Why nobody trusts your production dashboard — and how to fix it

Who this is for

Lean and continuous-improvement engineers, production and plant managers at South African factories who have watched a dashboard, andon screen or visual-management board launch with energy and quietly become wallpaper — and who suspect the next, prettier screen will meet the same fate.

The three-month lifecycle of a factory dashboard

The pattern is consistent enough to schedule. An OEE dashboard or andon board goes up showing downtime, counts and output against plan. Week one, everyone looks at it. Month one, supervisors quote it in meetings. Month three, nobody consults it, nobody challenges it, and someone has quietly stopped updating it. In lean-engineering communities the post-mortems repeat the same finding — one engineer put it as "fancy screens filled with data that ultimately the employees don't trust", another as visual management that "has become decorative, not effective".

The usual response is to blame culture, training or the display itself, and to relaunch with a better screen. It fails the same way, because the screen was never the problem.

Why manual data entry kills trust

Trace any number on a distrusted board back far enough and you find a person: an operator asked to log stops between actual work, a team leader reconstructing the shift from memory at handover, a clerk retyping paper tallies into a spreadsheet on Monday.

The people on the floor know this — they are the data entry. They know the logged downtime was an estimate, that the short stops never got recorded, that the reason code was whatever was quickest to click. So when the board contradicts their experience of the shift, they don't doubt themselves; they doubt the board — and they're right to. Distrust of manually-entered production data isn't cynicism. It's an accurate assessment of how the number was made.

What machine data collection looks like

The alternative doesn't require new machines or a controller upgrade. Shop-floor data collection can happen directly at the equipment on almost any line, including decades-old machines with no PLC:

• Run/stop state and downtime tracking — read from the PLC over its native protocol where one exists, or from a current clamp on the drive motor where one doesn't. Every stop gets a true start time and duration, automatically, including the micro-stops no one ever logs by hand.
• Counts — from a photo-eye, proximity switch or the machine's own counter, not a tally sheet.
• Quality — a sensor on the reject gate, so scrap is counted where it happens.
• Reasons — the one input that should stay human. A machine knows it stopped at 10:42 for six minutes; only the operator knows it was a jam on the infeed. A one-touch terminal captures the why in seconds, and because it's context rather than measurement, it's the one entry nobody disputes.

All of this is read-only and non-invasive — installed while the line runs, with no change to the machine's controls. It's the same instrumentation that delivers OEE on an old line without a new PLC.

The South African clause: the record must survive load shedding

There's a second way a dashboard loses the floor's trust here: gaps. If the data collection dies with every stage of load shedding, the record develops outage-shaped holes — and the stops that cluster around power events, often the most interesting ones, go missing. A board that visibly missed yesterday's chaos is a board nobody believes about today. Edge devices that buffer readings locally and sync when power returns keep the record continuous, so an outage is an event in the data rather than a hole in it.

How to fix it: the sequence

• 1. Audit provenance. For every number on the board, write down where it actually comes from. Anything a person types from memory is a trust liability to be retired.
• 2. Instrument events at the machine. PLC reads where possible; current clamps, photo-eyes and reject-gate sensors where not.
• 3. Keep humans for reasons only. One-touch reason codes; no event logging by hand, ever again.
• 4. Buffer at the edge. Local logging through outages, syncing on return — no gaps.
• 5. Feed what you already run. Machine-sourced data can flow into the lean boards, huddle screens and systems your teams already know, giving them credible numbers for the first time — replacing the data source, not the workflow.

Then apply the only test that matters: do people consult the board unprompted? When an operator checks the screen to settle an argument about last shift — because the 10:42 stop on the board is the 10:42 stop they remember — visual management has stopped being decorative.

This is the foundation under our OEE management solution, and the pattern behind the smart-factory OEE case study. It runs on the addaNet platform, so real-time production monitoring sits alongside energy, water and asset condition in one operational picture — and for the dashboards themselves, see Grafana for industrial IoT.

Frequently asked questions