Changeovers: Shorter Setups and Smarter Sequences

On most food and beverage lines, changeovers are one of the largest planned losses. Every product switch means stopping, cleaning, adjusting, restarting and waiting for the line to settle. With more SKUs and shorter runs every year, the number of changeovers keeps going up.

Most improvement work focuses on the mechanical side: faster tooling swaps, better kits, pre-staged parts. That matters. But there are really two levers. One is how long each changeover takes. The other is how often changeovers happen and in what order. Production data helps with both, and the second one is often where the bigger gains are.

What a changeover really costs

On the schedule, a changeover is usually a fixed block of time. Forty-five minutes between product A and product B.

On the line, it rarely works that way. The real changeover starts with the last good unit of product A and ends when product B is running at full speed with rejects back to normal. That includes teardown, cleaning, setup and start-up, but it also includes a recovery period that usually doesn’t appear on any schedule: the first half hour of running at reduced speed, with extra stops, adjustments and more rejects than usual.

When that recovery tail is included, a 45-minute changeover often turns out to cost 70 or 80 minutes of effective production. That’s the number that matters, and it’s the one most plants don’t measure.

Measure the whole changeover

Machine data makes it possible to measure changeovers from start to finish without relying on manual logs. The PLC already records when the line stops, when the product code changes, when cleaning sequences run, when the first unit of the new product is made, and how speed, stops and rejects develop afterward.

From those signals, each changeover can be broken into phases: rundown, cleaning, setup, start-up and recovery to stable production. Once every changeover is measured the same way, they can be compared directly.

Variation is the first clue

The most useful thing that comes out of this is usually not the average. It’s the spread.

It’s common to find that the same changeover, between the same two products on the same line, takes 50 minutes one day and 110 the next. That range is the opportunity. If the fast ones are possible, the question is what makes the slow ones slow.

Comparing them phase by phase usually points to a few recurring causes. Waiting is a big one: waiting for a quality release, for materials, for a mechanic or for the next shift to arrive. Trial and error at start-up is another, where settings get adjusted several times before the product is in spec. And sometimes cleaning runs longer than needed because it’s based on a fixed time rather than on the actual condition of the equipment.

An example: the start-up that kept changing

Here’s an example drawn from a few sauce and dressing lines. Changeovers between products, including an allergen clean, had a median of about 70 minutes, with a range from 50 to 120.

Breaking them into phases showed that cleaning time was fairly consistent. The variation was almost entirely in start-up and recovery. On the slow changeovers, the filler stopped several times in the first 30 minutes while fill weights and product temperature were adjusted. On the fast ones, the line came up close to target almost immediately.

The difference came down to where the settings started. On fast changeovers, the crew started from the settings of the last good run of that same product. On slow ones, the line started from wherever the previous product had left it, or from generic recipe values, and worked its way in.

Making the last good settings for each product the default starting point, and adding a check on quality release timing so swab results came back before the line was ready rather than after, brought the median down by about 20 minutes. Across several changeovers a day, that’s a significant amount of capacity, without touching the mechanical work at all.

The order of products matters as much as the speed

Not all changeovers are equal. Every plant knows some version of this. Going from light to dark, from non-allergen to allergen or from low to high viscosity is usually easier than the reverse. Switching between two formats that share parts is faster than switching between two that don’t.

These rules of thumb are usually right, but they’re rarely quantified. A year of production data changes that. Every product-to-product transition the line has made is in the history, along with how long it actually took and how long the recovery lasted. Put together, that becomes a changeover matrix: for each pair of products, the typical changeover time, how much it varies and how often it happens.

That matrix often holds a few surprises. Some transitions that everyone assumed were easy turn out to be consistently slow. Some that were considered difficult are quick, because the crew has done them so often. And a handful of transitions usually account for a large share of total changeover time, simply because they come up so frequently.

Using the data to schedule better

Once the real cost of each transition is known, scheduling becomes a much more concrete exercise.

The most direct gain is sequencing. With the same products and the same number of changeovers in a week, the order in which they run can change total changeover time considerably. Grouping products by allergen, color, viscosity or format, and running them in an order that follows the easier direction, reduces both the changeover time and the cleaning required.

Timing matters too. When changeovers are compared by when they happen, patterns often emerge. Changeovers that overlap a shift change tend to take longer, because the work gets handed over halfway through. Changeovers right after a weekend sanitation or on Monday mornings are often slower as the line warms up. Some crews are noticeably faster at certain changeovers than others, usually because they do them more often. Scheduling the most difficult transitions away from shift changes, and when the most experienced crew is on, is a simple change with a measurable effect.

In one case, applying the changeover matrix to a typical week with about 14 changeovers reduced total changeover time by roughly a quarter. The products were the same and so was the number of changeovers. Only the sequence and timing changed, mostly by reordering a few allergen and color transitions and moving two changeovers away from shift handovers.

Fewer changeovers, or better ones?

There’s a natural tension here. Fewer changeovers mean longer runs, which mean more inventory and sometimes shorter shelf life at the customer. Most plants can’t simply cut the number of product switches, because the market is asking for more variety, not less.

That makes predictability as valuable as speed. A changeover that reliably takes 55 minutes is often more useful to a planner than one that averages 45 but sometimes takes 100. When changeovers become predictable, schedules hold, overtime drops and planners can commit to tighter sequences with confidence.