BMW’s move to digital wheel alignment makes the headlines this week: trust, traceability and speed. On a factory floor, those three words decide whether signage is a help or a hindrance. A screen that freezes during a shift change, or a safety counter that resets after a power blip, is worse than no screen at all. The harder parts, offline playback, instant overrides, and proof that the message got through, are the ones that keep production running when the network does not.
What factory signage actually has to do
A manufacturing or warehousing floor runs on three types of information: live status, safety metrics, and emergency interruptions. Each has different rules, different audiences, and different failure modes.
| Type | Audience | Update frequency | Tolerance for delay | Tolerance for error |
|---|---|---|---|---|
| Shift board | Operators, supervisors | 1 to 5 minutes | Seconds | Low |
| Safety counter | All staff | Real-time | None | None |
| Line status | Team leads | 10 to 30 seconds | Seconds | Medium |
| Emergency alert | Everyone | Instant | None | None |
Shift boards show targets, actuals, and downtime reasons. They are updated by the MES or a simple spreadsheet. A five-minute lag is acceptable; a frozen screen is not. Safety counters, days since last accident, near-miss reports, must never reset. If the screen reboots, the number must reappear exactly as it was. Line status screens tell team leads which stations are running, which are blocked, and why. A 30-second delay is tolerable; a blank screen is not. Emergency alerts, fire, gas, evacuation, must interrupt whatever is showing, regardless of what the screen was doing before.
The network is not reliable
Wi-Fi coverage in a warehouse is patchy. Metal racks, forklifts, and moving stock create dead zones. A signage system that depends on a live connection will freeze or show a spinner. Devices must cache content locally and play it without waiting for the server. When the network returns, the device should fetch only what has changed, not the entire playlist. This keeps data usage low and playback smooth.
Caching also protects against server outages. If the MES goes down, the screens should still show the last known good state. When the MES comes back, the screens update automatically. No manual intervention, no USB sticks, no lost time.
Interrupting the noise
An emergency alert must override everything else. On an Android device, this is done with the "Display over other apps" permission. The alert appears full-screen, with a loud tone, and stays until acknowledged. It does not matter if the screen was showing a shift board, a safety counter, or a video. The alert takes priority.
Not all Android devices handle this the same way. Some televisions and tablets require the permission to be granted once, by hand. Others need a single adb command during setup. A few refuse to start the app from the background at all; for those, a full-screen notification is the fallback. The system must work with all of them, not just the ones that behave.
Proof that the message got through
A safety counter that resets after a reboot is useless. A shift board that shows yesterday’s numbers is misleading. An emergency alert that no one saw is dangerous. The system must prove that the message was received, displayed, and not overwritten.
Each screen acknowledges what it received and sends a heartbeat every 30 seconds. The server logs these acknowledgements and can ask for a screenshot at any time. If the screenshot does not match what was sent, the system flags it. This gives operations teams confidence that the numbers on the wall are the numbers they expect.
Licensing and hosting: per site, not per screen
A factory with 50 screens does not want to pay for 50 licences. A system licensed per site, not per screen, keeps costs predictable. The customer installs the server on their own hardware, inside their own network. No data leaves the building unless they choose to put the server outside it. This meets IT policies that forbid cloud-based signage.
The server does not need WebSocket support or a persistent worker. Devices poll the server, so it runs on ordinary shared hosting. This makes it easy to deploy and easy to maintain.
What breaks, and what to do about it
| Failure mode | Effect | Mitigation |
|---|---|---|
| Network outage | Screens freeze or show spinners | Cache content locally; play from cache; update when network returns |
| Device reboot | Screen goes blank | App restarts automatically; cache ensures last state is restored |
| Server outage | No new content | Screens continue playing cached content; update when server returns |
| Permission not granted | Alert does not interrupt | Use adb command during setup; fallback to full-screen notification |
| Screen offline for hours | Heartbeat missed | Server flags the screen as offline; operations team investigates |
A reboot should not lose state. A network blip should not cause a blank screen. An emergency alert should not fail because the Wi-Fi dropped. The system must handle these failures without manual intervention.
When digital signage is the wrong tool
Not every problem is solved by a screen. A safety counter that resets after a power cut is worse than a whiteboard. A shift board that lags by ten minutes is ignored. An emergency alert that no one hears is useless. If the system cannot guarantee resilience, traceability, and instant overrides, it is not ready for the factory floor.
Screens are good for live status, safety metrics, and emergency interruptions. They are bad for static information, complex instructions, or anything that needs a signature. A screen that shows a PDF is a screen that is not showing live data. A screen that requires a touch is a screen that is not visible from across the room.
What to do next
- Map the three types of information: shift boards, safety counters, line status, emergency alerts. Decide which screens will show which.
- Test the network. Walk the floor with a Wi-Fi analyser. Note the dead zones. Plan where to place access points or use wired connections.
- Choose devices that support the "Display over other apps" permission. Test the emergency alert on each model. If it does not interrupt, do not use that model.
- Install the server on a machine that is always on. Shared hosting is fine; a Raspberry Pi is fine. Just make sure it is inside the network.
- Cache the content. Test what happens when the network drops. Test what happens when the server reboots. Test what happens when the screen reboots.
- Set up the heartbeat and screenshot requests. Make sure the panel shows what is actually on the wall, not what was sent.
BMW’s digital wheel alignment is about trust, traceability, and speed. Factory floor signage must deliver the same. The screens that do are the ones that keep production running when everything else goes wrong.
