primary button never moves, and selection changes the fill so the grid never shifts.

05

Measure

How we'd know, and what's missing

App open→First playPrimary metric. End-to-end, so people who quit early still count.
Time to first playDay-7 retentionSkip rateLost-setup support tickets

Ship rule: if completion rises but day-7 retention falls, it optimised the wrong thing.

Artists ignore genre picks→Build the artist grid from chosen genres
No age gate yet→Test it up front vs. at the end
Skipping still says “picked for your taste”→Separate copy for the starter mix
Picks lost if you leave at the form→Design the abandonment path
No social sign-up→Add Google and Apple at the last step
← All work