Project 04 · 2022
Decision Tree
IntelliFlux Controls · UX Engineer
Owned and shipped end to end.
Interactive prototype
Open prototypeMy Role
I owned this product from requirements to shipped software — not one slice of it.
Requirements
Translated the vision from the VP of Technology into a real, buildable roadmap.
Research
Ran user research myself, directly with the engineers who lived in the tool.
Design
Owned the UX and product design decisions start to finish.
Delivery
Led project management through implementation with the development team.
The Users
Four roles. Four completely different relationships to the same screen. I sat with all of them — and they landed on the same complaint from four different angles.
- Data Scientists
- Chemical Engineers
- Data Analysts
- PLC Engineers
The tool wasn't broken. It was outgrowing itself.
The Roadmap
Research became a plan. I clustered everything I heard from the four roles into themes, then mapped the product's future as a tree — from the core Decision Tree out to each discipline's needs — so growth had a shape before it had a schedule.
Before the Redesign
Four screenshots, each matching one of the four structural problems engineers kept running into.
The Problems
The Vanishing Menu
Every node lived behind a right-click. No palette, no sidebar, no visible home — just a list that got longer, and longer, and eventually ran out of room to hold itself together.
Bolt-On Syndrome
Save · Load · New became Save · Load · New · Export · Import. Not a menu system — just more buttons, wedged into the same row, every time.
"Coming Soon" Forever
A whole panel of planned fields — flow rate, volume, pressure, temperature, concentration — sat stubbed out. The team knew growth was coming. The layout didn't.
One Dropdown to Rule Them All
Every saved flowsheet, ever made, in one flat list. No folders. No search. No mercy.
The pattern: add it wherever there's room — repeated at every layer of the product, until there was no room left.
The Call
One feature was fully scoped, roadmapped, and ready to build: a comment and notification system. I killed it. It would've looked good in a demo. It wouldn't have fixed why engineers were fighting the tool. Structure would. So structure is what shipped.
The Fix — Simple, but Mega
One idea, applied everywhere: replace every flat list with a real hierarchy.
- The vanishing right-click menu → a persistent, categorized Add Node panel (Data / Logic / Actions / Advanced) that scales with the library instead of drowning in it
- The bolt-on toolbar → a real top bar, with File, History, Settings, Help as actual destinations
- No version control → a history panel, so a bad edit stopped meaning "start over"
- Click-to-place → true drag-and-drop
Not a redesign of every screen. One structural fix, applied consistently, that made every future feature easier to add instead of harder.
After the Redesign
The same tool, after the structural fix: a persistent left rail, organized Add Node panel, dedicated run results area, and a history panel that turned 'start over' into 'step back.'
Latest Iteration
Where the redesign landed — a clean, scannable surface that keeps the canvas central and puts the right information within reach.
The Result
Speed
Engineers built and edited flows faster, with organized menus and real drag-and-drop.
Scale
The node library kept growing — the interface finally grew with it instead of against it.
Rework
History and version tracking meant a bad edit stopped meaning starting over.
The tool became something IntelliFlux could put in front of clients with confidence — directly powering geolocation-based pipeline analysis, leak detection, and automated reporting workflows in production.