The Fastest Route to What’s Broken
Client
Devicethread
Year
2026
devicethread is an IoT-based hospitality SaaS platform connecting smart locks, thermostats, and sensors into a single cloud ecosystem. The backend processed millions of real-time data points flawlessly, but the operational dashboard staff relied on daily couldn't turn that data into anything actionable.
Scope of Work
Millions of data points, zero visibility, staff couldn't see what needed fixing.
Rather than iterating on the existing table-based layout, we tested how far a fully spatial model could go — mapping every alert to a physical location and giving staff a live surface to fix issues from, instead of just a place to read about them.
TL;DR
devicethread's backend captured real-time signals from thousands of IoT devices, but the frontend buried that intelligence inside paginated tables staff couldn't parse under pressure. I helped turn an early spatial concept into a production system that surfaces critical issues instantly and lets staff resolve them without leaving the screen.
Context & Problem
The dashboard was built like a Property Management System, with an occupancy grid, when devicethread actually needed an error-management tool. Staff couldn't see failures, prioritize them, or fix them without leaving the screen.
Critical hardware issues were buried in paginated tables; no spatial view made property health visible at a glance.
Resolving a single alert meant jumping across four different screens, wasting time during live incidents.
Staff couldn't distinguish a critical failure from a minor warning without manually checking each device.
Business Goal
The engineering was world-class, but the interface was too rigid to use. The mandate: shift from a developer-centric database to a user-centric operational tool before usability issues turned into churn.
Approach & Constraints
The core insight: lists are for reading, hotels are for living. A spreadsheet can't show that a leak in Room 304 sits next to an electrical failure in Room 305, so we moved from a list view to a spatial-first architecture, mapping alerts directly onto floor plans.
Real hardware, two environments, one existing direction to build on.
IoT devices respond more slowly than servers, staff worked from two completely different physical contexts, and the concept already had an early direction set; this wasn't a blank canvas.
Hardware responds more slowly than servers; the interface had to feel instant regardless.
Staff worked from a desk and from hallways; one interface had to serve both contexts.
Early spatial concepts already existed; execution had to build on, not replace, that vision.
Research & Exploration
Rather than iterating on the existing table-based layout, we tested how far a fully spatial model could go — mapping every alert to a physical location and giving staff a live surface to fix issues from, instead of just a place to read about them.
Lists are for reading; hotels are for living.
We moved from an industry-standard list view to a spatial-first architecture, mapping digital alerts to physical floor plans to give staff instant situational awareness and cut the cognitive load of diagnosing hardware problems.
Exploring solutions
Two features turned a passive database into a live action system.
Floor plan rendered as the interface; room markers shift color by device status.
Auto-aggregates every active failure into a single, urgency-sorted task queue.
The hover-and-click model brings hardware controls directly into the map view.
Collaboration
Joined mid-vision, owned execution, turning an early concept into a production system.
3 Core Collaborators
Lead Designer (direction), Product Managers, Engineering
Field-Validated
Iterated directly against real operational scenarios, not just internal review
Evolved the Lead Designer's early low-fidelity concepts into production-ready UI.
Worked with product managers to translate business rules into interaction logic.
Aligned every design decision with real hardware behavior, latency, and network limits.
Outcomes
The redesign turned a data-rich but unusable dashboard into a tool staff could trust under pressure, cutting detection time, removing training overhead, and letting issues get resolved without ever leaving the map.
99%
Faster hardware error detection
100%
Remote issue resolution enabled
4→1
Screens needed to resolve an issue
1st Shift
Staff operational with zero training
Learnings
The screen is only half the story, context shapes everything else.
Displaying every data point creates noise; effective dashboards suppress the normal to surface the critical.
IoT hardware is slower than servers; optimistic UI made sluggish responses feel instant.
A feature useful at a desk is often useless in a hallway; the environment shapes what the interface must do.
