Dashboard Wallpaper: How a Useful Screen Becomes Decor, and How to Rescue It
There is a screen in your office that is on, and nobody is reading it. You know the one. It was built by someone who cared. It shows real data from a real monitoring system. At some point in the last year it stopped being a tool and became part of the room, like the carpet. People walk past it. Visitors are impressed by it. Nobody has looked at it on purpose since spring.
This is not a failure of the person who built it, and it is rarely a failure of the monitoring behind it. It is what happens to every dashboard that is not deliberately kept alive. It happens by a mechanism so gradual that nobody can name the day the screen died. I have let it happen to two screens of my own and only noticed when a new starter asked what the orange one meant.
By the end of this guide you will be able to recognise wallpaper and explain how a screen gets there. You will be able to run a test that proves whether anyone is reading it. You will be able to prune it on a schedule with a panel inventory you can copy. You will be able to decide when a rotating carousel helps and when it hides the one panel that matters. And you will be able to bring a dead screen back to life in an afternoon.
In this guide
- Wallpaper, defined
- How a screen becomes wallpaper
- The signs
- The "would anyone notice" test
- The quarterly prune and the panel inventory
- Rotating screens: help or hiding place
- The rescue plan
- The prune checklist
- Where to go next
Wallpaper, defined
Wallpaper is a dashboard that is switched on but not read. It still refreshes. Its panels still draw real numbers. The difference between it and a working dashboard is entirely on the human side: nobody's behaviour changes because of anything the screen shows. Say a panel turned red at ten in the morning. The first person to act on it did so because of an email at eleven. In that case, the screen is wallpaper, however accurate it is.
The definition matters because the usual test — "is the data correct?" — does not catch it. Wallpaper is usually correct. The guide on what dashboards are for defines a dashboard as a screen that answers one question, for one audience, at a glance. Wallpaper is a screen that still answers, to an audience that stopped asking.
This guide is about one screen and how to rescue it. If the problem is that you have nine screens and nobody knows which one is true, that is a portfolio problem. In that case, dashboard sprawl covers it. The two are related, because a screen becomes wallpaper faster when it has siblings, but the cure for one screen is the same either way.
How a screen becomes wallpaper
Four things turn a dashboard into wallpaper, and they usually arrive in this order.
Panels are added and never removed
Every panel on a screen was added because someone wanted it. No panel is ever removed because someone stopped wanting it, because nobody says "I have stopped wanting that graph". So the count only goes up. A twelve-tile wall becomes a twenty-panel wall by the end of the year. At twenty the tiles are too small to read from the door. So the screen fails the reading-distance test that the NOC wall guide describes. People stop reading it from the door. That is the only place they ever read it from.
Something goes amber and stays amber
One panel crosses a warning threshold for a reason that is real but not urgent. It might be a disk at 86 per cent that is being dealt with next quarter. The panel stays amber. Then a second one does. Within a month the screen has a permanent amber patch. The eye learns that amber is a colour the screen just is. The first time a new amber appears among the old ones, nobody sees it, because the screen has trained the room not to.
Nobody owns it
The person who built the screen changes role, leaves, or simply moves on to the next project. The screen becomes everyone's, which is the same as nobody's. A panel breaks and nobody knows who to tell. A service is decommissioned and its tile goes grey forever. The screen is now a museum of a network that used to exist.
The room stopped changing
A wall screen is only read because the room notices a change on it. A screen that is green for six months has no changes to notice. The eye stops visiting. This is not a flaw in the room; it is how attention works. It means a wall that is healthy for long enough will drift to wallpaper without any panel being wrong. The exception is when something on it is deliberately made to change. Dashboards do not decay because they are bad. They decay because they are left alone.
The signs
You can diagnose wallpaper from the corridor, without touching the monitoring system, because the symptoms are in the people.
- Nobody reacts to red. A tile goes red and the room does not turn its head. Someone notices ten minutes later from an email. When you ask, they say "oh, that one goes red sometimes".
- "When did that go amber?" Someone asks it, and nobody knows, and the honest answer turns out to be weeks. A screen that is being read cannot produce that question.
- Dust on the remote. The television's remote, the keyboard of the kiosk machine, or the mouse hanging from the bracket has not been touched since the screen was set up. Nobody has changed the page, fixed a panel, or acknowledged an alert from it.
- The screen is described, not read. People say "the big screen" or "the graphs", never "the file server tile". They know it is there. They do not know what is on it.
- A dead panel has been dead for a while. A grey tile, a "no data" graph, or a panel showing a value that stopped changing months ago, and nobody has mentioned it. Dead panels are the clearest evidence, because they are visible to anyone who looks.
One dead panel is a bug. Three dead panels that nobody has reported are a diagnosis. (Five is an art installation.)
The "would anyone notice" test
The test for wallpaper is to change something on the screen and wait to see who says so. It is the only honest measure of whether a dashboard is read, because it measures reading rather than asking about it. People always say they read the dashboard.
Run it like this:
- Pick a change that is visible but harmless. Swap the positions of two tiles or rename one panel to something slightly wrong. Or set one green tile to a test state that reads "TEST" in amber.
- Note the date and time you made the change, and tell one other person so it cannot be mistaken for a real fault.
- Wait. Do not mention it. Do not look at the screen meaningfully in front of colleagues.
- Record who mentions it first and how long it took. If nobody mentions it within a working week, write "nobody" in the log.
- Put the screen back.
A screen that is read produces a comment within the hour, usually from someone mildly annoyed. A screen that is wallpaper produces silence, and the silence is the result, not a failure of the test. Do the test once a quarter and keep the log. A screen whose time-to-notice is growing is a screen on its way out. You can catch it before the silence.
A variant works for the manager's page. Change a status word to something unambiguous and wrong. For example, mark a healthy service "Watch" with a note reading "test, please tell IT you saw this". See whether the reader tells you. Keep it short-lived. A manager's page that says "Watch" for a week has taught its reader something you did not want taught.
The quarterly prune and the panel inventory
The quarterly prune is a fixed appointment where every panel on a screen has to justify itself or go. It is the only mechanism I have found that reverses the "added, never removed" drift. It creates the moment when someone can say "I have stopped wanting that graph" without having to bring it up. Put it on the calendar with the dashboard's owner and one person from its audience, and give it half an hour.
The tool for the prune is the panel inventory. It has one row per panel, with the question the panel answers, the last time it changed a decision, and a verdict. Fill it in before the meeting, in the meeting, from memory and from the monitoring system's own history. It is deliberately blunt. A panel that has not changed a decision in a quarter is not earning its space, however handsome it is.
| Panel | Question it answers | Last time it changed a decision | Keep / cut / move |
|---|---|---|---|
| Internet link tile | Is the internet up right now? | Outage six weeks ago; room saw it first | Keep |
| Guest Wi-Fi client count | How many visitors are online? | Never | Cut |
| Core switch CPU graph | Is the core switch overloaded? | Never on the wall; useful on the desk view | Move to admin view |
| Backup status tile | Did last night's backup succeed? | Late job caught last month | Keep |
| Printer toner levels | Which printers need toner? | Never; facilities has its own list | Cut |
| Old VPN concentrator tile | Is the old VPN up? | Device retired; tile grey for four months | Cut |
| Disk space, file server | Is the shared drive about to fill? | Amber for three months; purchase approved | Move to manager page with a date |
| Open incidents count | How many things are being worked on? | Daily | Keep |
"Move" is the verdict that saves the most arguments. A panel that is wrong for the wall is often right for the desk view or the manager's page. Saying "it goes there" is easier for everyone than "it goes". Dashboards that earn their place covers deciding which few screens exist for a panel to move to.
Dead panels are easy to find before the meeting if your monitoring database can be queried. A panel is only as alive as its newest sample. Table and column names differ between products. So treat this as the shape of the query rather than the query. Check the product documentation for yours:
SELECT sensor_name, MAX(sample_time) AS last_sample
FROM samples
GROUP BY sensor_name
HAVING MAX(sample_time) < DATETIME('now', '-7 days')
ORDER BY last_sample;
Every sensor that comes back has not reported in a week. Every panel that reads from it has been showing stale data or nothing for at least that long. Bring the list to the prune. It is the one piece of evidence nobody can argue with, and it usually explains the grey tile that everyone had stopped seeing.
Rotating screens: help or hiding place
A rotating screen (or carousel) cycles through several dashboard pages on one television, showing each for a set number of seconds. It is the most common answer to "we have more panels than fit", and it is sometimes a good one. It is also the most effective way ever devised to hide a red tile from a room.
Rotation helps when every page is the same kind of thing and none of them is urgent. Take six branch offices, one page each, all tiles, rotating every twenty seconds. A passer-by sees each office in turn. A red tile on any of them is visible within two minutes. That works because the pages are peers and because the cycle is short enough that the eye catches a change.
Rotation hides the one that matters when the pages are not peers. Suppose page one is the status wall and pages two through five are graphs somebody wanted. In that case, for eighty per cent of every cycle the status wall is not on the screen. The red tile on it is not either. The room learns that the screen is usually showing graphs, and it stops looking. This is the mechanism by which a carousel turns a working wall into wallpaper in a single change.
The rule that survives both cases: the page that answers "is anything broken?" never rotates away. If the wall must carry more than fits, split the screen and rotate only the lower half. Or, in that case, rotate only within pages that are all status walls. Or put the extra pages on a second television. And if any rotating page has a panel with no fresh data, the rotation is now periodically displaying a lie. That is worse than displaying nothing.
At Bellweather Logistics the operations wall rotated through five pages. Page four was the old warehouse management system, which had been migrated away from the previous winter. Its panel had shown "no data" every ninety seconds, for eight seconds at a time, for a little over a year. Nobody reported it, because nobody was reading page four, or, it turned out, pages two, three, or five. The review recorded that the wall had been "partially informative". It now shows one page, which does not rotate, and has eleven tiles on it. The twelfth space is empty, and the team has been asked to leave it that way.
The rescue plan
A wallpaper screen can be rescued in an afternoon if you do it in this order and resist doing it in any other. The temptation is to start by adding a new panel that will make people look again. Do not. The room has learned to ignore the screen, and a new panel is just more of the thing they are ignoring.
- Name an owner. One person, by name, responsible for the screen until the next prune. Write it on the panel inventory. Nothing else on this list happens without it.
- Run the panel inventory. Every panel, the question it answers, the last decision it changed, keep or cut or move. Be harsh. A screen that comes out of this with half its panels is a screen that has recovered its reading-distance.
- Clear every permanent amber. Fix the condition or raise the threshold so the panel is honestly green. Or move the panel to the manager's page with a date on it. Amber on the wall must be temporary or it must be gone.
- Remove every dead panel. Anything with no fresh data goes, today, and the query above tells you which ones.
- Add a freshness marker and a grey rule. Use a last-data clock and a rule that turns a tile grey after three missed polls. That way, the screen can never again quietly show last month's news.
- Stop the rotation, or restrict it. The status page stays on screen. Anything else that must rotate does so in its own space.
- Rearrange for hierarchy. The panel that answers the question goes top-left and big; visual hierarchy covers how.
- Tell the room. Announce that the screen has changed, what is on it now, and what red means. A rescued screen that nobody was told about is still wallpaper for a week.
- Run the "would anyone notice" test a month later. Log the result. If the answer is still "nobody", the problem is not the screen. In that case, the room does not need one. That is a legitimate finding.
Rescuing a screen once is an afternoon. Keeping it rescued is the quarterly prune, and the prune is the part people skip. The dashboard maintenance burden is honest about what that upkeep costs. The argument here is only that the cost of the prune is smaller than the cost of the screen nobody reads. The screen nobody reads still costs a television.
Tip: a screen that is healthy for months can be kept alive by one deliberate change a month that the room is expected to see. Examples are the on-call name in the status strip, the date of the next maintenance window, or a count of days since the last incident. Something true, small, and moving. The eye keeps visiting a screen where something moves. It will be there when the thing that moves is a tile turning red.
The prune checklist
Use this at every quarterly prune. Each line is a yes or no, and the meeting is not finished until every line is a yes.
- The screen has a named owner, written on the panel inventory.
- The panel inventory has a row for every panel currently on the screen.
- Every row has a question, a last-decision date, and a verdict.
- Every "cut" panel has been removed.
- Every "move" panel has been placed on its destination screen.
- No panel on the screen reads from a sensor with no data in the last seven days.
- No panel has been amber for more than thirty days.
- The freshness marker is present and shows the time of the newest data.
- Tiles turn grey after three missed polls.
- The page that answers "is anything broken?" does not rotate away.
- Every remaining panel passes the reading-distance test from the door.
- The "would anyone notice" test has been run since the last prune, and its result is logged.
- The next prune date is on the calendar with the owner and one audience member invited.
Where to go next
Wallpaper is not a screen that went wrong. It is a screen that was left alone. Every screen left alone goes the same way: more panels, permanent amber, no owner, a room that stopped looking. The cure is not a better screen; it is a calendar entry, a panel inventory, and one person willing to say "cut". Do that four times a year and the television on the wall goes back to being a tool. That is the only thing it was ever meant to be.
The design review checklist turns the prune into a full review with a worked redesign of a forty-panel screen. The NOC wall and the boss's browser describe the two screens most panels get moved to. All of them are part of Dashboard Design: The NOC Wall vs. the Boss's Browser.
Put this into practice with SysPeep Monitor
A screen stays alive when the data behind it stays honest and the panels follow the network as it changes. SysPeep Monitor auto-discovers devices and interfaces. It polls them over SNMP, WMI, and SSH from a central Windows service. It shows the results on live dashboards in one web console. So a retired device stops being a grey tile nobody explains. The free trial installs on a Windows machine if you want to rebuild a rescued wall against your own network.
Download Free TrialLearn About SysPeep