Redesigned a factory-floor tool that works in loud, high-pressure environments.
Tychas is a factory-floor tool used by blue-collar workers to track production in real time.
The tool was deployed at a small-scale manufacturing unit producing ~50k industrial torches per day.
Each torch moves through five sequential checkpoints before it is dispatched to retailers or warehouses.
Lifecycle of an industrial torch on the assembly line
Assembly lines are loud, bright, chaotic, and cognitively demanding.
Shadowed operators on the shop floor for a full day
Observed interactions across 5 production checkpoints
Conducted short, in-context interviews with blue-collar workers
There were 3 aspects where the tool fell apart.
No scanning feedback
Operators already knew whether a product passed or failed.
Digital workflow didn't match physical reality
Operators already knew whether a product passed or failed.
But digitally they had to:
• Scan the barcode
• Search for the product in a table
• Click into a row
• Manually update the status
No error proofing
The system didn't account for real-world failures:
• Unreadable barcodes
• Duplicate scans
• Missed scans
Mapping operators' physical and digital actions helped identify friction points.
Eventually I defined a digital workflow that mirrors how operators actually work on the line.
Reduction of these 2 steps decreased the overall time per product by 4 seconds
Tested an intermediate UI on the live factory floor. Found that while design met AA standards, operators still missed scan confirmations!
Before - no feedback upon scanning
After - feedback upon scanning
Accessibility on factory floors has to be impossible to miss, not just standards-compliant.
Scan feedback states of the intermediate UI
Each decision below reflects learnings from on-floor testing.
Real-time scan feedback.
Physical actions reflected digitally.
Reduced the interaction to a single, repeatable loop: scan → decide → move on.
Defined error states for different failure types.
Also designed a KPI dashboard for line leaders to monitor throughput across checkpoints.
Operators cut click-throughs by 1.5×. Everything felt faster, lighter.
...which is a precursor for production volumes to go up, and error rates to go down.
Blend in.
Just adding tech isn't enough.
You have to ground it in the reality of the people using it. By listening, watching, and building alongside shop-floor operators, design can smooth out the rough edges and really help work flow.
