The Ten-Second Piece Undo at the EDGEBIC Kiosk

At the EDGEBIC by User Solutions kiosk, an Undo button appears for about ten seconds after any piece tap: tapping it decrements that counter by one and writes an audit row recording the reversal. It is one of the smallest features in the platform and one of the most quietly important, because it settles a question every shop-floor terminal has to answer. What happens when a finger lands twice?

The wrong answers are familiar. Some systems make the count uneditable, so operators stop trusting the counter and start keeping a tally sheet beside it. Others let anyone edit anything at the machine, so the count becomes an opinion. EDGEBIC takes a third path: fix it instantly, but never invisibly.

The Window

During a run punch, the operator taps a green Good counter for each finished piece and a red Scrap counter for a bad one. Every tap is written immediately, so the count on screen is always the count in the record.

The moment a tap lands, an Undo button appears. For roughly ten seconds it is live. Tapping it reduces the counter that was last incremented by one, clamped at zero so a counter can never go negative, and then the button fades. If nothing was mis-tapped, the operator ignores it and it goes away on its own.

That is the whole interaction. No dialog, no reason picker, no numeric keypad, no confirmation. On a touchscreen next to a running machine, every extra step is a step that will not get taken.

The Audit Row

Here is the part that matters more than the button. An undo is not a silent rollback. It writes an adjustment record noting that a piece was reversed within the operator undo window, and that record sits in the same append-only history as every supervisor correction on that operation.

Two things follow.

Operators cannot hide a slip. A shift that shows fifty good pieces and eleven undos is telling you something about the counting habit at that station, and a supervisor can see it without asking anyone.

Nobody is punished for honesty. Because the undo is fast and visible rather than shameful, operators use it. A system that made correction awkward would get uncorrected counts instead, which is strictly worse. The record shows what happened and what was fixed, in order, with no gaps.

That discipline is the same one behind reading the kiosk punch history drawer, where every adjustment carries who, when, and why.

After the Window Closes

Ten seconds is deliberately short, and it will often have passed by the time an error is noticed. That case has its own path, and it involves a second person.

A supervisor opens the punch history drawer on the operation, finds the closed punch, edits the piece count or hours, and types a mandatory reason. The original value is preserved next to the correction rather than overwritten, so the history reads as a chain of events rather than a final state. Then the daily totals are rebuilt from the current punch set, so the corrected number flows through to the plan and the reports. Both halves of that flow are covered in how a supervisor corrects a kiosk punch and how punches roll up into daily actual hours.

Why Piece Counts Deserve This Care

A piece count is not just a number on a screen. Good pieces are what roll into the day's actual output, and only good pieces: scrap is recorded separately with its cause, and rework is counted on its own field. See good, scrap, and rework counts explained.

From there the count travels. It sets the observed rate for the day, which is pieces divided by production hours and which is what tells you whether the routing standard still reflects reality. It feeds the quality side of OEE. Where pieces are the measure a shop records, it is the basis from which hours are derived, so a double-tapped count becomes overstated hours and a plan that thinks more work is behind you than actually is. See how hours and pieces convert on the shop floor.

One extra piece on one job barely moves anything. A habit of uncorrected extra pieces across a quarter moves everything, and it moves it in the direction that makes your shop look better than it is, which is the worst direction for a number to drift.

A Worked Moment

CNC-1, mid-run, operator counting as parts come off.

The operator taps Good for the fourteenth piece. The counter reads 15. They had already tapped for that part while it was still in the fixture. The Undo button is on screen, they tap it, and the counter returns to 14. Total elapsed time: about two seconds. An adjustment row is written noting the reversal.

At shift end, the day's totals are correct with no reconciliation, no note left for the supervisor, and no memory required. The supervisor reviewing the operation later sees one undo on a fourteen-piece run and thinks nothing of it, which is the appropriate reaction.

The Bottom Line

The ten-second undo gives an operator one fast, obvious way to fix the mistake they just made, and it makes that fix part of the record rather than a hole in it. Beyond the window, corrections require a supervisor and a reason, and the original value is kept forever. Between them those two rules produce piece counts a planner can quote without hedging. See the whole capture loop in the shop floor execution guide, or explore EDGEBIC.