Spotify Recently Played: Pin, Remove, Pause. Three reversible controls for a shelf that records everything.
Recently Played is where most listening starts on shared screens, and the shelf records every play with no controls. Community forums document users force-quitting tracks to push favorites up; the workaround is the evidence. The hard part was deciding what not to ship: global history delete, recommendation tuning, and private mode all pulled toward a larger system.
Role
UX/UI Designer
Type
Solo · Concept
Timeline
Nov to Dec 2025
Methods
Constraint mapping · Community signal synthesis
Outcome
Three-sprint concept plan with decisions log + hypothetical test synthesis.
Three actions. All shelf-native. All reversible. No settings menus, no buried toggles.
01
Premise & Constraints
Recently Played sits at the top of every Home feed and updates automatically from listening history. One of Spotify's highest-traffic surfaces. As a daily user, I started noticing moments the shelf wasn't designed to handle. An accidental tap that lingers. A favorite I want within reach. A borrowed phone with my history still logging. The shelf is always visible, on your screen and on anyone else's. That's the opening.
Today. The shelf records every play and orders by recency. That's the entire interaction model. Workarounds exist: community forums document users playing a track for one second and force-quitting to push it to the top of the shelf. That kind of behavior is signal. People are reaching for affordances that don't exist yet.
Concept. Three additive controls, all native to the shelf, each completing in 1 to 2 taps. Pin to keep a favorite within reach. Remove to hide an accidental or contextual listen without deleting it from the library. Pause to stop the shelf from logging for a time-boxed session. Reversible by default. No buried settings, no global toggles, nothing that touches the recommendation engine.
Constraints. The rules every decision had to honor:
Preserve discovery. The shelf exists to surface what users want. Any control that weakens that signal is out of scope.
ML training signal integrity. Permanent pause was ruled out. Time-boxing is what survived.
Per-device, not global. Remove on one device shouldn't propagate. Privacy at the device boundary.
Reversible only. No destructive deletes. Every action has an undo.
Shelf-native. No settings menus. The control lives where the action happens.
These came from mapping stakeholder priorities (Product, ML, Privacy/Legal, Engineering) before any visual design happened. They turned the obvious first idea, global history clearing, into a non-starter and pointed toward the three that remained.
What users said
“Recently Played is literally the ONE AND ONLY feature of the app I use.”
Reddit, Oct 2024
02
Research & Discovery
No live users, no internal data, no brief. I built the research layer from public signals: Spotify community discussion threads, App Store reviews, UX forum posts, and a competitive audit across seven platforms. I used AI-assisted synthesis to cluster behavioral themes from 200+ community posts. It cut the analysis time significantly, but the judgment calls were still mine.
Spotify Personas
These three personas are synthesized from public sources: Spotify Community forum threads, App Store reviews, and Reddit discussions of the Recently Played shelf. The framing was also informed by conversations with practitioners inside the music streaming industry. I treated them as a research lens, not a research finding. They represent the behavioral range visible in public discourse, and the tension between them is what made the design hard: Melanie needs social curation, Dave needs fast access under time pressure, and Stephen tells you where the feature boundary is.
Melodic Melanie
Primary Persona
“Why pay for it when it’s free?”
Goals
Inexpensive access to music for herself and for sharing with friends.
Pain point
Recently Played surfaces accidental or contextual plays when she shares her screen. No way to clean it up before someone sees it.
What they need from this design
Pin to surface favorites she’d actually share. Remove to clear plays she wouldn’t.
Design tension: Melanie is the social pressure case. If the shelf feels curated, she’s comfortable sharing. If it doesn’t, she stops showing her phone.
Show 2 more personas (Secondary & Negative)
Ranger Dave
Secondary Persona
“You should pay for music to support the artists.”
Goals
Supporting music creatives. Listening without interruption from ads.
Pain point
The shelf is noisy from commute listening and his son’s music. He can’t quickly get back to what he actually wants to hear.
What they need from this design
Pin for fast access to go-to albums. Pause to protect recommendations during his son’s listening sessions.
Design tension: Dave is the time-pressure case. He has 45 minutes on BART. Every tap matters. If the control takes more than two steps, he won’t use it.
Stephen Tan
Negative Persona
“I stream my music to hear it first before committing to buy the ones I like.”
Goals
Listen to a small set of favorite songs. Keep hassle to a minimum.
Pain point
None related to Recently Played. He doesn’t engage with the shelf or discovery surfaces.
What they need from this design
Nothing from this feature set. Stephen defines the boundary.
Design tension: Stephen is the person these controls are not for. He validates the scope: if the design started trying to serve Stephen, it would mean the feature was too broad.
Competitive Audit
I audited seven platforms to map where the category sits. YouTube Music is the closest reference point with three of the four controls. On-shelf pinning is the open space. No one in the audit offers it, which made it worth designing for.
Platform
Remove
Clear All
Pause History
Pin to Shelf
YouTube Music
✓
✓
✓
✗
Spotify (current)
✗
✗
✗
✗
Spotify (proposed)
✓
✓
✓
✓
Point of Reference
User Journey Map · Ranger Dave · 7 Stages
The gap between seeing a control and trusting it.
Mapping Ranger Dave’s 7-stage journey surfaced the highest-friction moment: Stage 3 to Stage 4, from spotting a long-press affordance to committing to an action. No visual cue that the gesture existed. The shelf gave no signal that anything was interactive. That gap drove the decision to design for immediate discoverability, not power-user access patterns.
03
Design Decisions
Pin, Remove, and Pause came out of a longer list. Early directions included global history clearing, a private listening mode, and surfacing controls from Settings. Each was ruled out: too broad in scope, too deep in the navigation stack, or too likely to degrade recommendation signals. The three that remained could live on the shelf, complete in 1–2 steps, and reverse without permanent consequence.
One rule applied to every option: don't break discovery.
Mid-project pivot
I originally led with Pin, the differentiation play. Mapping stakeholder priorities re-sequenced it. Community threads showed people asking for control over accidental and contextual plays, not curation. Remove moved to the front. That re-sequencing changed the entire build order.
Stakeholder Priorities: Mapped Before Any Decision
Product
Differentiation + engagement lift
Pin is the greenfield feature; Remove + Pause reach parity first
ML / Recs
Training signal integrity
Permanent pause ruled out. Time-boxing is the constraint that survives
Privacy / Legal
Data handling scope
Remove is device-scoped; global deletion requires a different flow
Engineering
Sync model complexity
Pin capped at 4; chip-based state avoids a heavy backend rewrite
The Design
Three controls, one surface
Every control lives behind the same long-press gesture on the Recently Played shelf. No settings menus, no buried toggles. Each action completes in 1–2 taps and every destructive action includes an Undo safety net.
Pin
Lock a favorite to the front of the shelf. No competitor offers on-shelf pinning. The item moves to position one and stays there until unpinned.
01
Long-press any item on the shelf. The action sheet surfaces all controls in one place.
02
Select Pin on top. Microcopy clarifies: removing hides it from this shelf, not from your library.
03
The Roses moves to position one with a pin indicator. The rest of the shelf shifts to accommodate.
Long-press hold (500ms) · Prevents accidental triggers. Tap opens the item. Hold reveals controls.
Remove
Hide an item from the shelf without deleting it from your library. Device-scoped, not global. The Undo toast is the safety net that made a trash layer unnecessary.
Removed from Recently PlayedUndo
Frame 2
One surface, three controls.
Long-press reveals Pin, Pause, and Remove in a single iOS-native action sheet. Same entry point for every interaction.
Frame 3
Undo safety net. Device-scoped, not global.
The item is hidden, not deleted. Removal is scoped to this device. The toast makes the action reversible for 3–4 seconds. No trash layer needed.
Loops every 18 seconds
01
Same entry point. Long-press surfaces the same action sheet as Pin.
02
Select Remove from Recently Played. The item is hidden from the shelf, not deleted from your library.
03
Item gone. Shelf reflows immediately. The Undo toast sits above the Now Playing bar: reversible by default.
Auto-dismiss completes the removal · No second confirmation. Inaction is consent. The Undo toast is the entire safety net.
Pause
Stop the shelf from logging activity without leaving the app. Time-boxed to protect ML training signal. One toggle, same action sheet, instant feedback.
01
Same entry point again. All three controls share one gesture and one surface.
02
Pause listening history toggle in its default off state. The toggle sits inline with the other controls.
03
Toggle flipped. History pauses immediately. Time-boxed: resumes automatically to protect recommendation quality.
Time-boxed auto-revert · Protects ML recommendation signal. The user doesn't have to remember to turn it back on.
System Model
Every state is reachable and reversible
One entry point, three branches, and every path returns to Default. The dashed node is the only terminal state, and it takes deliberate inaction to reach it.
States covered, states deferred
The diagram covers reachable controls. Four shelf states sit outside the happy path and need naming.
Discoverability
Long-press has no current affordance on the Recently Played shelf. First-time users get a one-time tooltip on the first scroll after release, plus a small “Tap to manage” inline hint that appears once the shelf accumulates 5+ items. Both decay after the first successful long-press. Deliberately not added: a permanent kebab or chevron next to each tile. It would teach the gesture but compete with album art and erode the calm shelf the design depends on.
Loading
The shelf already renders skeleton tiles while history fetches. The long-press gesture stays suppressed until the fetch resolves, so the existing skeleton covers the controls without new design.
Error
A failed Pin, Remove, or Pause sync surfaces inline retry inside the action sheet, not a global toast. The user keeps their place in the sheet. Full error spec lives in the engineering handoff.
Persistent Pause
When listening history is paused, the shelf header shows a small “Paused” chip beside the section title. The global state is legible without re-opening the action sheet, which closes the visibility loop Pause otherwise leaves open.
Empty
Recently Played hides itself at zero items, so the controls never need to render against an empty shelf. No empty-state design required at this level.
Prototype Walkthroughs
Three controls. Each one walked through end to end.
PIN
Pin a favorite to the top
0:09
What the walkthrough shows
0:00Instruction pill fades in: “Long-press The Roses.”
0:02User taps The Roses (right column, row 2, marked by the green affordance outline).
0:03Action sheet rises with a 200ms ease-out. Pill swaps to “Tap ‘Pin on top.’”
0:05User taps the Pin on top row.
0:06Tile reflows from position 3 to position 1 in 250ms. Toast slides up: “Song pinned.” with green Undo button.
0:07Conclusion pill: “The Roses pinned to the top.” Loop dissolves back to start.
REMOVE
Remove an item without library impact
0:10
What the walkthrough shows
0:00Instruction pill fades in: “Long-press This is Madonna.”
0:02User taps Madonna (left column, row 2, marked by the green affordance outline).
0:03Action sheet rises. Pill swaps to “Tap ‘Remove from Recently Played.’”
0:05User taps the Remove row in Accent / Red.
0:06Madonna’s tile fades out, shelf reflows. Toast slides up: “Removed from Recently Played.” with green Undo button.
0:07User taps Undo on the toast (400ms smart-animate).
0:08Madonna fades back in at her original index. Conclusion pill swaps to “Removal undone.”
0:09Demo dissolves back to pre-intro so the flow replays.
0:02User taps Hollow Coves (right column, row 1, marked by the green affordance outline).
0:03Action sheet rises. Pill swaps to “Toggle ‘Pause listening history.’”
0:05User taps the Pause listening history row.
0:06Toggle crossfades from off to on in 300ms. Sheet holds 0.6s so the state change is visible.
0:07Sheet slides down. Shelf dims to 50% opacity. Toast slides up: “Listening history paused for 12h.” with green Undo button.
0:08Conclusion pill: “History paused for 12 hours.”
0:10Demo dissolves back to pre-intro so the flow replays.
04
Feature Details
Here's the reasoning behind the specific decisions that shaped Pin, Remove, and Pause.
01
Pin
Bring your favorites to the front
Community forums documented users playing one second of a track to push it to the top of Recently Played. That workaround is the evidence Pin isn't speculative. Cap at 4 keeps it lightweight and out of playlist territory.
Key Decisions
Cap at 4 pins: prevents scope creep into playlist management; keeps taps-to-pin ≤2.
Visible Pinned row: makes the feature discoverable; signals clearly that pinning does something.
Undo on pin/unpin: reversibility is a first-class affordance, not an afterthought.
Drag/↑↓ reorder: gives power users control over order without requiring a separate screen.
02
Remove
Hide an accidental or contextual listen without deleting it
Per-item removal, device-scoped. Hides the item from the shelf without deleting listening history. The distinction is explicit in microcopy because Engineering and Support both flagged it as a top confusion risk.
Key Decisions
Device-scoped: satisfies privacy use case without touching global history or affecting recommendations signal.
Does not delete history: protects recommendation integrity; avoids a misleading mental model.
Undo available: keeps the action reversible; reduces CSAT risk.
Microcopy matters here: 'Removed from this device. Doesn’t delete your listening history.'
03
Pause
Stop the clock on a session you'd rather not save
Time-boxed pause on history logging: 15 min, 1 hr, 3 hrs, or Until tomorrow. Auto-resumes at the selected time. This constraint came directly from ML: permanent pausing degrades training signal.
Key Decisions
Time-boxed only: permanent pause would degrade ML training signal. Ruled out in alignment with Recommendations & ML.
Chip + Resume affordance: makes active pause state visible; gives users an easy escape hatch mid-session.
Auto-resume: reduces cognitive load; users shouldn't need to remember to turn it back on.
Start/see/stop in ≤3 clicks: the key metric for this feature; drove the chip placement decision.
05
Constraints & Trade-offs
These were the trade-offs that shaped the design. The reasoning behind each decision matters as much as the decision itself.
Cross-platform
Desktop users access the same controls through right-click, matching how Spotify already handles other context actions. The interaction model stays consistent across surfaces without forcing mobile patterns onto desktop.
Out of scope
Editing global listening history
This touches a different surface (Settings → Privacy) and a different stakeholder group. Scoping it here would expand the project beyond one sprint and dilute focus.
Out of scope
Changing ranking algorithms
Recommendation logic is owned by a separate team. Any ranking changes would require cross-team alignment that's out of scope for a shelf-level UX feature.
Out of scope
Profile privacy redesign
The privacy concerns are real, but they're systemic. They need their own scope. A shelf control would only patch the surface, not the underlying model.
Out of scope
History Trash
The research included a 'History Trash' concept: a secondary space holding 30 days of removed items. It was cut. The Undo toast accomplishes the same goal with less cognitive load and no additional surface to manage. A trash layer introduces a mental model question (is this gone, or just hidden?) that the current design avoids entirely. Undo is the answer. Trash is scope creep.
Constraint
Pause must be time-boxed
Recommendations & ML: permanent pause degrades training signal quality over time. Time-boxing was the negotiated middle ground.
Constraint
Remove is device-scoped only
Privacy/Legal: global history deletion requires explicit user confirmation flows and different data handling. Device-scoped remove is a much lighter lift with the same UX benefit for the target use case.
Constraint
Pin cap at 4
Engineering: a lightweight chip-based sync model breaks down past 4 items. Keeps the feature fast and avoids overbuilding into playlist territory.
06
Validation Plan
This is a concept project. No live data exists. Instead of reporting numbers I don't have, I built the evaluation framework I'd use to gate a launch decision: a moderated usability test with specific acceptance criteria per feature.
Test Protocol
Moderated usability test. 6 participants across the four archetypes (Social Curator, Commuter, Parent on shared device, Explorer). Each participant completes three task scenarios: pin a specific item, remove an accidental or contextual listen, and pause history before handing the phone to a friend. Sessions recorded. Think-aloud protocol. 45 minutes per session.
Feature
Metric
Threshold
Rationale
Pin
Taps to pin
≤ 2
Long-press + tap. If it takes more, the action sheet hierarchy failed.
Pin
Reorder task time
< 3s
Benchmark: iOS Home Screen widget reorder averages 2.5s.
Remove
Time to remove + undo
≤ 6s
End-to-end: long-press, tap Remove, spot toast. 6s is generous. Under 4s signals strong discoverability.
Pause
Start / see / stop
≤ 3 taps
Toggle on, confirm state change, toggle off. Three taps is the ceiling.
All
Task error rate
≤ 5%
Industry benchmark for low-complexity touch tasks. Above 5% means the affordance is unclear.
All
Post-task CSAT (top-2 box)
≥ 80%
Collected after each task, not at end of session. Isolates per-feature satisfaction.
All
UMUX-Lite lift vs. baseline
+8 pts
Baseline: current Spotify shelf with no controls. +8 is a meaningful usability gain, not noise.
Decision Criteria
All feature-specific thresholds must pass for that feature to ship. If error rate or CSAT fails on a single feature, that feature gets a redesign cycle before re-test. UMUX-Lite is a directional signal, not a gate: if task metrics pass but UMUX-Lite is flat, the features ship and the team investigates perception gap in follow-up research.
07
Reflection
The constraints did the design work. Mapping what Product, ML, Privacy/Legal, and Engineering each cared about before any visual design turned global history clearing into a non-starter. Remove outranked Pin, and Pause stayed time-boxed, because those were the correct decisions once the full system was visible. Device-scoping Remove and time-boxing Pause weren't compromises; they were the right calls from the start.
The pivot from Pin-first to Remove-first (documented in Section 03) was the most consequential decision in the project. What I'd add now: ship Remove in sprint one, observe what users actually do with it, then build Pin on validated behavior rather than assumed need. That sequencing change is a judgment call, not a scope call.
