Changeover Time measures the elapsed window from the last good unit of a running product to the first good unit of the next product, including all retooling, calibration, and material reload steps on the same asset [S1][S2]. It sits inside the production efficiency KPI family alongside OEE, throughput, and cycle time, and is widely treated as a leading indicator of hidden capacity loss rather than a lagging accounting metric [S3][S5].
Plants that formally track changeover as a KPI rather than as informal "setup minutes" report 24% higher OEE and 18% better on-time delivery versus plants without structured KPI programs, per a User Solutions summary of Aberdeen Group research published on 2026-04-10 [S5]. On a single line, reducing changeover from 30 to 7 minutes is typical of Single Minute Exchange of Die (SMED) projects targeting sub-10-minute changeovers, with the 95%+ reduction figure appearing in the same 2026-04-10 KPI guide [S5].
Definition, Formula, and What "Last Good to First Good" Actually Means
The accepted formula is Changeover Time = Time of first good unit of next product minus Time of last good unit of previous product, measured on the same asset and the same shift [S1][S2]. This is deliberately tighter than "machine stopped to machine restarted" because it captures both the setup work and the stabilisation time before quality is back in spec [S5]. DataPARC's 2022 KPI reference classifies changeover as a process-data metric, sourced from sub-second to per-minute sensor streams rather than from manual logs, which is why it pairs naturally with high-frequency OEE and throughput KPIs on the same dashboard [S1].
The metric is asset-focused, not product-focused: a single changeover event retooling for a new SKU is one observation, regardless of how many units the line will run of that SKU [S2]. SimpleKPI's 2021-11-19 reference groups changeover time under machinery- and asset-focused KPIs alongside machine uptime, defects, and cycle rate, which matters because it means the KPI rolls up to the equipment, not to the part number [S2].
Target Ranges and the SMED Benchmark
The 2026-04-10 User Solutions 50-KPI guide sets the changeover target as "minimise; 95%+ reduction measured against original customer-defined baseline" and ties the KPI to SMED methodology developed for sub-10-minute die exchanges [S5]. In practice the same guide states world-class OEE is 85%+ while average manufacturers operate at 60%, and OEE is the composite that exposes changeover losses through the Availability factor (Run Time / Planned Production Time) [S5].
For the Availability factor specifically, a plant at 90% availability, 95% performance, and 99% quality reaches 84.6% OEE, just under the 85% world-class line, and most of the headroom on Availability in mid-range plants traces back to changeover and minor stops rather than breakdowns [S5]. PlanetTogether's 2026-02-11 APS KPI piece flags setup and changeover time as the "hidden capacity loss" lever that turns a 7:12 a.m. Monday reschedule from guesswork into a decision [S3].
Data Sources, Cadence, and Dashboard Wiring
Changeover data is most reliably captured as high-frequency process data (state changes, machine cycle signals, vision-system rejects) rather than operator-entered timestamps, and dataPARC's KPI reference explicitly lists changeover time among the metrics that benefit from sub-second to per-minute sampling [S1]. JODOO's 2026-04-15 dashboard guide adds that the same dashboard should also surface cycle time, downtime minutes, scrap pieces, and changeover duration, so the trade-off between setup loss and OEE gain is visible in one view [S4].
A working changeover KPI on a dashboard usually carries four fields: the from-SKU, the to-SKU, the asset ID, and the changeover duration in minutes, with rolling weekly and monthly averages plus a Pareto of changeover types by total minutes lost [S4][S6]. Thrive's 2026-09-23 dashboard reference repeats the same data-flow point: cycle time, downtime minutes, scrap pieces, and changeover duration are metrics; the changeover KPI is the metric tied to a reduction target and a SMED project owner [S6].
Who Should Track It, and Where It Misleads
Changeover time is the right KPI for high-mix, lower-volume environments (job shops, printed-circuit assembly, contract packaging, automotive Tier-1 changeovers) where every product switch costs a measurable slice of the day, and where the trade-off is setup time versus batch size [S2][S3]. It is a poor primary KPI for a continuous process plant running one grade for weeks at a time; there, the changeover event is rare enough that the sample size per month is too small to drive a statistical process control chart, and OEE on availability is the better signal [S3].
It also misleads when reported as a single weekly average: a 9-minute average can hide a 4-minute best case and a 38-minute worst case tied to one specific changeover type. PlanetTogether's APS framing pushes for per-event reporting and for using APS scenario comparison to weigh "more setups, smaller batches" against "fewer setups, larger batches" against the same demand plan [S3]. Silvon's 40-KPI list treats changeover time as lost minutes, not as a percentage, and recommends a Pareto by changeover type to target the worst 20% of events first [S7].
Comparison: Changeover Time Against Related KPIs
Changeover Time, Machine Downtime, Cycle Time, and OEE each measure different aspects of production using different data sources and units, so a manufacturing KPI dashboard should keep them separate: Changeover Time represents the time lost from switching a production line from one product to another, Machine Downtime tracks downtime minutes, Cycle Time measures per-unit timing, and OEE is a composite score that manufacturers with formal KPI programs use to benchmark performance [S1][S4][S5][S6][S7]. A planner reading only OEE cannot tell whether a 6-point drop came from a slow changeover, a speed loss, or a quality spike, which is why changeover time is run as a standalone metric with its own target [S4].
The OEE in Continuous Process Industries: Yes, With Specific Adaptations reference makes the same point for process plants: OEE is meaningful, but changeover as a discrete event has to be re-defined as a grade-transition event before it can be tracked the same way. For high-mix discrete lines, the Plunger Pump Operating Boundaries: Pressure and Speed Envelope article shows the parallel pattern in maintenance, where each operating boundary is a tracked event with a duration, not a monthly average.
Failure Modes, Common Mistakes, and Verification
Three failure modes show up in every published KPI list: (1) measuring stopped-to-started rather than last-good-to-first-good, which understates changeover by 5-15 minutes on lines that need warm-up or first-piece inspection [S1][S5]; (2) aggregating across assets, which hides a single problem cell inside a healthy average [S4]; and (3) letting the changeover KPI live in a spreadsheet updated at end of shift, which means a 30-minute blind spot on a packaging line costs missed output, premium freight, and overtime by the time anyone sees the number [S4].
Verification of the data path should be done quarterly: pull 10 random changeover events from the dashboard, walk to the asset, and reconcile the dashboard timestamp against the machine's own state log and the operator's shift handover, since Silvon notes that "reducing changeover" is one of the most misreported numbers on plant dashboards [S7]. For shop-floor equipment tied to construction machinery and equipment or lamps and light fittings assembly lines, the same audit pattern applies, with the added step of checking that the changeover clock pauses during operator breaks if the plant policy is to exclude them.
Standards, Sourcing, and What to Anchor Targets To
There is no ISO or IEC standard that names a specific changeover-time target value; SMED is a methodology from the lean manufacturing literature, and the 95%+ reduction figure cited by User Solutions is a benchmark drawn from project practice, not a regulatory threshold [S5]. What the published KPI guides do anchor to is the SMED definition (last good unit to first good unit) and the OEE definitions (Availability = Run Time / Planned Production Time, Performance = Actual Output / Theoretical Maximum, Quality = Good Units / Total Units) which together define how changeover losses should be quantified [S5].
For a reader building the KPI from scratch, the minimum viable data set per event is: asset ID, from-SKU, to-SKU, last-good-unit timestamp, first-good-unit timestamp, and a reason code from a fixed list (tooling change, material reload, calibration, cleaning, change of program) [S1][S5]. For comparison, the pressure transmitter and time relay reference pages show the same pattern: a short, fixed taxonomy of failure modes is what makes a metric auditable rather than a spreadsheet of free text.
Next signal to watch: whether a plant's changeover KPI is run per-event with a reason-code Pareto or only as a weekly average, because that single design choice decides whether the SMED project gets a hit-list or a guess. A second trackable signal is the ratio of changeover minutes to total available minutes per shift, which user-side 2026 dashboards increasingly publish alongside OEE availability so setup loss and breakdown loss are visible in the same view [S4][S6].