Distro 2.0 - Web Platform

UnitedMasters' distribution flow is the highest-traffic product surface. Every release starts here. Artists submit metadata, artwork, audio, and platform selections through a multi-step form that feeds directly into the DSP delivery pipeline. The quality of what enters this form determines metadata integrity across Spotify, Apple Music, and every downstream platform.

The existing flow had been built incrementally over years. The team operated under an 80/20 prioritization framework for the release pipeline: focus on the 20% of fixes that resolve 80% of submission failures. The redesign was scoped under that lens. Not a full reimagining, but a structural intervention at the points where the most submissions were dying.

Completion sat at 11.87%. That number alone doesn't capture the real cost. Every abandoned submission was an artist who had music ready to release and couldn't get through the form. Every rejected submission created manual ops overhead: someone on the operations team had to review, diagnose, and often communicate the fix back to the artist.

The root cause wasn't missing features. It was sequencing and feedback. Steps weren't ordered by dependency, so artists would fill out distribution details before entering the metadata those details depended on. Validation was generic ("invalid field") rather than diagnostic ("artwork must be 3000x3000px minimum, yours is 1400x1400px"). Artists couldn't tell the difference between a blocking error that would prevent submission and a warning they could safely ignore.

The metadata schema was fixed. DSPs require specific fields in specific formats, and we couldn't simplify what was collected. More importantly, backward compatibility with existing drafts was non-negotiable. Artists had in-progress releases saved in the old format. The new flow had to ingest those drafts without data loss or requiring artists to re-enter information.

Engineering capacity was one sprint for the initial build. This forced a decision: build the flow first and retrofit the component library, or build the component library first and risk shipping late. I chose to design the form components (inputs, selectors, date pickers, validation states, multi-step navigation) as reusable primitives from the start, accepting the upfront cost because this flow would set the pattern for every form surface that followed.

My director ran an initial design pass and user testing sessions that surfaced key signals around validation comprehension and step sequencing. I took those findings and built the final flow architecture, component system, and field-level remediation patterns.

Reordered the entire flow by dependency chain: identity (who is this release by), content (what's on it), distribution (where does it go), review (confirm everything). Each step validates before the artist can advance, but the critical design decision was the distinction between blocking errors and warnings. A blocking error prevents submission: wrong file format, missing required field. A warning flags something that might cause a DSP rejection but doesn't stop the artist, like a title that uses special characters some platforms strip.

I mapped every metadata field to its downstream consumer with data engineering. That mapping became the foundation for contextual help text. Not generic tooltips, but specific explanations of why a field matters. "Release date must be at least 7 days out" became "DSPs need 5 to 7 business days to process your release. Submitting with less lead time risks your release not appearing on release day." The goal was connecting artist action to platform outcome at every field.

Draft auto-save was a trust decision. Artists had been burned by lost work in the old flow. Auto-save with visible recovery ("You have a draft from March 12") addressed that directly. The persistent progress indicator wasn't just navigation. It was a contract showing what's done, what's left, and that nothing entered would be lost.

Submission completion increased from 11.87% to 25.0%, a 2.1x improvement. But the more telling metric was subscription churn: it dropped 55% among artists who used the new flow. The interpretation is that submission friction wasn't just a UX problem. It was actively eroding artists' perception of the platform's value. Removing that friction reinforced why they were paying for a membership.

The component library built for this project became the foundation for every subsequent web form: Content ID enrollment, Waterfall Releases, internal tools. The validation pattern (blocking vs. warning with specific remediation) was adopted as the standard across the product. That was the real systems-level return: one project's design decisions became the default for the org.

The auto-approval pipeline is the next structural intervention. Using fingerprint data, artist history, and membership tier to route high-confidence submissions past ops review entirely. The flow captures clean enough data now to make that feasible, which was part of the original design intent. Expanded metadata fields (producer credits, lyrics, Canvas video) are in planning but gated on DSP readiness.