Delete the screens that did not end in a decision. That is the whole test, and you can apply it to every manufacturing dashboard you own in an afternoon. For each one, ask a person to name a single decision it changed in the last month. The screens with an answer stay; the ones that come back with a story about how useful they could be are not being used, and they take attention away from the two or three screens that actually run the place. I was in a plant boardroom last spring. I do not run lines. I write words for a living. There were nine tiles on the wall screen and one of them was red. I asked what happens when that tile goes red. Four people looked at each other, and then the maintenance lead said nothing happens, it has been red since the sensor got moved. Everybody laughed, because everybody already knew. The screen had been reporting an accurate fact about a machine that was no longer in that spot, every ten seconds, for months. That is the easy version. Anyone can spot a broken tile. The expensive version looks healthy: numbers correct, colours right, refreshed on schedule, and nobody has made a different call because of it since the day it was built. Which manufacturing dashboard is worth keeping? A manufacturing dashboard is a screen that turns plant data into a decision somebody makes on a schedule. A screen is worth keeping when someone can say a sentence out loud about it, in this shape: when this happens, this person does this thing, by this time. When fill weight drifts past target for ten minutes, the line lead pulls a sample and adjusts before the run is scrapped. When the changeover on line 3 crosses forty minutes, the shift supervisor calls maintenance instead of waiting for the end of shift. When a lot is flagged on hold, the quality tech opens the record and clears or escalates it before the truck is loaded. Say the sentence for each screen. If it needs a paragraph, or a maybe, or the name of somebody who left in 2024, you are looking at a report rather than a decision. This is a language test, which is the part of the job I actually know. A screen’s name is a claim about why it deserves your attention. “Line 3 Overview” claims nothing. It is a filing label. “Fill weight drifting” names a moment, and a moment implies a person, and a person implies an action. Most plants I have walked through have a wall of screens whose names say nothing about when to act, and call it visibility. The audit you can run in an afternoon Get the list first. Whoever administers your reporting tool can export every manufacturing KPI dashboard, report, and scheduled email in about ten minutes, including the ones you forgot existed. Include the emailed PDFs. They are dashboards that arrive whether or not anyone asked for them. Then build one row per screen and fill in four columns. Who opens it. A name, not a department. “Operations” means nobody. When they last opened it. Most tools log this. The log is usually the whole audit. The decision sentence. When this happens, this person does this thing, by this time. Written out, in full. What breaks if it is turned off for a week. Guess, then find out. Column four is the hardest one to answer honestly, and the only way to fill it in properly is to turn the screen off. Two weeks dark. If nobody calls, you have your answer. If somebody calls in an hour, you have learned something better than an audit: you found a decision that depends on that screen and nobody had written down. Two people in a room should get through thirty or forty screens in an afternoon. The list matters more than the tool. What the audit usually turns up The same number under three different names. Availability on one screen, uptime on another, run rate on a third, all built by different people in different years from different definitions. Nobody trusts any of them, so a supervisor keeps the real number in a notebook. That notebook is the actual reporting system. The screen built for a person who left. It made sense when they were here. It answers their question, in their vocabulary, and it has outlived them by two reorganisations. The monthly review nobody in the room can act on. Reviewed monthly, by someone who cannot change what it measures, in a meeting where the only available action is a comment. That is not a decision. Nobody in that room can act on what the screen shows. The alert everyone mutes. It fires so often that people stopped reading it. The mute is a decision, and it was made once, quietly, and it applies to every future firing including the one that matters. The finding worth the afternoon is not on that list. Somewhere in that room, someone will describe a call they make every single day, with real money attached, and there is no screen for it at all. It lives on a whiteboard, or in a text message to a supervisor, or in one person’s head. You went looking for screens to delete and found the decision nobody built a screen for. What to do with the ones you keep Fewer screens, each one closer to the person who acts. That usually means the surviving screens stop being screens. A decision that has to be made in the next ten minutes should not wait for somebody to walk past a monitor. It should go to the person who makes it, where they already are, with enough of the record attached, a number that gets back to the shift that caused it, so they can act without opening three other systems to find out whether this has happened before. Archive what you delete, and write down why. Six months from now somebody new will propose building it again, and the note is cheaper than the rebuild. The cost of a dead screen is not the licence. It is that an operator walks past nine tiles a day, learns that screens are for other people, and stops looking. Then the one screen that would have saved the run is up there, correct, refreshed on schedule, glowing at an empty room. This week: List every screen your plant runs, and for each one write the sentence: when this happens, this person does this thing, by this time. The screens without a sentence are your afternoon's work.