Shop Floor Data Collection: Stop Guessing What the Job Actually Cost

It is 4:40 on a Thursday and the traveler for job 1187 has finally come back to the office.

Somebody has written "≈6 hrs" in the margin. Two people ran that job. One of them was pulled onto a rush order halfway through and never came back. The quantity column says 60. Inspection, two operations earlier, only ever passed 40.

Now you have to decide what this job actually cost, and the honest answer is that nobody in the building knows.

That is not a people problem. It is a shop floor data collection problem, and it is the single most common reason a manufacturing ERP produces numbers its own users quietly ignore.

Where the numbers actually go wrong

The clipboard is not the villain here. The delay is.

A paper traveler is a snapshot from the moment it left the printer. Work happens, and the record of that work sits in a pocket or on a bench until someone carries it to a desk and types it in. By then the shift is over, the operator has gone home, and the person keying it in is reconstructing a day they did not work.

This is still how most of the industry runs. A 2024 survey by the National Association of Manufacturers' Manufacturing Leadership Council found that 70% of manufacturers still enter data manually. In a separate survey of more than 600 US manufacturing leaders, L2L reported that half still run on manual frontline logs, paper trails and spreadsheets, and that 65% of frontline supervisors lose up to four hours a shift reconciling disconnected data by hand.

Four hours a shift. Reconciling. Not making parts.

Here is what that delay does to the three numbers you care about most:

• Labor hours become estimates. When people log their own time hours after the fact, they round. When they forget, a supervisor guesses at the end of the shift. Your standard-versus-actual comparison is then a comparison between one estimate and another.
• WIP is invisible until it is history. If the floor reports on Friday, then on Wednesday afternoon you genuinely cannot say how far job 1187 has got. You ask someone to walk out and look.
• Scrap becomes a number with no reason attached. A quantity gets crossed out and rewritten. Nobody records why, so nobody can fix the cause.

And there is a quieter failure underneath all of it. When several people work one job, paper cannot hold that shape. Shop owners describe exactly this: travelers would be great, except that two or three people are on the same job in different departments at the same time. So the traveler gets one set of hours, and the labor cost is wrong from the moment it is written.

Putting a screen on the floor does not fix this by itself. A terminal that only displays the traveler earns the industry's own insult for it: "paper on glass." What actually changes the numbers is capture at the bench, meaning the clock, the count, the reason and the components all recorded where and when the work happens.

What shop floor data collection actually means

Strip out the jargon and it is one idea.

Shop floor data collection means capturing production data (labor time, part counts, scrap, material issue, operation status) at the moment and in the place it happens, instead of reconstructing it from paper at the end of a shift.

That is it. The value is not the screen. The value is that the number is recorded by the person who knows it, while they still know it.

The operator's screen is not your work order list

This is where a lot of ERPs go wrong, and it is worth being blunt about it.

An operator standing at a machine does not need your work order table. They do not need column filters, saved views, or a hundred-odd rows spanning every work center in the plant. They need to know what they can run right now, here, and they need to start it without thinking.

So in Qualis the shop floor is its own screen. The operator picks the work center they are stationed at, once. That choice is remembered on that device, so tomorrow morning the tablet at the assembly bench opens straight onto the assembly bench's queue.

What they get is a queue of big, touch-sized cards. Each one carries the work order number, the operation, the parent manufacturing order, and the two numbers that matter at the bench: Completed and Scrap. Three buttons. Nothing else.

Two toggles shape the list:

• Mine or All. "Mine" is not just work assigned to you. It is your work plus the unassigned pool, which is what operators actually pull from. Nobody gets stranded staring at an empty screen because a planner did not assign a name.
• Ready, Blocked, or All. "Ready" deliberately includes operations already in progress, because starting a job advances its status, and a card that vanished the instant you tapped Start would be a bug you feel rather than read.

The header is honest about it too: "Your work queue by work center: start, report and move on."

Start is one tap, and it is your clock, not the job's

Tap Start. Three things happen at once.

The operation moves from Ready to In Progress. A green Labor running marker appears on the card. And the Start button becomes Pause, because a stopped timer should never offer you a pause, and a running one should never offer a second start.

Now the part that fixes job 1187.

Clock entries belong to the person, not to the job. Two operators on the same operation each accrue their own labor time. When one of them clocks out, it closes their entry and only theirs, because a colleague's running clock is not theirs to stop. The work order shows "2 other operator(s) clocked in" so nobody is working in the dark, and it accumulates the total as "Logged: 145 min labor / 90 min machine."

Labor time and machine time are tracked separately, which matters more than it sounds. An hour-long unattended cycle where the operator swapped a part in two minutes is not an hour of labor, and costing it as one is how shops end up quoting themselves out of work.

This also closes the loophole that makes shared terminals untrustworthy. When a floor runs on one shared login, you get buddy punching and per-operator data you cannot defend. Here the clock is attached to whoever is signed in, so the labor record survives contact with an audit.

Report: good, scrap, and why

Tapping Report opens the bench sheet, which holds everything that happens at that station in one place.

Top to bottom: the clocks, the quantities, this operation's component lines with Required / Issued / Remaining, and the finished-goods rows with Expected / Received / Remaining. The operator who consumes the components and the operator who books the finished units are the same person at the same bench, so they are not sent to three different screens to say so.

Two details are worth pausing on.

Materials are filtered to this operation. The sheet lists only the components consumed at this step, not the whole bill of materials. If a component is lot or serial tracked, issuing it forces a lot or serial pick, so the traceability chain never gets an optional field where it needs a mandatory one.

Scrap is never an unexplained number. Type a scrap quantity above zero and a Scrap Reason appears. Choose Other and a Scrap Details box appears, marked required, and the Report button greys out until it is filled in.

That last behaviour is rarer than it should be. Plenty of systems ship scrap reason codes. Far fewer make one mandatory. Oracle Fusion treats the reason as explicitly optional when an operator reports scrap at a workstation, Dynamics 365 lists "None" among the valid reasons on its production floor execution screen, and standard Odoo records the scrap reason as an optional tag. In SAP, tying a reason to scrap so that it is required only when scrap is actually reported is typically reached through a transaction variant or a BADI enhancement rather than standard configuration.

Three scrapped parts with no reason is a number. Three scrapped parts marked cracked during assembly is the start of a corrective action.

The guard that stops phantom good parts

Back to the 60 that should have been 40.

In a Qualis routing, an operation can be flagged as a control point, meaning the step where the part is actually checked. From that moment it caps what comes after it. A downstream operation cannot report more good quantity than the preceding control point has passed.

Try it and the report is refused outright, in plain language:

Cannot report 60 completed at operation 30: a preceding control point has only completed 40. Complete the earlier control-point operation first.

Not a warning that logs and lets you carry on. The report does not land.

There is a second ceiling on top of it: you cannot report more good parts than the manufacturing order asked for. That one is a per-organisation setting, so shops that legitimately over-produce can allow it, and shops that count scrap toward order completion see the remaining ceiling shrink accordingly.

This is the difference between a gate and a note. The MES literature has a good test for it: if a check fails and work can still continue, what you have is a reporting tool, not a control tool. A control point that cannot stop the next operation is documentation.

The market is genuinely split here, and it is worth knowing which side a system sits on before you buy it:

• Oracle Fusion Cloud Manufacturing builds the ceiling into its standard work execution model. Quantity becomes ready at an operation only once it is completed at the prior count point operation.
• NetSuite can enforce it, but the preference that checks completed quantity against prior operations ships set to "No Verification", so out of the box a downstream operation can be completed for more than the operation feeding it, without a warning.
• SAP raises a message rather than a fixed ceiling in standard production order confirmation, and how hard that check bites is a Customizing setting per plant and order type, ranging from no check at all, through information and warning, up to a blocking error. It is also generic: it applies to any predecessor operation, whether or not that operation is an inspection.
• Epicor Kinetic flags over-reporting with a shop warning and then lets the operator continue. A hard cap is custom BPM work.
• Odoo's standard Shop Floor screen lets an operator register whatever quantity they type. There is no native over-production ceiling, and none inherited from the quantity an upstream quality check passed. Its quality control points record a failed check and prompt a quality alert, but blocking the work order is a separate action an operator has to take.
• Katana never reaches the question. Produced quantity lands on the manufacturing order rather than on each operation, so operation 10 never reports a quantity for operation 20 to exceed.

Qualis ties the cap to the quality checkpoint itself, on by default, with no configuration to remember. A failed check limits what the next operation can consume.

Where the operator screen lives in other systems

Here is the part nobody puts on the pricing page.

In most manufacturing ERPs, the screen your operators stand at is not part of the thing you just bought.

• Odoo runs shop floor execution through a dedicated Shop Floor app, available on its higher tiers. The open-source Community edition covers manufacturing orders, bills of material and work orders, but ships no native operator terminal for the people on the line.
• Katana sells shop floor control as a separate add-on, on top of a subscription that already prices manufacturing management as its own add-on.
• NetSuite licenses the shop-floor operator seat separately, and puts per-operation reporting behind the Manufacturing Routing and Work Center feature, with deeper shop-floor control arriving as the Advanced Manufacturing SuiteApp.
• Epicor keeps machine-integrated execution in Advanced MES, a separate product that moves data back to the ERP through its own exchange module.
• Infor provisions Factory Track apart from CloudSuite Industrial, with its own databases and its own web address, connected by a point-to-point integration. That is a second system to administer, right down to linking user accounts across both.
• SAP puts the Production Operator Dashboard in SAP Digital Manufacturing, a separate cloud subscription. Standard S/4HANA gives the operator a confirmation screen; the guided cockpit with work queues and in-line quality capture comes from a separately priced product extension.
• MRPeasy is the fair counterexample. Its worker-facing kiosk is standard on every plan, right down to entry level. It gates on seats instead: kiosk reporting requires a login, so every operator who reports their own work needs a paid seat.

Across the tier-1 market that operator seat is quote-only. SAP, Oracle NetSuite, Epicor and Infor all route shop-floor licensing to an account executive rather than a published rate.

In Qualis, the shop floor is part of the manufacturing module. Same login, same permissions, same database. The work orders your operators clock on to are the same work orders that feed the work center load board your planner uses to promise dates, and closing the final operation is what mints the lot and serial genealogy your auditor asks for.

Finishing the job

Mark Done does one thing first that most systems forget: it stops your running timer before it closes the operation, so a terminal status never strands an open clock. The confirmation says so plainly: "WO-00063 will be closed and can no longer be reported against. Your running timer is stopped first." A colleague's clock is left alone.

Closing the final operation is what triggers finished-goods receipt and writes the genealogy record linking the components consumed to the units produced.

And when the machine is the problem rather than the part, Report a problem takes the operator straight into the maintenance request form. The person standing at the machine sees the fault first, and making them find a supervisor to relay it is how small faults become long stoppages.

Meanwhile the supervisor gets the opposite surface: the full work order table, filterable by manufacturing order, work center, operator, status and date, plus a Time Logs tab on every operation showing type, operator, start, end and duration. Not a summary. The actual entries, by name.

Frequently Asked Questions

What is shop floor data collection?

Shop floor data collection is capturing production data (labor time, part counts, scrap, material issue and operation status) at the moment and in the place the work happens, rather than reconstructing it from paper travelers at the end of a shift. The point is timeliness and attribution: the number is recorded by the person who knows it, while they still know it.

Do I need an MES, or can my ERP handle shop floor data collection?

It comes down to where the data originates. If you need machine-level signals such as PLC tags, sensor feeds and second-by-second line status, that is MES territory. If what you need is operator-reported execution, Qualis covers it natively: work orders reported at the operation, produced and scrapped quantities captured as operators declare them, materials issued with lot and serial capture, and complete genealogy minted when the final operation closes. All of that sits inside the same system that already holds your stock, costs and orders, with no second database to integrate.

How do you track labor time when two operators work the same job?

Each operator gets their own clock entry on the same operation. Both accrue time independently, clocking out closes only your own entry, and the operation shows how many other operators are currently clocked in. Because entries are tied to the signed-in user rather than to a shared terminal session, the resulting labor record is per person and stands up to an audit, which is exactly what a single shared login cannot give you.

How do you stop an operator reporting more quantity than the work order allows?

Two ceilings. The first caps reported good quantity at the manufacturing order's ordered quantity, with a per-organisation setting for shops that legitimately over-produce. The second is the control-point ceiling: an operation flagged as a quality control point caps what every downstream operation can report, so assembly cannot claim 60 good units when inspection only passed 40. Both refuse the report with a message naming the number and the operation, rather than logging a warning and continuing.

What data should you collect on the shop floor?

At minimum: labor and machine time per operator, good quantity, scrap quantity with a reason code, material consumption with lot or serial where the component is tracked, and operation status. Add finished-goods receipt at the final operation and you can close the loop from order to cost to traceability. Machine performance signals are a separate layer that requires equipment integration, so start with the operator-entered set, because that is where the costing errors actually live.

How long does it take to implement shop floor data collection software?

The operator side is deliberately shallow. Each person picks their work center once and the choice persists on that device, so there is no per-station configuration project and no equipment integration required to start reporting. The real preparation is upstream: having routings with sensible operations, deciding which of those operations are control points, and agreeing your scrap reason codes. Shops that already run structured routings are typically reporting from the floor far sooner than they expect.

The bottom line

The gap between what happened on the floor and what your ERP believes happened is where manufacturing margin quietly disappears. You cannot close it with a better spreadsheet or a firmer memo about paperwork, because the problem is not effort. It is distance and delay.

Close it by making the easiest place to record the number the same place the number is created: one screen, at the bench, with the operator's own clock running, the scrap reason mandatory, and a quality control point that can actually stop the next operation instead of merely noting its objection.

Then the hours on job 1187 are measured, not remembered. And the 60 is a 40, on the day, by the person who counted it.

Want to see it on your own routings? Explore Qualis and put a tablet on the bench.