blog

Late Scrap and Rework: Correct OEE Without Rewriting the Shift History

By: Lauren Dunford

By: Guidewheel
Updated: 
November 14, 2026
8 min read
Late Scrap and Rework: Correct OEE Without Rewriting the Shift History

No items found.

When quality checks identify scrap or rework after a shift, correct overall equipment effectiveness (OEE) for the run that made those units. Retain its original report and runtime. Keep production time separate from quality-confirmation time, and show the revised first-pass quality and OEE beside the earlier provisional values.

A unit that needed rework remains a first-pass quality loss even if it later ships. Record its eventual recovery separately. Do not add it as another original production unit or move its defect into the shift that happened to discover it.

A six-facility plastic packaging manufacturer struggled with unreliable, after-the-fact OEE reports. With Guidewheel's granular machine, product and shift data, its team found a correlation between downtime and non-conforming products. Its quality leader reported using downtime alerts to help keep non-conforming material from customers. That case shows why production and quality evidence belong together; the correction method below addresses the separate question of what to do when quality evidence arrives later.

Record What it answers
Production interval When and where the original unit was made
Quality confirmation When its disposition became known
Provisional OEE What the shift report showed with then-available evidence
Finalized first-pass OEE What the originating run achieved after reconciliation
Rework/delivery record What happened to the same unit afterward

Keep production time and confirmation time separate

Attach a defect to its originating run, using traceable production identity. The discovery timestamp explains the correction; it does not decide which run owns the quality loss.

A useful correction record includes asset, job, product, lot or unit identity, production start/end, inspection time, disposition time and the person confirming the outcome. Keep a separate entry for when the reporting record was changed. These timestamps may be different, and each serves a different question.

Label shift-end values provisional when inspection is incomplete. If the initial report treats produced units as provisionally good, explain that assumption and the outstanding inspection scope. Do not silently present unconfirmed quality as a finalized result.

Agree the quality boundary before correction. This article uses first-pass good units: output that met the declared requirements without rework. The first-pass-yield and scrap-rate guide explains the distinction between initial success and eventual recovery.

The confirmation point depends on the process. Consider these possible workflows.

Possible workflow Link the later result to Keep separate from production time
Lumber graded after drying and surfacing The traceable originating lot and run Final grading and disposition times
Packaging checked after the recorded run The job, lot or identified units Check result and disposition times
Food manufacture or packaging awaiting release The declared production stage and traceable lot The responsible team's release decision and time

A running machine cannot confirm any of these outcomes. Use the relevant quality record, and distinguish when the condition was detected, when its disposition was confirmed and when the report was restated.

Keep planned production time and runtime tied to the original interval. A late inspection result changes the good count, quality factor and OEE. It does not establish that the machine ran fewer minutes. Correct time only when separate evidence shows the original time record was wrong, and log that as its own correction.

Preserve the initial snapshot and append the confirmed version. This is a recommended reporting control. The system of record might be an agreed production system, quality record or controlled worksheet; confirm the actual implementation rather than assuming a monitoring platform has a particular version-history feature.

Restate two runs with a visible quality correction

Recalculate each originating run with the revised first-pass good count and the original time boundary. The following two-run example is illustrative, with invented inputs.

Run A has 240 planned production minutes, 200 runtime minutes and 300 original units at an approved ideal cycle of 30 seconds per unit. Its initial report records 294 first-pass good units and six known first-pass failures. Later inspection identifies another 12 scrap units and eight units requiring rework, all from the initially accepted group.

Run B has 240 planned minutes, 210 runtime minutes and 400 original units at 25 ideal seconds per unit. It initially records 392 good units and eight known failures. Later inspection finds four additional scrap units and six requiring rework from that run's initially accepted group.

Measure Run A Run B
Original total units 300 400
Initially reported good units 294 392
Additional first-pass failures confirmed later 20 10
Finalized first-pass good units 274 382
Initial quality 98.0% 98.0%
Finalized quality 91.33% 95.50%
Initial OEE 61.25% 68.06%
Finalized OEE 57.08% 66.32%

For A, finalized quality is 274 ÷ 300 = 91.3333%. Finalized OEE is (274 × 30) ÷ (240 × 60) = 57.0833%. For B, quality is 382 ÷ 400 = 95.5%, and OEE is (382 × 25) ÷ (240 × 60) = 66.3194%.

Availability and performance retain their original inputs. The correction reduces first-pass good output; it does not remove runtime or add another stop. If a later investigation also finds a time-record error, keep that change separate so reviewers can see both causes of the restatement.

Every unit is assigned once to its original production total. The additional failures came from the initially reported good group, so subtracting them is appropriate. If the initial report had already classified those units as rejects, subtracting them again would create a second quality loss.

Notice that Run B's larger percentage is not a reason to assign A's defects to B's confirmation shift. The production identity governs attribution. The confirmation-time view should show the inspection workload and discovered defects, while the production-time view shows the revised performance of the runs that made them.

Track rework without creating a second good unit

Keep rework activity linked to the original unit, and separate first-pass performance from eventual accepted delivery output.

Suppose later rework recovers six of A's eight newly identified rework units and five of B's six. Eleven existing units become deliverable. They add no first-pass good units to either originating run. A remains at 274 first-pass good units and B at 382.

The recovery record should show the original run, rework operation, processing or labor time, result and final disposition. The remaining units may still be pending or ultimately scrapped. Quality must confirm their outcome before they enter accepted-output reporting.

If rework uses the original machine during a later shift, that machine's later time record must still reflect the activity. Mark the work as rework and use the agreed separate operation or order convention. Do not hide its capacity consumption because the original unit was already counted.

A production-pass counter and a unique-original-unit count can therefore differ legitimately. A second processing pass is an event, not a new original unit. Keep that distinction visible when reconciling cycles and quantities.

For delivery reporting, count each recovered unit once under its confirmed status. Do not add the same unit both through a corrected original good count and through a rework-recovery posting. The recommended identity link makes the relationship reviewable even when production and quality systems use different event timestamps.

Keep the correction ledger and review both reporting views

Maintain a correction ledger that explains the old value, new value and evidence. Review production-period performance separately from the period when quality confirmed the change.

Correction field Example purpose
Originating run and unit/lot scope Identify the production record affected
Production interval Attribute the loss to the right run
Confirmation timestamp Explain when the new evidence arrived
Original quantity/status Preserve what the initial report showed
Confirmed change Record scrap, rework or acceptance without overlap
Revised first-pass good count Support recalculation
Reason and evidence Link inspection or disposition record
Owner and revision Show who confirmed and changed the result
Affected reports Identify shift, monthly and other downstream summaries

Agree a normal finalization window with quality and operations. It should reflect the process's inspection delay, not a universal deadline. Define what happens when evidence arrives after that window. A finalized report may need reopening, while the prior version remains available.

Use the OEE governance definition sheet for the standing policy. Use the OEE systems-boundary guide to decide which record owns time, original quantities and quality disposition.

OEE resource hub connects the corrected number to the choice of which loss to improve next.

Guidewheel's machine information can support the time and operating context. Teams still need production identity and confirmed quality outcomes to make this correction. The ledger described here is a recommended method, not a claim that machine current proves accepted output or that a specific native restatement workflow is available.

When a report changes, tell its users which runs were restated and why. Separate the effect of better quality information from any physical deterioration in the current shift. This prevents a late discovery from becoming a misleading explanation of today's machine behavior.

Discuss your production and quality reporting boundaries with Guidewheel.

Frequently asked questions

Does late scrap change the original runtime?

No, unless independent evidence shows the time record itself was wrong. Late scrap changes the originating run's good count and quality-dependent OEE. Retain its runtime and log any separate time correction independently.

Which shift owns defects found the next day?

The originating production run owns the quality loss when traceability supports that attribution. The next day's confirmation record shows when the defect became known. Preserve both timestamps instead of using discovery time as production time.

Can recovered rework restore first-pass OEE?

Recovered rework does not restore first-pass success. Keep the original failure in first-pass quality and record later acceptance separately. Rework processing also consumes capacity in the interval where it occurs.

What if the originating run cannot be identified?

Hold the correction in an unresolved quality exception with the available lot and inspection evidence. Do not assign it arbitrarily to the discovery shift. A quality and production owner should resolve the scope before a run-level restatement.

Should a previously published monthly report change?

Follow the agreed correction policy when later evidence materially changes its production-period result. Retain the earlier version, explain the revision and identify affected users. A finalization date should not make confirmed quality losses disappear.

About the Author

Lauren Dunford is the CEO and Co-Founder of Guidewheel, a FactoryOps platform that empowers factories to reach a sustainable peak of performance. A graduate of Stanford, she is a JOURNEY Fellow and World Economic Forum Tech Pioneer. Watch her TED Talk: the future isn’t just coded, it’s built.

GradientGradient