I have a test for every information system on a shop floor. I call it the five-second rule. If I walk past a display — a board, a screen, a chart, a dashboard — can I tell, within five seconds, whether things are normal or abnormal? If the answer is yes, the system works. If the answer is no, the system is noise.
Most digital dashboards fail this test. Most visual management boards pass it. And after twenty years of implementing both, I can tell you that the cheap, simple, physical board will outperform the expensive, complex, digital dashboard almost every time. Here is why.
What visual management actually does
Visual management is not about making information look nice. It is about making abnormality visible. The entire purpose is to create a visual environment where the normal state is obvious and the abnormal state is immediately, unmistakably different. When this works, anyone — a visitor, a manager, an operator from a different line — can walk into the area and, within seconds, see whether the process is under control or not.
The best example I have ever seen was at a Toyota supplier in Poland. The line had a physical board — a simple steel panel painted green — with magnetic squares representing each workstation. Green square: station running normally. Yellow square: minor issue, line still running but needs attention. Red square: station stopped, line down. The squares were moved by the team leader, manually, as conditions changed.
I walked past this board at 14:32 on a Tuesday. Station seven was yellow. I knew, within one second, that something at station seven needed attention. I did not need to read a chart, interpret a trend, open a dashboard, or wait for an alert. A yellow square on a green board. That is visual management.
The power of visual management is not in what it shows. It is in what it does not show. It strips away everything except the answer to one question: is this normal?
Why dashboards fail the five-second test
Digital dashboards are designed by software engineers, not shop-floor operators. They are optimised for data density, not for cognitive efficiency. A typical dashboard I audited last year showed, on a single screen: OEE by line, defect rate trend, SPC status for twelve characteristics, maintenance backlog, attendance, safety incident count, and a list of active nonconformances. Forty-seven data points, updated every thirty seconds, on a 55-inch screen.
No human being can process forty-seven data points in five seconds. The dashboard is not providing information — it is providing the illusion of information. The operators on the floor do not look at it because it is meaningless to them. The shift leader glances at it every hour or so, looking for the one number they care about. The plant manager sees it during walkthroughs and feels reassured that "we have real-time data." Nobody actually uses it to make decisions.
Compare this to the board at the Toyota supplier. One piece of information per station. One glance. Five seconds. Action.
The cognitive cost of digital
There is a second reason visual management beats digital: the cognitive cost of interpretation. A physical board with colour-coded markers requires zero interpretation. Green is normal. Red is not. The information is pre-processed — the system has already determined what is normal and what is not, and the visual representation maps directly to the state.
A digital dashboard requires interpretation. The operator sees a number — 87.3% OEE — and must mentally compare it to the target, consider whether it is trending up or down, decide whether the deviation is significant, and determine what action to take. Each step requires cognitive effort. And cognitive effort, under time pressure and production stress, is in short supply.
This is not a criticism of operators. It is a recognition of human cognition. Under load, humans default to the simplest available information. A red square is simpler than a number. A physical indicator that has been moved from its normal position is simpler than a trend line. Visual management wins because it works with human cognition instead of against it.
The board that saved 200,000 euros
In 2023, I was asked to approve a 200,000-euro investment in a real-time production monitoring system at one of my plants. The system would collect data from twelve machines, process it through a cloud platform, and display real-time OEE, defect rates, and downtime reasons on screens throughout the plant. The business case was compelling — the vendor had calculated a 14-month payback based on projected OEE improvement.
I asked the plant manager a simple question: "When a machine goes down today, how do you know?" The answer: "The operator calls the maintenance team on the radio." I asked: "How long does that take?" The answer: "About thirty seconds."
I then asked: "What will be different with the new system?" The plant manager thought for a moment and said: "The dashboard will show the downtime within fifteen seconds of the machine stopping." I asked: "And then what?" He said: "The maintenance team will be dispatched."
The current system — operator calls on radio, maintenance responds — had a detection-to-response time of about thirty seconds. The proposed system would reduce detection time to fifteen seconds but would not change the response — the maintenance team would still be dispatched by a human, using a radio. The fifteen-second improvement in detection was worth 200,000 euros to a plant where the average downtime event lasted forty-five minutes.
I rejected the investment and spent 200 euros on physical boards. Each machine got a two-colour flip indicator: green for running, red for stopped. The team leader's job was to flip the indicator when the machine stopped and note the time. At the end of the shift, the data was entered into a simple spreadsheet for trend analysis.
The result: downtime detection was still about thirty seconds (operator sees the red indicator and calls maintenance). But the visual indicators created a secondary benefit the dashboard would not have: any manager walking through the plant could see, at a glance, which machines were running and which were stopped. The plant manager, who had never walked the floor regularly, started walking it twice a day because the visual indicators gave him immediate, actionable information. He spotted patterns — machine six was always red during the afternoon shift — that led to maintenance investigations that the dashboard data, had we installed it, would probably have buried in a chart.
The four rules of effective visual management
After twenty years of implementing visual management systems, I have four rules:
1. One board, one question. Each visual element should answer one question. Is the machine running? Is quality on target? Is the schedule on time? Mixing multiple questions on one board creates noise. If you need to answer three questions, use three boards, physically separated.
2. Physical beats digital. A physical indicator — a magnet, a flip card, a lamp — has a presence that a pixel on a screen does not. It exists in the same space as the process it represents. When the operator flips the indicator, they have acknowledged the state change. When a sensor updates a screen, no human has acknowledged anything. Physical creates accountability.
3. The abnormal state must be obvious to a stranger. If I walk into your area for the first time, I should be able to identify an abnormality within five seconds, without explanation. If I need someone to explain the colour code, the layout, or the metric, the system is too complex. Test it with a visitor. If they cannot identify normal versus abnormal in five seconds, redesign it.
4. The visual must connect to a response. A red indicator that does not trigger a response is worse than no indicator — it teaches people to ignore abnormalities. Every visual signal must have a defined response: who acts, what they do, and how fast. If the response does not exist, do not deploy the visual. Fix the response first.
When digital is the right answer
Visual management is not anti-technology. It is anti-complexity. There are situations where digital is the right answer — when data needs to be aggregated across multiple sites, when trend analysis requires historical data, when regulatory reporting demands traceable records. But these are reporting needs, not operational needs. The operational need — running the process, detecting abnormalities, triggering responses — is almost always best served by simple, physical, visual systems.
The test is always the same. Five seconds. Can you see the abnormality? If yes, the system works. If no, you are spending money to create noise. And noise is not management. Noise is just noise.
