Case study · 7 min read

How I redesigned the supply chain platform

A 25-year Windows load-planning tool, rebuilt as cloud software people could finish a plan in, without the paper checklist beside the monitor. What research showed, which functions left the first screen, and what we caught before the engine.

Three beats of the product, in the order a planner meets them:

1. 01 Pick up where they left off. If the tool forgets the plan, people will not trust the next screen either. Paper existed because state did not.
2. 02 One path, not forty functions. High-frequency work is the primary path. Expert tools stay one disclosure away.
3. 03 Catch the miss before the engine. Impossible values used to reach calculation, then the plan died. Inline flags beat a failure ten minutes later.

My approach

Research

• Shadowed packaging engineers and logistics planners
• Paper checklists and a handful of the 40+ functions told us what the cloud product actually had to be

Design

• Replaced modal islands with one linear workflow and progressive disclosure
• Pre-populate common dimensions. Flag impossible values. Put CO2 on the planning path

Build

• Component library for dense tables and results
• Responsive layouts for desktop, tablet, and remote access. Keyboard and high-contrast for the floor

Test

• Time-on-task, data-entry errors, fleet impact
• Audited against WCAG 2.1 AA and the European Accessibility Act, not as a later overlay

Twenty-five years of features is not a workflow

The tool was the industry standard for load and packaging optimization. FMCG, logistics, and manufacturing teams used it to define cases, simulate loads, and fill trucks. It only ran on Windows. Work was tied to a machine. Nobody could plan together across a global chain in real time.

Feature accretion had produced a dense interface with no hierarchy. Research showed two things that decided the redesign. Planners kept parallel paper checklists because they did not trust the desktop app to hold state. And in a typical week they touched only a handful of the 40+ specialized functions. Mixed-case packing and compression analysis were the steps where people backed out entirely.

What changed

One path instead of modals

Case creation, load building, and truck loading stopped living in separate dialogs. They became one continuous flow, so state had somewhere to sit.

Progressive disclosure

High-frequency functions are the primary path. Expert tools stay available one layer down. Casual users see less. Experts do not lose range.

Stop bad data early

Common dimensions pre-populate from prior plans. Impossible values fail inline, before they reach the calculation engine. Mixed-case packing sat on that path because that is where people used to abandon the flow.

CO2 on the path

Fleet and fuel impact is not a report you remember to open. It sits in the same workflow as the load.

How I checked it

35% less time-on-task for core planning workflows. 50% fewer data-entry errors after launch. 10% reduction in fleet requirements for FMCG clients. One cold-chain firm also logged thousands of planning hours saved. The number I care about more is whether the paper checklist left the desk. When state is trusted, it does.

Password-protected detail sits behind the original project file. This page is the public write-up of the decisions.