When Is a Downtime Action Really Closed? An Evidence Checklist for CI Teams

Close a downtime action when the evidence shows that it met the agreed loss-reduction test under comparable operating conditions. A completed repair records work performed. Action closure records an operating result. Continuous improvement (CI) teams need a defined baseline, verification window, acceptance owner and reopen rule to connect the two.
In Guidewheel's Fibrix Filtration case study, the team identified repeated winder stops associated with falling air shafts, modified the tracking system and checked that the issue no longer caused unplanned downtime. The useful lesson is the sequence: observe the recurring loss, make the change and examine what happened afterward.
This checklist turns that follow-through into an evidence audit. The fictional example below first fails its closure test despite a completed repair. After rework, a second check passes. Keep this record in the maintenance or improvement workflow your team already uses.
Scroll horizontally to view all columns.
| Evidence | What it establishes | What it cannot establish alone |
|---|---|---|
| Completed work and handover record | The agreed work was performed and handed back | The target loss fell |
| Comparable event history | The observed loss before and after the change | The cause of the change |
| Operator observations | How the intervention behaved during the reviewed work | Performance in conditions never tested |
| Production and quality records | Output and quality results within their defined boundaries | Recovery across the whole production route |
| Acceptance record | Who reviewed the evidence and what they accepted | That future recurrence needs no review |
For the broader route from data capture to improvement, start with the machine monitoring guide. Use the checklist here when deciding whether one action can leave the open list.
1. Agree the downtime action closure test before work starts
Define the result that would justify closure before the team sees the outcome. A threshold chosen afterward can turn almost any change into apparent success.
Write down the specific loss, affected asset and operating condition. If the action addresses recurring material waits, measure those waits. A decrease in total downtime could conceal an unchanged material problem.
Use seven fields to establish the test:
- Problem: the observed event, its boundary and the conditions in which it occurs.
- Baseline: the original period, loss measure, denominator and source records.
- Action: the proposed change and person authorized to perform it.
- Acceptance rule: the numerical threshold and required quality or other checks.
- Verification window: the relevant shifts, jobs and exposure to the problem.
- Acceptance owner: the person responsible for reviewing the outcome.
- Reopen rule: the result or adverse effect that requires renewed action.
Choose the window around meaningful exposure, not an arbitrary number of calendar days. Five shifts that never run the troublesome product do not test a product-specific intervention. For an infrequent failure, a short period without recurrence may be insufficient evidence.
The logic follows the Plan, Do, Check, Act improvement cycle: evaluate the change against the intended result, then standardize it or investigate again. The thresholds and records below are a practical application, not a prescribed industry standard.
2. Compare the same loss on the same basis
Keep the measurement boundary consistent between baseline and verification. Record differences that could explain the result before crediting the intervention.
Check the asset, product or job family, operating mode, shifts, stop threshold and reason-code definitions. Confirm that both windows contain relevant opportunities for the problem to occur. Also inspect missing signal intervals and unclassified events. A smaller reason category is inconclusive if more stops became unknown.
Define the denominator in plain language. Stop minutes per scheduled production hour, stop minutes per running hour and minutes per changeover answer different questions. Use the same definition throughout the comparison.
In the worked example, scheduled production hours include the measured unplanned stops. Planned nonproduction time is excluded consistently. This keeps the denominator independent of whether the intervention shortened the stops. It is an exposure measure for the chosen loss, not an OEE calculation.
After a repair, retain four distinct milestones:
Scroll horizontally to view all columns.
| Milestone | Evidence to retain | Responsible role |
|---|---|---|
| Work completed | Authorized work and handover record | Maintenance or work owner |
| Machine restarted | Observed activity consistent with the validated machine state | Operations and technical owner |
| Accepted production resumed | Output crosses the agreed count boundary and passes required quality checks | Production and quality owners |
| Improvement sustained | The target loss and adverse-effect checks pass over the agreed exposure | Acceptance owner |
The post-repair production recovery guide explains the gap between restart and sustained accepted output. Recovery from one stop does not establish that a recurring loss has been reduced over representative work.
Ask the operators who used the changed process whether the symptom returned or another task became harder. Review their account alongside measured records. Where counts disagree, use the runtime and production-count reconciliation guide before accepting an output claim.
3. Reject completed work that fails the agreed test
Keep the action open when its evidence fails the acceptance rule, even if the repair is finished and production has restarted.
Consider a fictional continuous-profile line with a recurring stop. Maintenance completes the intervention, operations confirms restart and quality accepts the profiles being produced. The CI team must still check whether the recurring loss fell enough.
The baseline contains 60 minutes of the defined stop across 40 scheduled production hours. The team agrees to accept no more than 0.75 stop minute per scheduled production hour, with comparable jobs and shifts and no unacceptable adverse effects. Each verification window covers 40 scheduled production hours with matched trigger opportunities.
Scroll horizontally to view all columns.
| Window | Target-stop minutes | Scheduled production hours | Loss rate | Closure result |
|---|---|---|---|---|
| Baseline B1 | 60 | 40 | 1.50 min/hour | Reference |
| V1: after initial work | 52 | 40 | 1.30 min/hour | Fails the agreed limit |
| V2: after rework | 24 | 40 | 0.60 min/hour | Meets the numerical limit |
Calculate each rate as target-stop minutes divided by scheduled production hours. The limit permits at most 0.75 × 40 = 30 stop minutes in each verification window.
V1 improves on the baseline by (1.50 − 1.30) ÷ 1.50 × 100 = 13.3%. It still exceeds the 30-minute allowance by 22 minutes. The work is complete; the closure test has failed.
After authorized rework, V2 records 24 stop minutes, six minutes below the allowance. Its loss rate is 60% lower than baseline: (1.50 − 0.60) ÷ 1.50 × 100. The acceptance owner then checks the operating context, operator observations and adverse effects before closing the action.
All inputs, windows and thresholds in this example are invented. They are neither a Guidewheel customer result nor a universal performance target. Passing this before-and-after check supports the agreed operating decision; it does not isolate causation if other relevant conditions changed.
Use three distinct dispositions:
- Pass: comparable evidence meets every acceptance criterion.
- Fail: usable evidence shows that at least one criterion was missed.
- Inconclusive: missing data, insufficient exposure or changed conditions prevent a defensible decision.
For an inconclusive result, name the missing evidence and next verification step. Do not record it as either a successful intervention or a proven failure.
4. Attach the evidence and record who accepted it
The closure record should let another person reconstruct the decision without relying on the work owner's memory.
This completed illustrative record follows V2. B1, V1 and V2 identify separate source windows in the example; a real record also needs their exact dates and timestamps.
Scroll horizontally to view all columns.
| Closure field | Example entry |
|---|---|
| Action and asset | CI-47, recurring stop on continuous-profile Line A |
| Work history | Initial intervention completed; V1 failed; authorized rework completed |
| Baseline | B1: 60 target-stop minutes across 40 scheduled production hours |
| Verification | V2: 24 target-stop minutes across 40 comparable scheduled production hours |
| Acceptance rule | No more than 0.75 min/hour, with comparable exposure and no unacceptable adverse effects |
| Measured result | 0.60 min/hour; numerical criterion met |
| Evidence sources | Timestamped event export, production schedule and job record, operator handover notes, quality review |
| Operator check | Reviewed shifts confirm intended behavior; no new operating difficulty reported |
| Adverse-effect check | Quality review and adjacent-process checks found no unacceptable effect during V2 |
| Acceptance owner and decision | Line A CI lead reviewed V2 with production and quality owners; action closed |
| Reopen trigger | Rate above 0.75 min/hour in a comparable review, or an unacceptable adverse effect |
Retain V1 alongside V2. Record the rework performed and any other changes between the windows. Preserve the original baseline rather than replacing it with a more favorable comparison.
Attach the actual event, schedule and quality records, with their timestamps and owners. Keep planned nonproduction time, missing intervals and event corrections visible. A summary table is useful navigation; the source records substantiate it.
Match the claim to the action. Reducing a recurring line stop does not prove that a downstream packing constraint disappeared or that an order shortfall was recovered. A delivery-capacity claim needs accepted-output evidence through the relevant production route.
Where a maintenance or improvement system owns the action, retain that responsibility. Monitoring evidence supports the review; this checklist does not assume automatic closure approval in the monitoring software.
Use the action-closure sheet in the Downtime Baseline and Recovery Workbook to capture the comparable baseline, acceptance threshold, result and reopen rule. The workbook includes illustrative completed examples and blank records with calculation recipes. Example results are static: calculate and check new inputs manually.
5. Standardize the change and keep a reopen rule
After acceptance, update the instructions, handover or maintenance practice affected by the change. Name who will check that the new method remains in use.
Use the downtime reduction playbook to connect verification with standard work and the next improvement. In this example, the CI lead also keeps the 0.75-minute/hour limit in the recurring review.
Reopen CI-47 if the defined loss rate rises above that limit in a comparable review, or an unacceptable adverse effect appears. The acceptance rule allows some residual stops. Recurrence below the limit prompts review; it does not automatically overturn closure.
A new product, altered stop threshold or revised reason mapping can also invalidate the comparison. Record the change and decide whether to investigate the original intervention or establish a new scope. Preserve what the team knew at the original closure date.
Guidewheel's practical operating platform for manufacturing brings machine information and operating context together, with software and expert support to help teams act. For action closure, that shared evidence helps the customer assess what changed and what deserves attention next.
Choose one completed action and check whether another reviewer could follow its baseline, verification and acceptance record. Discuss the operating evidence your team needs with Guidewheel.
Frequently asked questions
Is a finished work order enough to close a downtime action?
It establishes work completion. The improvement action also needs evidence against its agreed outcome test. Keep maintenance completion and CI verification as separate statuses when the systems allow it, so the team can see what remains to be checked.
How long should the verification period be?
Use enough representative exposure to test the specific loss. Relevant jobs, shifts, operating hours and trigger opportunities matter more than a fixed number of days. Agree the window before reviewing results, and record any reason for extending it.
Who should accept action closure?
Name a person with authority to assess the operating outcome, supported by the work owner and relevant production or quality owners. The record should identify the evidence they reviewed, the decision date and any limits to the conclusion.
What if output improved but the same stop still occurs?
Assess the original loss against its own acceptance rule. Output may improve because of another change while that loss remains. Residual stops may be acceptable under the agreed threshold, but a claim that the stop was eliminated would require different evidence.
What evidence should reopen a closed action?
Use the agreed recurrence or loss-rate threshold and adverse-effect triggers. Also review changes that make the original comparison invalid. Preserve the closure history, record the new evidence and assign the next investigation or verification step.