The setup is marked complete at 10:35, but the next acceptable product does not arrive until 10:50. Measure changeover time using a declared last-good to next-good production boundary, not only the maintenance or setup ticket. Retain adjustment, trial output and required quality acceptance inside that boundary so an earlier status change cannot hide the remaining loss.
Vorne’s SMED explanation measures the interval between the last good part at full speed and the first good part at full speed. Define what acceptable output and operating rate mean for your process before collecting timestamps. This workflow separates the elapsed transition from its labor and task detail, then tests which work can move outside the stopped period under approved procedures.
Declare the production boundary before comparing shifts
Record the machine or line, outgoing product, incoming product, tooling or recipe revision and transition identifier. State which event starts the clock and which evidence ends it. For a batch process, specify the equivalent accepted output event with the process owner. Do not substitute a machine-running signal for a quality result when the end definition requires both.
Keep the event time separate from the time someone entered it. If a quality result is entered later but identifies an earlier sample, preserve both timestamps and the approved clock rule. A missing last-good timestamp should remain an exception. Backfilling it from the scheduled stop can make the report appear complete while changing actual elapsed time into a planning assumption.
Retain cleaning, safety checks and required process controls under their actual procedures. A time-reduction project does not authorize bypassing an isolation step or accepting unverified output. Have operations, quality and the appropriate safety owner approve any changed work sequence before the revised method is used.
Build the timeline without adding overlapping tasks
Capture each element with a start, finish, owner and evidence source. Mark waiting for tooling, materials, settings or inspection as distinct intervals. Use a transition ID to link production counts, setup work and quality results. The production downtime review supplies the wider loss context; this analysis follows the specific product-to-product transition.
Elapsed machine time and labor effort need different totals. If two technicians work during the same 15-minute interval, the clock advances by 15 minutes while their combined effort may be 30 person-minutes. Preserve those units in the report. Adding every task duration gives the wrong elapsed total when activities overlap.
Map which task must finish before the next task or restart can occur. If a tooling wait overlaps a cleaning step that takes longer, removing the tooling wait may save no elapsed time. Keep task intervals and predecessor relationships together so an opportunity calculation follows the work that controls restart. Labor effort can fall without changing that controlling sequence.
For an illustrative transition, assume the last accepted product A occurs at 10:00, setup work finishes at 10:35, and product B meets the declared quality and operating-rate condition at 10:50. Elapsed changeover is 50 minutes. A setup-ticket report would show 35 minutes and omit the final 15 minutes of adjustment or acceptance. The example is a clock calculation, not an observed plant result.
| Illustrative event | Time | Interpretation |
|---|---|---|
| Last good A at the agreed rate | 10:00 | Start of the transition |
| Setup ticket complete | 10:35 | A task status inside the transition |
| Good B at the agreed rate | 10:50 | End of the transition |
| Elapsed changeover | 50 minutes | 10:50 minus 10:00 |
Retain startup rejects and material consumed during trials alongside the time result. A shorter clock accompanied by more rejected output may have shifted the loss into quality or material cost. Use the same acceptance rule when comparing the original and revised methods, including any required confirmation of the operating rate.
Separate stopped-machine work from preparation
The Lean Enterprise Institute’s SMED definition separates internal setup, which requires a stopped machine, from external work that can occur while it runs. SMED means single-minute exchange of die and refers to a single-digit-minute target. That name does not establish a suitable target or a feasible change sequence for every process.
Review the current method before designing a faster one. Label elements as internal or external under approved operating conditions, and retain the reason for each classification. Finding tools, checking a released instruction or staging materials may offer preparation opportunities. Tasks involving machine access, hazardous energy or process cleanliness need their own engineering and safety assessment.
Distinguish moving work from removing it. Preparing a fixture before the line stops can reduce elapsed stopped time while leaving the labor requirement unchanged. Record the preparation owner and due time, then verify that the work does not interfere with the current run. Without that resource check, the proposed improvement may create a queue elsewhere.
Declared start
Tasks, waits, checks
Declared acceptance
Reconstruct one transition from its source records
Bring approved production timestamps, setup activity and the associated quality evidence.
Test the revised method on comparable transitions
Compare the same outgoing and incoming product families, equipment and required controls. A color-to-color transition and a transition requiring a different cleaning procedure may have different work. Keep that sequence in the dataset and explain any grouping. An overall average can fall because the schedule contains easier transitions, even when no method improved.
Suppose the illustrative 50-minute transition includes eight minutes locating a fixture and seven minutes waiting for its approved instruction. Assume these intervals do not overlap and each delays restart (no parallel task controls either interval). If both can be prepared beforehand without affecting other work, the arithmetic opportunity is 15 minutes. Calling the resulting 35-minute scenario a verified saving would be premature. Measure repeated comparable transitions after the authorized change, retaining quality results and preparation effort.
Assign preparation to a real shift and resource before testing that scenario. Record when the fixture and released instruction became ready, then check the next transition against those records. If the previous run overran or the same technician was needed elsewhere, retain the failed preparation condition with the time result. Compare transitions where the revised method was followed with those where it was not, rather than attributing every short transition to the method change.
Review the result in the production schedule. Recovered time at the manufacturing bottleneck may affect flow differently from recovered time on equipment that is waiting for work. Keep capacity assumptions separate from actual additional good output. A time reduction does not prove extra sales, and it does not by itself justify reducing batch sizes without reviewing the remaining constraints.
Update the approved work instruction only after the owner accepts the evidence. Record what changed, its effective date and the training or resources required. Preserve the earlier method and its baseline. If results later worsen, reviewers need to distinguish method drift from a different product mix or a missing preparation step.
How Permute preserves the transition and its evidence
Permute is the connected-data platform we sell for analysis and custom business applications. We can combine approved SAP or NetSuite production records with setup and inspection extracts in Excel or Google Sheets. We reconcile transition, equipment and product identifiers while retaining the source records and extraction times.
We define the customer’s accepted clock boundary, sequence grouping and exception rules for repeated analysis. A production lead can ask which changeovers exceeded the expected method and inspect the waiting intervals or missing end evidence. The resulting analytical view can support recurring review without turning an inferred timestamp into an observed event.
Keep operating instructions and machine control outside the report
This workflow analyzes approved records and targets hourly or daily review where supported. It is not a machine-control or safety system, and it does not certify that a setup sequence can be changed. Missing event capture remains a measurement gap. Operations, engineering and quality owners approve work methods; supported external actions require their own permissions and approval path.
Accept a reduction only against the same production boundary
Changeover time should retain the full interval required to return to acceptable production under a declared rule. Separate elapsed time from labor effort and identify overlapping tasks before calculating an opportunity. Compare like transitions after a controlled method change, with startup quality and preparation still visible. Use the verified result in planning only after the owner accepts the evidence and its operating conditions.
Review the gaps in your changeover evidence
Trace one disputed transition from the last good run through the next accepted production.
Questions about setup and changeover measurement
Should the changeover clock equal the OEE downtime figure?
Keep both definitions visible and reconcile them. An OEE report may classify startup rejects, reduced speed and planned stops under different loss rules. The elapsed transition here follows the declared production boundary, so equal labels do not guarantee equal totals.
What happens when a changeover crosses a shift boundary?
Keep one transition identifier and preserve all timestamps across the handover. Show the elapsed interval and assign activities to their actual owners without restarting the clock. Apply any approved calendar exclusions as a separate, disclosed measure.
Can a machine-start signal replace a missing good-product event?
Only if the process owner has defined and verified that signal as sufficient evidence for the chosen endpoint. Otherwise mark the end event missing and retain the machine-start timestamp as partial evidence. An operating signal alone does not establish the quality of the output.