Insights / Scheduling

Cutting Changeover Time by Resequencing Production

On many food and drink lines, the time a changeover takes depends on what ran before and what runs next. Going from a mild product to a strong one may need a quick flush; going the other way may need a full clean. Running the same products in a better order can therefore cut total changeover time without new equipment or faster changeover methods. This guide shows how to measure sequence-dependent changeovers, how to find a better order, and why the problem needs software once the range grows.

Sequence-dependent changeovers

A changeover time is sequence-dependent when it varies with the pair of products involved. Common patterns in food, drink and personal care:

  • light colours before dark ones
  • mild flavours before strong ones
  • allergen-free products before products containing an allergen, with a validated clean after
  • a full clean before organic, halal or kosher runs
  • pack size changes that need a tooling change, as opposed to a label change only

When changeovers behave like this, a single average changeover time is misleading. The same week’s production can lose very different amounts of time depending on the order it runs in.

Build a changeover matrix

The starting point is a from-to table: for each pair of products, how long the line is down when changing from the first to the second. Collect it from:

  • line logs or MES records of past changeovers, grouped by the products either side
  • cleaning and hygiene procedures, which set the minimum for allergen and flavour changes
  • operators and shift leaders, for pairs that have rarely or never run back to back

With a large range, group products into families that change over the same way, such as by flavour strength, allergen or pack format. The matrix then works at family level, with a smaller allowance for changes within a family.

A worked example

The figures are illustrative. A sauce line runs four products: plain tomato, garlic, chilli and a mustard sauce that contains a declared allergen. The table shows changeover times in minutes, from the product in the row to the product in the column.

From \ To Tomato Garlic Chilli Mustard
Tomato 20 25 30
Garlic 60 25 30
Chilli 70 45 35
Mustard 90 85 80

Each product runs once per cycle, and the cycle ends back on tomato for the next one. There are six possible orders:

Order Changeover time per cycle
Tomato, garlic, chilli, mustard 170 minutes
Tomato, chilli, garlic, mustard 190 minutes
Tomato, garlic, mustard, chilli 200 minutes
Tomato, chilli, mustard, garlic 205 minutes
Tomato, mustard, garlic, chilli 210 minutes
Tomato, mustard, chilli, garlic 215 minutes

The best order runs mildest to strongest with the allergen last, then takes one full clean back to tomato. It loses 45 minutes less per cycle than the worst order, a 21% difference, with the same products, the same volumes and the same changeover methods. On a line that runs the cycle ten times a week, that is seven and a half hours.

Why it gets hard quickly

Four products have six possible cycles, which can be checked by hand. Twelve products have 39,916,800. Twenty products have more than 1017. The number of orders grows factorially, so checking them all stops being possible very quickly.

The real problem is also larger than the order alone:

  • orders have due dates, so the best order for changeovers may deliver something late
  • products have shelf lives, which limit how far ahead each one can be made
  • batch processes upstream have minimum batch sizes and their own changeovers
  • tanks, cleaning sets and crews are shared with other lines
  • some products can only run on some lines

This is a scheduling optimisation problem. Planners handle it with rules of thumb such as light to dark, which get a long way. Optimisation software working on a model of the plant can search far more options and check each one against the real constraints, so the sequence it proposes can actually be run.

Sequencing and SMED

SMED (single-minute exchange of die), developed by Shigeo Shingo, shortens each changeover: doing preparation while the line is still running, standardising fittings, removing adjustments. Sequencing reduces how much changeover time the schedule needs in total. The two work together. SMED lowers the numbers in the matrix; sequencing chooses the cheapest path through it.

Sequencing is usually the cheaper of the two to start with, because it needs no equipment changes and no new procedures on the line. It does need the schedule to hold to the sequence once it is set.

Run length and range

Fewer, longer runs mean fewer changeovers, but they raise stock levels and push against shelf life. Many plants settle this with a fixed product wheel: a repeating cycle with a set order and a target run length for each product, reviewed as demand changes.

The range itself is the other lever. At a pickles and condiments plant we modelled, 12% of SKUs contributed just 3% of margin but consumed 25% of changeover time. Resequencing production cut total changeover time by 20%, and together with removing the low-margin SKUs it helped the plant move from 24/7 to 24/5 operation.

Results from our projects

  • Pickles and condiments: 20% less total changeover time from resequencing, with sequences that were physically achievable given tank capacities, minimum batch sizes and crews.
  • Vinegar: 56% less idle time and 80% of labour excess hours eliminated, from AI-optimised scheduling across milling, fermentation and bottling.
  • Biscuits: alternative product combinations for a rebuilt line compared on modelled throughput, changeover time and crewing before the mix was chosen.

Common questions

Can our ERP or MRP system do this?

MRP plans what to make and when materials are needed. It does not usually optimise the order of runs on a line. Advanced planning and scheduling (APS) software can, for the plant as it is. To test a different range, shift pattern or line allocation before changing anything, a model of the plant is the better tool.

How much changeover data do we need?

Enough to fill a family-level matrix. A few months of line logs, combined with the cleaning procedures and operator knowledge, is usually sufficient to start. The matrix can be refined as more changeovers are recorded.

What is a product wheel?

A fixed repeating production cycle: every product in a set order, with a target run length, repeated on a regular period such as weekly. It makes the schedule predictable and keeps the changeover sequence efficient.

Related guides

Test your sequence in a model