← Back to all work

Self-Service Workflow: From 53% to 96% Completion

Rebuilding a confusing multi-touch credential request process into one guided change-order flow that scaled completion from 53% to 96%.

53% → 96% completion rate
Role
Product Designer
Timeline
Aug 2024–Present
Tools
Figma, FigJam, React (recreation)
Team
Solo design, cross-functional engineering partners

The Problem

Self-service completion for API, app, and entitlement access requests was stuck at 53%. It wasn’t one clean failure point: manual back-and-forth with asset owners, unclear ownership on requests, and a slow, confusing flow were all compounding on each other. Developers who needed access couldn’t reliably get through the process on their own.

The Process

Applications list with Change Order actions for each registered app (representative recreation) Explore the live, interactive demo

A representative recreation, genericized and rebuilt in React, not the real internal tool.

The work started with a full content and component inventory audit, going through the existing flow screen by screen to identify what could be cut and what needed to be made clearer before redesigning anything.

The Decision

Of all the tangled problems, the flow itself was the one piece fully within design’s control, and the lever most likely to unblock everything else. So that’s where the redesign focused: rebuilding the request process around a single change-order pattern. Select the asset, a side panel opens with a dropdown of the available metadata changes, add any additional fields the change needs, then submit and get confirmation. What had been a confusing, multi-touch process became one guided path.

The change-order side panel: select a change type, fill in just the fields it needs, submit (representative recreation)

The Outcome

An app owner's own Pending Requests view: every outstanding request tied to their application, with nothing left to chase down manually (representative recreation)

Reflection

This case study is a good example of why “redesign the screen” and “redesign the flow” are different jobs. The original screens were reasonable individually; the problem was how many of them you had to cross, in what order, with how much ambiguity about who owned what. Fixing that meant collapsing steps, not polishing them.