I Built a Family Dashboard (The Honest Engineering Writeup)

Six months ago I published a piece about my e-paper family dashboard. Since then, three things happened. The build was adopted by 4 other families in my social circle. The build failed twice in production. The hardware is now cheaper. This is the honest followup with the engineering details, the failure modes, and the updated cost. If you are considering building your own, this is the version you should build.

The Original Build (Recap)

The original build: 7.5-inch e-paper display from Waveshare, Raspberry Pi Zero W, 3D-printed frame, $45 total. The display shows the family calendar, weather, school pickup schedule, and dinner plan. The Pi fetches the data on a 15-minute schedule. The data sources are all free. The total maintenance: zero. The build was a weekend project. The result: the family uses it daily. The right answer for the original build is to do it. The right answer for the production-hardened build (described below) is to do it differently in a few specific places.

What Failed in Production

Two failures in 6 months across 5 builds (mine plus 4 family/friends). First, the Raspberry Pi Zero W is not reliable for the e-paper refresh. The e-paper refresh takes 2-3 seconds and uses significant current. The Pi Zero W has limited current on the USB port. The combination occasionally causes the e-paper to draw too much current during refresh, which causes the Pi to brown out (experience a voltage drop that causes a crash or restart). The fix: use a Raspberry Pi 3 or Pi 4 instead. The Pi 4 has more current and a more stable USB. The Pi 3 is also fine. The trade-off: the Pi 4 costs $35-50 instead of $15. The right answer for production is the Pi 4. The right answer for a quick test is the Pi Zero. The failure mode was: the e-paper would show partial data and freeze. The diagnostic was: the Pi was rebooting during the refresh. The fix was: use a Pi with more current.

Second, the Google Calendar API rate limit is hit by accident when multiple family members add events. The fix: use a service account (an API credential that authenticates as the app, not a user) with read-only access to the family calendar. The service account does not have user-based rate limits. The right answer is a service account. The wrong answer is OAuth user tokens (which have lower rate limits). The failure mode was: the calendar would stop updating for a few hours. The diagnostic was: 429 errors in the logs. The fix was: switch to a service account.

The Updated Build

The current production build: 7.5-inch color e-paper display from Waveshare ($80), Raspberry Pi 4 ($45), 3D-printed frame ($5), misc cables and SD card ($10). Total: $140. The display is color (the new model is $80 instead of $20 for grayscale). The Pi 4 is reliable. The 3D-printed frame is a more permanent wall mount. The total cost is higher than the original ($140 vs $45), but the reliability is dramatically better. The right answer for a quick test is the original $45 build. The right answer for a production build is the $140 build.

The Software Stack, Updated

The current software: a custom Python script (about 300 lines) that fetches the data, renders the display, and updates it on a 15-minute schedule. The script runs as a systemd service. The data sources: Google Calendar (via service account), Open-Meteo (no auth), a hard-coded JSON for the pickup schedule and dinner plan. The script is open source on GitHub. The right answer for the software is to fork the GitHub repo. The right answer for the data sources is the ones I described. The right answer for the schedule is 15 minutes. The right answer for the rendering is Python with the Pillow library. The right test is whether the build runs for 6 months without intervention. The answer for the production build is yes.

What I Would Do Differently

Three things. First, I would use the color e-paper from the start. The $60 difference is worth the visual upgrade. The grayscale is fine. The color is better. Second, I would use the Pi 4 from the start. The $30 difference is worth the reliability. The Pi Zero is fine for a test. The Pi 4 is required for production. Third, I would add a power-loss recovery script. The original build assumed the Pi was always on. The reality: power blips happen. The Pi needs to recover from a power blip. The fix: a systemd service that starts on boot. The right answer for a production build is a recovery script. The right answer for a quick test is to skip it.

The Adoption Story

The 4 other families who adopted the build: 1 friend who runs it on a Pi 4 in the kitchen, 1 cousin who runs it on a Pi 3 in the hallway, 1 sibling who runs it on a Pi 4 in the mudroom, 1 colleague who runs it on a Pi Zero in their home office (despite the reliability issue, they accept the trade-off). The total installs: 5. The total success rate: 5/5. The total failures: 0 in the last 3 months (after the fixes). The right answer for adoption is to share the build. The right answer for the maintainer is to maintain. The right answer for the user is to use.

What This Is Not

This is not a smart home hub. This is not a home automation controller. This is a display that shows information. The right answer for a smart home hub is Home Assistant. The right answer for a family dashboard is an e-paper display. The trade-off is the simplicity. The dashboard is simpler. The dashboard is more useful for the specific use case. The right test is whether the family uses it. The answer for the 5 installs is yes. The right test for the long term is whether the maintenance stays zero. The answer for the first 6 months is yes. The right test for the next 6 months is TBD. The right answer for now is “yes, build it.” The right answer for the future is “if it breaks, fix it.”

Final Thoughts

The e-paper family dashboard is one of the few tinkerer projects that has survived a year of actual family use across 5 installs. The original build is $45. The production build is $140. The maintenance is zero. The use case is real. The right answer for a tinkerer who wants to build something their family will actually use is the production version of this dashboard. The build is a weekend. The cost is $140. The result is a glanceable family display that the family uses daily. The trade-offs are real (color e-paper is slower, the Pi 4 uses more power, the 3D-printed frame is custom). The result is the right answer for the use case. If you have been looking for a tinkerer project that your family will actually use, this is the one. The build details are in the GitHub repo. The wiring diagram is in the README. The software is the same as the original. The hardware is the production version. The result is a glanceable family display that the family uses daily. Build it.