How Many Kiosk Terminals Does Your Floor Actually Need?

A kiosk session in EDGEBIC by User Solutions binds to one work center and one machine instance, not to a person, so you count terminals by machines where punch timing matters rather than by headcount or by your work center list. How many shop floor terminals you need is a siting question with three inputs: which machines are contested, how far apart they stand, and where the network actually reaches.

Most shops get this wrong in one of two directions. Either they buy one terminal per work center because that felt symmetrical, and half of them go untouched, or they buy one terminal for the whole plant and it becomes a bottleneck in front of the bottleneck. Neither is a hardware problem. Both come from not looking at what the binding actually does.

What a Session Is Bound To

When the kiosk launches, the Select Work Center dialog asks for two things: the work center this session serves, and the Instance, numbered from one, meaning the physical lane or machine within it. Everything that follows is scoped to that pair. The NEXT UP panel offers only work the scheduler has queued at that work center. COMING UP shows what is behind it in the same queue. Punches land against that instance.

The kiosk is deliberately not a plant-wide viewer. There is no job search and no browsing other machines' queues, which is a scope decision rather than a gap: see what an operator sees at the kiosk.

Two consequences follow directly, and they are the whole basis of the terminal count.

First, a terminal is a machine's screen, not an operator's screen. It has no login and no account, only a free-text Operator box that stamps a name onto punches. Two people can use one terminal in a shift with no logout ritual between them, because there is nothing to log out of. See why the kiosk has no login. So headcount is the wrong denominator.

Second, the binding is not permanent. The header's Switch WC button reopens the Select Work Center dialog mid-shift. That single button is what makes a shared terminal a legitimate pattern rather than a compromise.

Four Patterns, and What Each Costs

The pattern that fails most reliably is one per department. It is not the sharing that breaks it, it is the distance. The kiosk's value comes from timestamps taken when the tap lands, and a terminal thirty paces away converts every tap into a decision about whether the walk is worth it. Operators answer that question honestly, and the answer is usually no, so the counts arrive at shift end from memory. At that point you have paid for a terminal and are getting planner-grade data out of it.

The pattern that works most reliably is a shared cell terminal within arm's reach of two to four machines, with a dedicated screen on any machine where changeovers are frequent enough that the operator is at the screen anyway.

Count From the Contested Machines

Start the count from a list you probably already have: the machines where people argue about what happened.

That list typically includes the bottleneck, because everything flows through it and the reschedule is most sensitive there. It includes the machine with the long or variable changeovers, because setup phase times are the one thing typed totals can never reconstruct. And it includes the machine that gets blamed when a job is late, because that argument is only settled by a punch history with times and reasons on it.

Every machine on that list gets a screen. Then look at what is left and group it by walking distance, not by department code or by process family. Machines an operator can reach without turning around can share. Machines that require a walk cannot, whatever the org chart says they have in common.

What is left after that is the honest "no terminal" set, and there is nothing wrong with it. A booth that runs one color a day, or a machine whose operator wears gloves all shift, is better served by a planner typing two grid rows each morning. See who should log actuals: the operator or the planner.

The Instance Question

Work centers with several physical lanes make the count less obvious, because the kiosk binds to one instance at a time.

The deciding question is whether the lanes run independently. If your four-spindle work center always runs one job across all four lanes together, one terminal bound to one instance is enough, and the operator never thinks about it. If the four lanes genuinely run four different jobs, one terminal means the operator re-binds through Switch WC every time attention moves, and re-binding four times an hour is the kind of friction that ends in nobody punching at all.

For genuinely independent lanes there are only two honest answers: a screen per lane, or accept that this work center is planner-primary. A single shared screen across independent lanes looks like a saving and reads as a bad punch history within a fortnight.

Siting: Power, Reach, and Network

Three physical constraints decide where a terminal can actually live, and one of them is easy to forget.

Power and wall space are obvious. Reach is nearly as obvious: the screen has to be usable by someone whose hands are busy, which in practice means at working height, out of the coolant spray, and not behind the guard that has to be opened to load a part.

Network coverage is the one shops underestimate, because the kiosk has no offline mode. It reads and writes the shared database live, so a terminal in a dead spot is not a slower terminal, it is an unreliable one, and unreliable is the state that teaches operators to stop bothering. See what the kiosk does when the network drops for the honest fallbacks. Survey coverage at the actual machine positions before you commit to a count, and be willing to move a planned terminal to planner-side entry rather than install one that works most of the time.

The Remember Setting Is Pattern-Specific

The Select Work Center dialog offers Remember this selection, which saves the binding so the terminal reopens on the same machine. Whether to tick it is not a preference, it is determined by which pattern that terminal is in.

On a dedicated terminal, tick it. The shift starts correctly bound and nobody spends a thought on it. If it ever reopens on the wrong machine, Switch WC fixes it in two taps and the new selection can be remembered instead.

On a shared cell terminal, leave it off. A remembered binding on a shared screen means the terminal reopens on whichever machine somebody last used, and the failure that follows is silent: an operator taps Start Setup on a job that belongs to the machine next door. Forcing the dialog on every launch makes the binding a deliberate first act of the session, which is exactly what you want when the screen serves several machines. See setting up the kiosk for your first shift.

A Worked Count

A fabrication shop has fourteen work centers. Two are welding cells with three independent bays each; the rest are single-machine.

The contested list is three long: the assembly bottleneck, the punch press with two changeovers a shift, and the mill everybody blames. Three dedicated screens.

The two welding cells run independent bays, so the choice is per-bay screens or planner-primary. The shop picks per-bay for the busier cell, because its bays run short jobs and the punch detail is worth having, and planner-primary for the quieter cell, where each bay runs a job for two days. Three more screens.

Four remaining machines stand in two adjacent pairs. Two shared terminals, Remember left off on both.

The last three machines are a booth, a deburr bench, and an oven. No screens; the planner types their days.

Final count: eight terminals for fourteen work centers and seventeen physical machines. Six machines have no screen at all, deliberately, and the punch data from the eight that do is better than it would have been from seventeen half-used screens.

The takeaway

Terminal count follows the binding: one work center, one machine instance, no person attached. Start from the machines where punch timing is contested, group what is left by arm's reach rather than by department, treat network coverage as seriously as power, and let Switch WC and the Remember setting do the work on shared screens. Deliberate gaps served by planner entry beat a full set of terminals nobody trusts. Explore the platform at EDGEBIC, see the upgrade path in RMDB to EDGEBIC, and read on in what is a shop-floor kiosk and the daily rhythm of a kiosk-run shop.