What's Next?
Wear tracking has the most left in it. Right now it maps usage against a fixed replacement interval. The version worth building would learn from the conditions the system is actually running in, so a filter in a dusty house reports differently from one in a clean one.
Results
Tradeoff: the calm surface hides things. Putting detail one level down means someone monitoring a specific reading navigates to it every time. That trade is wrong for a user actively diagnosing a problem, which is why the status line names the affected system rather than only flagging that something is off.
Tradeoff: sourcing thresholds externally constrains the design. Published limits aren't always the number a homeowner cares about, and some subsystems have no clean standard to point at. I took the constraint deliberately, because a threshold I invented would be a design opinion wearing the costume of a safety limit.
Tradeoff: a reorderable dashboard weakens the default. Letting users add and reorder cards means I can't rely on position to carry meaning, so every card has to be self-describing. The alternative was a fixed layout that suits nobody exactly.
The tests I would run: time to locate a specific reading against a stock app, whether people can say what a value means without leaving the screen, and how often a set-aside item gets brought back rather than forgotten.
Solution
A mobile-first dashboard that resolves to one line most days and opens into full telemetry when it doesn't. Three tabs, each answering one question. One card format across air, water, gas, and energy, with published safety thresholds drawn into every graph. Consumable wear tracked by usage rather than calendar. High-contrast dark palette, because people check a home dashboard late at night and early in the morning.
Research & Approach
Setting something aside moves it off the status line but keeps it on the list, and the status changes to "Nothing pressing. One set aside." Most apps make dismissal permanent, which teaches people not to dismiss anything. Separating "not now" from "never" means the status line stays trustworthy, because clearing it costs nothing.
Air, water, gas, and energy share identical card anatomy, typography, and graph baselines. Learning to read the air quality card teaches you the others for free, which matters because these are checked occasionally rather than daily. Time to locate a specific value, benchmarked against a stock app, is how I would test whether that holds.
A reading of 21 AQI or 180ppm dissolved solids means nothing alone. Every chart draws its threshold as a dashed baseline, sourced from published standards rather than my judgment: ASHRAE for ventilation, the EPA's PM2.5 standard for particulates, USGS classification for water hardness. When a spike crosses the line, that line is a regulatory limit.
This follows the split attention effect: separating two sources a reader must combine imposes an avoidable working memory cost, and integrating them physically removes it (Chandler and Sweller, 1992). That research is instructional rather than dashboard specific, but the structure is the same. A live reading and its safe range are two things the user has to hold together, so the interface should do that work.
Standard apps say "change the filter every 90 days" regardless of whether the system ran. Hearth maps consumable wear against usage and states it in days remaining, with an inline "Mark replaced" control so maintenance is one tap rather than a trip into settings. Same pattern for appliances, with "Log service" against last-serviced dates.
Reduce motion stops the background drift and counters. Larger text scales the whole interface by about 12 percent. High contrast trades the frosted panels for solid surfaces and stronger borders. Each one is a real alternative rendering rather than a filter over the same design, which meant designing the frosted look and its solid counterpart in parallel.
Client
Self-Initiated
Role
Solo Designer/Developer
Duration
6 months (2026)
Challenge
Smart home enthusiasts want the actual numbers. Consumer apps give them a green dot and an "on" label, which is why so many end up building their own dashboards to reach data their paid app already collects.
But a wall of live telemetry fails a different way. Most of the time nobody is diagnosing anything. They are checking whether the house is fine before bed, or working out why the energy bill moved. Fifteen readings answer neither question quickly.
So the problem was serving both: a glance that says the house is fine, and a path to the real data when it isn't.
Hearth
Data-Driven Smart Home Dashboard