Changeover Matrix: Building One From Real Machine Data

Every schedule assumes changeover times someone guessed years ago. The machines have been recording the truth all along.

Written by George Munguia · Published Jul 19, 2026

Ask a scheduler how long the changeover from SKU A to SKU B takes and you will get a number, usually a round one, usually inherited. The schedule is built on a matrix of these guesses. Meanwhile the machines have been timestamping the truth for years: when the last good unit of A ran, when the first good unit of B followed. The gap between those timestamps is the real changeover, and it rarely matches the folklore.

Measure last-good to first-good

Wrench time undercounts. The full cost of a changeover includes the ramp-down, the setup, the startup scrap, and the tuning until quality holds, everything between the last saleable unit of the old run and the first saleable unit of the new. Captured from PLC counts and quality data (any machine brand, via OPC UA or whatever it speaks), this measures itself: no stopwatch studies, no observer effect, every changeover a data point forever.

What measured matrices reveal

Sequence effects dominate: A-to-B and B-to-A are rarely equal, dark-to-light costs more than light-to-dark, and some pairs hide a full washdown the guessed matrix never priced (allergen plants feel this hardest, see allergen scheduling).
Crew spread is real: the same change runs 40 minutes for one crew and 90 for another. That spread is not a discipline problem, it is a knowledge-transfer opportunity: find what the fast crew does and make it the standard.
Startup scrap is half the cost: plants that only tracked downtime discover the first-good-unit tail, tuning scrap, is comparable to the stop itself.

Feed it back to the schedule

The matrix's purpose is prospective: measured times replace guesses in the scheduler, and sequence optimization gets honest inputs (the propose-and-approve architecture is in our scheduling software guide). Plants usually find the old matrix was optimistic on the pairs that mattered, which explains years of schedules that looked feasible at 6 a.m. and slipped by noon. Contract packagers, where changeovers are the margin, treat the measured matrix as quoting data too: the real cost of a customer's job includes the transitions it forces (see contract packaging operations).

Keep it alive

A matrix built once decays like the guesses it replaced. Because measurement is automatic, the matrix should be a rolling view, recent months weighted, crews compared, drift flagged when a pair's time creeps. The changeover that got 15 minutes slower over a quarter is a maintenance signal wearing a scheduling costume.

Frequently Asked Questions

What is a changeover matrix?

A table of how long each product-to-product transition takes on each line, used by scheduling to sequence runs. Most plants populate it with estimates; measured versions capture last-good-unit to first-good-unit from machine data automatically.

How should changeover time be measured?

From the last good unit of the outgoing run to the first good unit of the incoming run, including ramp-down, setup, startup scrap, and tuning. PLC counts plus quality capture measure this automatically for every changeover, no stopwatch studies.

Why do changeover times vary so much between crews?

Different methods, sequences, and preparation. The spread, often 2x between fastest and slowest, is a transferable-knowledge signal: capture what the fast crew does, make it the standard, and the whole matrix improves.

How does the matrix improve scheduling?

Honest transition times make sequence optimization real: the scheduler can minimize total changeover cost, avoid the expensive pairs, and produce plans that hold past noon. Guessed matrices produce schedules that were never feasible.

The Operations Brief

The best of what we publish, once a week.

Practical guides on high-production manufacturing, straight from the floor. No spam, unsubscribe anytime.

You're in. The next brief lands Monday morning.

See Harmony on your line

Harmony connects your machines, systems, and paperwork into one real-time operational layer — no rip-and-replace. See what it looks like in a plant like yours: read the CLS case study.

Book a Demo →

← Back to Resources