KPI Library / Production

Changeover Time (SMED)

Formula: Time from last good unit of the previous run to first good unit of the next run
Typical range: Many plants start at 45 to 90 minutes; SMED programs typically target under 10 minutes for the internal (machine-stopped) portion

Changeover time is the elapsed time between the last good unit of one production run and the first good unit of the next, covering teardown, setup, and any adjustment runs in between. SMED (Single-Minute Exchange of Die) is the discipline built around shrinking it, largely by converting steps that require the machine to be stopped (internal) into steps that can happen while it’s still running (external).

What good looks like

Most plants that haven’t run a formal SMED program sit somewhere between 45 and 90 minutes for a typical changeover. The first pass of a SMED event usually cuts that by 30 to 50% just by separating internal from external work and pre-staging tools and materials, no capital spend required. Mature programs on high-changeover lines push the internal (machine-stopped) portion under 10 minutes, which is where the “single-minute” name comes from.

The trap is where the clock stops. If changeover is logged as complete the moment the line restarts, but the first several minutes of output are scrap while the process stabilizes, the reported changeover time looks better than the actual disruption. Define the end of changeover as first good part, the same rule that should apply to MTTR, so the metric can’t hide behind a fast-but-scrappy restart.

Changeover Time in Power BI (DAX)

With a changeover events fact carrying start and end timestamps by line and part:

Changeover Minutes =
SUMX (
    fact_changeover_events,
    DATEDIFF (
        fact_changeover_events[changeover_start],
        fact_changeover_events[first_good_part_time],
        MINUTE
    )
)

Changeover Count =
COUNTROWS ( fact_changeover_events )

Avg Changeover Minutes =
DIVIDE ( [Changeover Minutes], [Changeover Count] )

Segment Avg Changeover Minutes by part-to-part pair, not just by line. Some changeovers are inherently harder than others, and blending them into one average hides which specific transitions are worth a SMED event.

Common mistakes

1. Ending the clock at line restart instead of first good part. This is the most common way changeover time gets understated.
2. Treating all changeovers as equally difficult. A color change and a full tooling swap don’t belong in the same average without at least a category breakdown.
3. Running a SMED event once and never re-measuring. Gains erode as new operators join and old habits creep back in; changeover time needs the same ongoing tracking as any other KPI, not a one-time project close-out number.