The company
Kettle & Co Foods makes sauces and condiments in the East Midlands: ketchup, brown sauce, barbecue, sweet chilli, mayonnaise and a wholegrain mustard. Most of it goes to two supermarket groups under their own labels, the rest under Kettle & Co's name. Turnover is about £18 million a year and the business makes an operating profit of roughly 6%, or £1.08 million. Food margins are thin, and everyone there knows it.
Everything goes through one bottling line with five stations: Mixing, Filling, Capping, Labelling and Case packing. Two shifts a day, Monday to Friday, one operator at each station. Sixteen people work on the line, plus agency staff when the rota runs short.
A week that didn't add up
Ask Kettle & Co's operations manager what the problem was and she'd have said "the filler". In peak weeks the line needed 360,000 bottles, and the filler could only make them with a Saturday shift. That happened for about 20 weeks of every year, from the barbecue season through to Christmas.
The odd thing was that the filler's rated speed was fine. At 5,400 bottles an hour, 360,000 bottles is under 67 hours of filling, and the line had 75 hours a week. On paper there were eight hours to spare. In practice there were none.
The planner, who had run the schedule for eleven years, built each week in a spreadsheet: customer orders in due-date order, then moved around by hand to keep mayonnaise and mustard runs apart from the rest. It worked. It also took most of Monday, and nobody could say whether a different order would have been cheaper, because nobody had time to try.
Finding the bottleneck
The first thing ModelWeave reported was the bottleneck station: Filling, in every week but one. That matched what the team felt, but it now came with a reason. Filling had the least spare time of any station once its changeovers were counted. Capping, Labelling and Case packing all had hours to spare.
This is the core of the Theory of Constraints. A line can only go as fast as its slowest station, so the bottleneck is where time matters most. An hour lost at the filler is an hour of output lost for the whole line. An hour saved at the labeller is welcome, but it doesn't make one extra bottle, because the labeller was already waiting for the filler.
The report said exactly that about the labeller's reel changes: the new plan saved some changeover time there, and that saving was money, not output. The one exception was a week with a planned day of maintenance on the capper. For that week ModelWeave named Capping as the bottleneck station, and the plan worked around it.
Where the overtime came from
Takt time is the pace the customer sets. You take the working time you have and divide it by the demand:
Seconds of filler time per bottle in a peak week, changeovers included. The black line is takt time, 0.75 s. Illustrative.
75 hours is 270,000 seconds. Divided by 360,000 bottles, that gives a takt time of 0.75 seconds a bottle. To keep up, the filler has to average one bottle every 0.75 seconds across the whole week.
Running flat out, it manages one every 0.67 seconds. But the filler doesn't run flat out all week. Kettle & Co's planned week had 19 changeovers at the filler: 12 quick flavour changes of 25 minutes and 7 full allergen cleans of 90 minutes. That's 930 minutes, or 15.5 hours, with the filler stopped. Spread those hours over the bottles and the filler needs 0.82 seconds each.
0.82 is slower than 0.75. The difference, about seven hours a week, is the Saturday shift. Nobody had been doing anything wrong. The overtime was simply what the run order cost.
Batch scheduling and cleaning cycles
Schedule optimisation looked for a cheaper run order for the same orders. Kettle & Co's changeover file told it what each switch takes: light sauces to dark ones need a short rinse, dark to light needs longer, and anything that follows mayonnaise or mustard needs a full allergen clean. The shift file gave each station its hours, and the orders file carried the planner's own sequence, so the comparison was with the plan Kettle & Co would actually have run.
The cheapest plan it found grouped the allergen products together and ran each block from light to dark. Fewer orders went out late, too: about 2% of orders instead of 6%, because the filler had more of its week to give.
| Changeover | As planned | New run order |
|---|---|---|
| Flavour changes, 25 minutes each | 12 | 8 |
| Full allergen cleans, 90 minutes each | 7 | 3 |
| Filler stopped for changeovers | 15.5 hours | 7.8 hours |
That's 7.7 hours a week handed back to the bottleneck. In a peak week the filler now needs 74.5 hours for 360,000 bottles, just inside the 75 it has. It is tight, and the report said so.
Four fewer full cleans a week also mattered outside the production office. Each clean-in-place cycle at Kettle & Co uses about 4,000 litres of hot water, the cleaning chemicals, and roughly 350 kWh of gas to heat it all. Every changeover also flushes product down the drain: about 60 kg on a flavour change and 150 kg on a full clean, once the line is purged and the first bottles are rejected.
Where these come from. ModelWeave counts and costs changeovers at the filler; it doesn't model water, energy or product waste. Kettle & Co's own team multiplied the changeover counts in the ModelWeave report by their own figures per clean, over a 48-week year. All of it is illustrative.
On a typical gas emission factor, that energy is around 14 tonnes of CO₂e a year. Kettle & Co's customers ask for that figure now, so the sustainability lead was as interested as the finance director.
Stock between stations, and shelf life
Before, Mixing started batches as early as it could, "to keep the filler fed". The result was holding tanks full of mixed sauce and pallets of filled bottles waiting for the labeller. On average there were two and a half days of production sitting between stations.
Inventory buffer optimisation set a stock level for each station within the range Kettle & Co gave it. In front of the filler the minimum was one full holding tank, so the bottleneck is never left waiting for sauce. It then released work to Mixing at the filler's pace, using a release list the planner can follow. In Theory of Constraints language that's drum-buffer-rope: the bottleneck sets the beat, a buffer protects it, and the release list is the rope.
Stock between stations fell from 2.5 days to 1 day. At an average cost of 45p a bottle, that freed £48,600 of cash once. Holding that stock had been costing about 10% a year in space and finance, or £4,860.
Shelf life was the other benefit. Mixed sauce at Kettle & Co has to be filled within 48 hours. Under the old way of working, batches sometimes came close to that limit, and two had been scrapped the previous spring. With work released at the filler's pace, sauce waited hours rather than days. Finished pallets left about a day and a half sooner, with more shelf life left on them for the supermarkets' checks on delivery.
The roster
Workforce optimisation built the rota from a workers file: an operator code for each person (no names), their skills, the shifts they can work, contracted and maximum hours, hourly rate and overtime premium. Agency workers went in the same file with their agency rate. It covered every shift the plan needed, kept within UK rest rules, and costed labour and overtime for each worker each week.
With 7.7 filler hours back, most peak weeks no longer needed the Saturday shift. Saturday overtime fell from 20 shifts a year to 6, in the heaviest weeks only. The roster also used Kettle & Co's own trained operators more evenly, so agency hours fell from about 24 a week to 6.
Why they trusted the numbers
The planner's first question was a fair one: how do you know this order is better? Thirty orders a week can be run in about 2.65 × 1032 different orders. Nobody checks that by hand, and a tool that sounds confident is no help if it can't show its working.
ModelWeave doesn't guess. Kettle & Co's rules, such as allergen cleans, shift hours and material arrival dates, become constraints that a plan must keep. Costs become a sum, and ModelWeave searches for the plan that makes that sum smallest. Then a separate check goes back over the plan and re-verifies every rule before anyone sees it.
Each statement in the report says how it is known. The capacity figures were proven, by arithmetic from the line's own hours. The breakdown test was measured: both plans were run through many simulated weeks with random breakdowns, and the new plan kept every order on time in 71% of them against 38% for the old one. Where an alternative was tested, such as putting all the allergen products at the start of the week, the report named the rule that ruled it out. Blanks in the files were filled with ModelWeave defaults, and the report listed them.
The report also doesn't claim the plan is the best possible. It calls it "the cheapest schedule found", because the search stops at a time limit. That honesty is what won the planner round.
"It didn't tell me anything I couldn't check. Every saving had the sum written next to it."
Kettle & Co's planner (illustrative)
What it did to the profit and loss account
Operations people talk in hours and changeovers. The board talks in pounds. Here is the year in the board's terms, with revenue held at £18 million so that none of the gain depends on selling more.
| Cost | Before | After | Saving | Figure from |
|---|---|---|---|---|
| Late-delivery penalties | £52,000 | £17,000 | £35,000 | ModelWeave's late orders, at the supermarkets' penalty rates |
| Product lost at changeovers | £76,464 | £40,176 | £36,288 | Changeover counts from ModelWeave × the team's kg per changeover × 90p/kg |
| Cleaning water, chemicals and energy | £29,976 | £14,904 | £15,072 | Changeover counts from ModelWeave × £63.50 a full clean, £15 a flavour change |
| Saturday overtime | £24,750 | £7,425 | £17,325 | ModelWeave's roster, from the workers file |
| Agency hours | £24,192 | £6,048 | £18,144 | ModelWeave's roster, at £21 an hour |
| Cost of holding stock | £8,100 | £3,240 | £4,860 | ModelWeave's stock levels, at 10% a year |
| Total | £215,482 | £88,793 | £126,689 |
48 working weeks a year; Saturday shifts are ten hours for five operators at time and a half on £16.50 an hour.
Every pound in that table is a cost that stops, so it falls straight through to operating profit:
That last figure is the one that landed in the boardroom. Kettle & Co would have had to win a contract worth more than £2 million to make the same extra profit, with all the cost and risk that brings. Here it came from running the existing line in a better order.
Two things are deliberately left out of the table. The £48,600 of cash released from stock is real money, but it's a one-off improvement in working capital, so it belongs on the balance sheet and in the cash flow, not in yearly profit. And outside peak season the filler now has 7.7 spare hours a week, enough for around 41,000 more bottles. If sales can fill them, that is extra profit on top. If not, they are still hours of protection against a bad week.
What it didn't do
It didn't change the recipes, the line or the people. Nobody worked faster. Nothing was bought. The six heaviest weeks of the year still need a Saturday, because no run order makes 75 hours stretch further than that.
It also only knows what it is told. If a changeover time in the file is wrong, the plan is built on the wrong number. That is why the report lists every default and every assumption it used, so the planner can correct them and run it again.
To be clear: Kettle & Co Foods is an illustrative example, not a real company, and every figure above is illustrative. The aim is to show how the parts connect, from a run order to a cleaning cycle to a line in the accounts. The quickest way to see what your own line could do is to try ModelWeave on our sample data, then on your own files.