A→B costs 30 minutes. B→A costs 55. Same two SKUs, same line — but most schedules price them identically anyway. That one wrong assumption is where a chunk of "unavoidable" changeover loss actually comes from. iFactory, an on-premise AI platform for production scheduling and OEE analytics, times every changeover pair straight from machine signals, so the schedule finally knows what a transition really costs instead of guessing. Bring your own changeover data to a live review.
Your Schedule Doesn't Know A→B Isn't The Same As B→A
Direction changes the cost. Pair changes the cost. Duration is only one piece of a bigger puzzle — and most plants are solving for the wrong piece.
Duration Was Never The Whole Story
Ask a planner how long a changeover takes and you'll get one number. Ask which direction, which pair, which crew — and the number falls apart.
That's the changeover matrix: every SKU, against every other SKU, with its own measured minutes. Shared tooling makes one direction cheap. A deep clean makes the reverse expensive. Light-to-dark is quick; dark-to-light drags. Same two products, two completely different bills — and a planner who only tracks "average changeover time" never sees the split at all.
Line up a week's worth of SKUs and there isn't one possible order to run them in — there are dozens, sometimes hundreds. Each path through that mix adds up to a different total changeover bill. Most plants pick a path by due date alone and never check what it cost against the cheapest available one.
Sequencing Fixes The Plan. SMED Fixes The Machine.
Confusing these two is how improvement budgets go to the wrong place first.
| Lever | What moves | Typical gain |
|---|---|---|
| Sequencing | Which changeovers happen at all | 20–40%, zero capital |
| SMED | How long each one takes | 30–60%, process work |
| Both | Count and duration together | Gains stack, don't overlap |
Sequence first. It's a scheduling rule, not a shop-floor project — nothing to install, nothing to retool, just a smarter order. Then point SMED at whichever specific pairs the matrix still flags as expensive once the easy order is already in place. Run them the other way around and you'll spend an engineering budget shaving minutes off a transition the schedule shouldn't have created that often in the first place.
How Much Of Your Loss Did Your Schedule Create?
Send this week's SKU mix and recent changeover history — the team will show you the gap between what's running now and the cheapest sequence available.
Your Shift Report Confirms A Number. It Doesn't Measure One.
A written standard says 25 minutes. The report says 25 minutes. Nobody actually timed it.
- Spread by pair → sequencing problem. Some transitions are just expensive, and the schedule keeps choosing them anyway.
- Spread by crew → method problem. Same job, 20–50% longer with one crew than another — the fast crew's trick was never written down.
- Missing ramp-up → hidden problem. Start-up stumbles after restart often cost nearly as much as the changeover itself, and almost nobody counts them.
Three different root causes, three different fixes — and a single average "changeover time" number flattens all of them into one meaningless figure that tells nobody what to do next.
25 Minutes On Paper. 26 To 52 On The Clock.
One mixed-SKU line. One written standard: 25 minutes, every pair, confirmed every shift. Four weeks of real measurement later: a spread from 26 to 52 minutes depending on who ran it and what followed what. The fast changeover was real — it happened almost daily. It just wasn't happening on purpose. Format change and start-up carried almost the entire gap between the fast crew and the slow one; every other stage barely moved between them. Nobody saw any of it until the clock did the measuring instead of the sheet.
That's the pattern worth hunting for on any line: a standard that nobody has disproven simply because nobody has actually timed it against the sheet that claims to confirm it.
Four Ways A Changeover Matrix Goes Wrong
- Guessed, not measured — memory is wrong about anything that doesn't happen weekly.
- Treated as symmetric — A→B and B→A are not the same number. Averaging them erases the exact thing sequencing needs.
- Never refreshed — a matrix from before the last tooling change is a matrix describing a line that no longer exists.
- Used without constraints — the cheapest sequence on paper can still break a due date, a quality rule, or a material delivery. Math first, reality check second.
Get any one of these four wrong and the matrix stops being a planning tool — it becomes a confident-looking spreadsheet that quietly points the schedule in the wrong direction every single week.
The Matrix Builds Itself. The Schedule Uses It.
No stopwatch. No form. iFactory reads the changeover straight off machine and PLC signals — first good unit to last, every stage in between — and turns it into a live matrix the scheduler can actually use.
- Every changeover timed automatically, start to finish.
- Direction tracked separately — A→B and B→A never get blended into one misleading average.
- Scheduling respects due dates, quality separation, and material limits — not sequence math in a vacuum.
- Drift flagged by pair and by crew, so a sequencing problem and a method problem never get confused for each other.
On-premise means the data never leaves the plant to get this smart — every signal the schedule runs on stays inside the four walls that generated it.
A Bad Sequence Doesn't Just Cost Minutes — It Costs Options
Every minute lost to an avoidable changeover is a minute the schedule can't spend anywhere else.
Run the expensive direction three times a week instead of once, and the plant hasn't just burned extra changeover time — it's quietly shrunk its own flexibility. Fewer SKU switches fit in a shift. Rush orders get pushed further out. A planner trying to squeeze in one more run discovers there's no slack left, and the schedule itself is the reason why.
Fix the sequence and that slack comes back without touching a single machine. The same shift that used to fit four transitions can often fit five or six — not because anything runs faster, but because fewer of the transitions it runs are the expensive ones.
What A Scheduling Lead Checks First
A changeover program is worth trusting when it changes next week's sequence — not when it produces a report agreeing with the standard everyone already believed.
What Planners Ask About Changeover Scheduling
No. SMED makes one changeover faster. Sequencing makes fewer expensive changeovers happen in the first place. Book a session and see both measured on your own line.
Shared tooling one way, a deep clean the other; a color or allergen shift that's cheap forward and expensive in reverse. See your own matrix built from real data, not assumptions.
Every time SKUs, tooling, or process change — and every six to twelve months regardless. A stale matrix quietly misleads every schedule built on it. Ask support about refresh cadence.
Easily. The math doesn't know about customer commitments, quality separation, or material delivery — so the optimal sequence always gets filtered against those first. Walk through your constraints with the team.
Last good unit to first good unit — ramp-up included. Most shift reports stop the clock the moment the machine restarts, which is exactly where the hidden cost begins. See it measured properly on a live line.
See Your Own Matrix, Built From Real Data
Book a session with the iFactory team and find out how much of your changeover loss your schedule is creating on its own.
