Factory dashboards: designing a screen that changes what somebody does next

One audience, one decision, one screen

A display built to satisfy the operator, the shift leader, the plant manager and the group reporting pack simultaneously will serve none of them. The operator needs to know whether the station is ahead or behind and whether help is coming; the shift leader needs to see where the constraint is right now; the plant manager needs yesterday against plan and the reasons. Those are different time horizons and different granularities. Separating them produces simpler screens that people read, and it also exposes requests for measures that no identified person would act on, which is the moment to decline politely.

Floor screens are read at a glance, from a distance, by someone holding a part

The physical conditions are unforgiving: bright light or poor light, a viewing distance of several metres, safety glasses, gloves that defeat a touchscreen, and a couple of seconds of attention. That rules out dense tables, small type, subtle colour differences and anything requiring interaction. It also means colour must carry meaning that survives colour vision deficiency, using position, shape or text alongside it rather than red and green alone. Test a prototype by standing where the operator stands, at the time of day when the light is worst, rather than approving it on a desktop in an office.

Refresh should match the decision, not the data

A figure that updates every few seconds implies somebody should react at that speed, and if nobody can, the movement is just visual noise that trains people to ignore the screen. Match cadence to the cycle of the response: an assembly station that can recover a small deficit within the hour benefits from a live count, while a quality trend reviewed at a weekly meeting should not be shown fluctuating hourly on a wall. Where a measure genuinely moves faster than anyone can act, showing a smoothed view with an alert on the exception is more honest and more useful.

Sprawl and the slow death of trust

Screens multiply because each is easy to add and none is ever retired. Eventually two displays show output for the same line with different figures, because one counts at final inspection and the other at the last machine, and the resulting argument discredits both. Keeping a register of screens, each with a named owner, a stated audience and a definition source, prevents most of this. So does a periodic review that asks who looked at each one and what changed as a result, with removal as the default answer when nobody can say. A smaller set of trusted displays outperforms a wall of contested ones.

The screen is not the management routine

Visual information only works inside a habit: a short stand-up in front of it at a fixed time, a person who asks about the outliers, an agreed response when a measure goes the wrong way. Plants that install displays without the routine find them ignored within weeks, and then conclude that operators are not interested in performance data. The routine can exist without any screen at all, as generations of paper boards demonstrated, but a screen without a routine has no mechanism to change anything and becomes expensive decoration on a wall people walk past.

Frequently asked questions

Are screens better than a whiteboard on the floor?

For anything the team writes on and discusses, a physical board often wins, because writing a reason by hand is quicker than navigating a system and the act of writing it involves the team. Screens win where data is automatic, where remote sites need the same view, and where history has to be retained. Many plants use both deliberately: automatic measures on a display, human commentary and actions on a board beside it, with nobody pretending the display captures what the team said.

Who should build them, IT or the production team?

The audience should specify what decision the screen supports and what they will do about each state it can show. Whoever builds it needs access to reliable definitions rather than raw tables, otherwise every screen re-implements a measure slightly differently. The arrangement that works is a small set of governed measures maintained centrally, with local teams composing their own views from them. Fully self-service building with no shared definitions produces the conflicting-figures problem quickly.

How do we handle a target that the team consistently misses?

Show it, but show alongside it the largest reason for the gap, because a display of failure with no route to a cause simply teaches people to stop looking. If the target has been unreachable for a sustained period, the honest options are to change it or to change the process, not to keep displaying an aspiration. Teams read a permanently red screen as evidence that management is not paying attention, which costs more credibility than the missed target itself.