Overfill is giveaway: reading the weight curve against the declared weight

The cost nobody invoices

A filler set to land a 250 g bag at 254 g is doing its job: almost nothing goes out short, and nobody gets a call from an inspector. It is also giving away 4 g on every bag, 1.6% of the product, and nobody sends an invoice for it.

Run the numbers on one line. At 20,000 bags a day, 4 g a bag is 80 kg of product a day. Over 250 production days that is 20 tonnes. At $10 a kilogram of finished product, the line ships $200,000 a year of cheese, chips or cream that no customer paid for.

Why the target sits above the declared weight

No filler puts exactly 250 g in every bag. The weights spread around the target, and the wider the spread, the further above the declared weight the target has to sit to keep the light ones from going out short.

Most net quantity rules, Canada’s included, judge a lot two ways: the average of a sample against the declared quantity, and individual packages against a tolerance below it. The exact tolerances depend on the product and the regulation that applies to it, so confirm them with whoever answers for labelling at your plant. What does not change is the arithmetic: a plant that wants almost no package below the tolerance sets its target roughly two to three standard deviations above the declared weight.

• With a standard deviation of 2 g, the target lands around 254 g for a 250 g bag.
• Bring the standard deviation down to 1 g and the same safety margin puts the target near 252 g.
• That is 2 g a bag, 40 kg a day on the line above, recovered without shipping a single short bag.

This is why the spread matters more than the average. Moving the average down without narrowing the spread only trades giveaway for short packages. Narrowing the spread is what earns the right to move the target.

Reading the weight curve

A weight curve plots every weighed sample of an order over time, against the declared weight and the limits. Read as a curve rather than a column of numbers on a paper check sheet, it answers the questions a supervisor actually has:

• Is the average where we set it? A curve that sits 6 g above the declared weight when the target is 254 g is a filler running hot, not a filler doing its job.
• Is the spread stable? A band that widens after lunch points at a hopper level, a product temperature or a worn part, not at the operator.
• Is it drifting? A slow climb across a shift is the classic sign of a filler compensating for something upstream. Caught at the third sample, it costs a few kilograms; caught at the yield report, it costs a week.
• Which machine? Two fillers on the same product with different spreads deserve different targets, and a Pareto by machine says which one to fix first.

Bringing the target down, in order

1. 1Weigh on a schedule, not when someone remembers. A fixed sample of bags at a fixed frequency is what turns weights into a curve.
2. 2Judge every sample against both limits the moment it is weighed: the minimum for each package, and the average against the declared weight.
3. 3Treat an out-of-spec weight as a non-conformity, with the product on hold, rather than a number to note and move past.
4. 4Fix the spread before touching the target: worn seals, hopper levels, product temperature, the filler’s own settings.
5. 5Only then, with a few weeks of narrower curves behind you, bring the target down, and keep watching the curve for the first lots that run on it.

How totalMOS handles it

In totalMOS, a weight check is a test like any other, with its own sample size, frequency and limits, including an average limit against the declared weight. Each reading is judged against the product’s spec when it is entered on the iPad at the line, an out-of-spec weight opens a non-conformity before the check can close, and every order draws its own weight curve against its declared weight on the web, next to the Pareto by machine. The dairy demo plant shows the weight checks it runs.