Shop floor data collection by work order and shift

Capture quantities and corrections at the operation, then reconcile machine signals and operator entries before shift close.

Kartik Deshpande··5 min read

Shop floor data collection starts at the point where someone can identify the work order, operation, and quantity that changed. Capture machine counts when the machine can identify what it ran, and ask an operator to supply context the controller lacks. Review both before closing the shift if the numbers will update production reporting.

Microsoft’s production floor execution documentation shows how workers can register jobs and report progress and scrap in an execution interface. A paper form or scanner can record similar events. Choose the interface based on how the operator knows the current job and how an entry reaches the system that owns production quantities.

Identify what happened before recording the quantity

Give each event an identifier, work order and operation, observed time and entry time, source, and quantity with unit. Keep a distinct event type for completion, scrap, rework, and correction. If a label scanner can send the same message twice, use the event identifier to count it once and save evidence of the second delivery when diagnosing a fault.

Specify whether a machine sends deltas or a running counter. With a running counter, store the observed value and calculate the difference within a known reset period. A machine that restarts at zero between batches needs a new counter period. Counting every reading as a new increment can inflate output even when every message arrives once.

Define the first and last point that count as completion. If pieces pass two stations, their sensors can each show throughput, but only the agreed finished operation reports good units to inventory. Keep the manufacturing work order as the reference for the operation and planned unit.

A useful station trial covers a normal run and a run that ends at shift change. Have the incoming operator take over the same work order without creating a new completion event for the output already recorded. Then compare both shifts with the work-order total and check that the sequence still shows which operator classified a held piece.

Reconcile one run without double counting output

Suppose a station’s counter rises from 12,410 to 12,530 for one work order. An operator reports 110 good pieces, seven scrapped pieces, and three awaiting inspection. The total explains the 120 pieces detected. Posting all 120 as good units overstates released output by ten.

If inspection releases two of the three held pieces and scraps the third, post an inspection disposition against the existing held quantity. The final totals are 112 good and eight scrapped, with zero remaining on hold. An inspection event carries its own time and reference, but it accounts for pieces the machine already counted.

Use the scrap rate only after setting the denominator and the cutoff for unresolved pieces. A shift report can keep held quantity separate and show that its scrap figure may change. Reconcile the inspected disposition before using the run to compare defect rates across stations.

RecordQuantity in the exampleTreatment
Machine counter difference120Observed pieces for the operation
Released good110, then 112Count the two later releases once
Scrapped7, then 8Keep the original disposition and revision
Held3, then 0Move these pieces when inspected
Sources behind the shift total
Observe
Counter and work-order reference
Classify
Good, scrap, or held
Review
Disposition and corrections tied to the original entry
An inspection disposition changes a quantity’s status without creating a second piece.

Check which records share a work order

Use a machine export and a shift log to inspect the keys that link their counts.

Choose how each station will capture and correct data

At a station with a stable work-order barcode and reliable network, scan at changeover and attach subsequent observations to the selected operation. If a station runs two jobs side by side, add a lane or fixture identifier so the operator can see which signal belongs to which job. Where a paper sheet is the best available source, keep the observer and original event time when entering it later.

During a pilot, unplug the connection long enough to observe what happens to one queued count. Restore connectivity and check whether the collector sends the same event twice or loses the original timestamp. Then correct a good count from the prior shift and verify the report shows both the original entry and the authorized revision.

Give a supervisor a queue of unassigned machine events, unexpected units, and unresolved dispositions. Review it before approving shift totals. For event intervals that carry equipment or cause details, keep the downtime record separate so quantity adjustments do not rewrite why a stop occurred.

Keep ownership of repairs clear. The collector should flag a missing identifier, while a supervisor with the production context can link the event to the right job. Require the approved reason before a corrected entry changes the published total. Limit review access where production details or labor identifiers should stay within a plant.

Keep machine setup records apart from output quantities. A setup event can explain why a line was unavailable for half a shift without changing how many pieces passed inspection. When changeover runs produce units, use the same good, scrap, and held rules as normal production and label the run type so analysts can choose whether to compare it with steady production.

Show which station last reported and when. If the operator entered a completion but the machine feed is silent, leave the machine comparison incomplete until it resumes. If a local tally is later uploaded, match its work order and event sequence before using it to clear the gap.

How Permute connects the evidence

Permute is the context layer we sell between business data and the AI tools people already use. Match work orders, station exports, and shift logs using the plant’s identifiers, with the original entries still available to compare.

Define reported good quantity and the rules for corrections once. A supervisor can then ask which operation accounts for a discrepancy and trace the answer to the source records they can access.

What stays in the source systems

Keep machine control and authoritative production postings in the existing execution system. Permute reads connected data for analysis and review. It does not run the line, record a machine signal that never arrived, or write the corrected quantity back.

Close the shift with a record you can trace

Shop floor data collection works when an event has enough context to follow it into the shift total. Use a machine signal to count observations and an operator or inspection record to explain their disposition. Tie corrections to the original work order and event. The supervisor can then approve a quantity without losing the evidence that explains it.

Walk through a shift quantity discrepancy

Bring the event history and the records the supervisor uses to approve the total.

Questions about shop floor records

Do operators have to enter every count manually?

Use machine counts where they are trustworthy and tied to the right operation. Ask operators to record decisions or observations the signal cannot explain, such as a hold or a correction.

How should you handle a device with the wrong time?

Keep its reported timestamp and the capture time separately. Flag the offset and correct reporting from an approved reference instead of rewriting the raw event.

Should rework count as a new completed piece?

Count the unit once in finished output when it meets the release criteria. Track the rework operation separately to retain the extra labor and process history.