The Cheapest Product You Ship Is the Free Ounce

The line that reported more than 100 percent

A lunchmeat co-packer had a year of clean production data and a digital twin that matched the real floor to 98.2 percent. That is a tight model. It ran five scenarios, from a do-nothing baseline to a full reconfiguration, and it found the real ceiling: cook capacity caps the plant at about 18 million pounds a year, three times current volume, and everything downstream of that is negotiable.

One line spoiled the picture. It reported an OEE above 100 percent, which is physically impossible, and it had been doing so for seven weeks. A broken feed. The number everyone on the floor had been trusting was lying, and it was lying in the flattering direction. Nobody caught it because nobody had a reason to distrust a number that made the line look good.

The expensive version is the number that looks fine

An OEE over 100 percent is the cheap version of this failure. It is absurd on its face, so eventually someone catches it. The expensive version is the number that looks completely reasonable and is still walking money out the door. That number is giveaway.

Fill weight is a target, and the target almost always sits above the label minimum. Operators pad it on purpose, because underfilling a labeled net weight is a legal and customer problem, and the person who runs the scale would rather be a little heavy than a little light. So the target drifts up, and it stays up, because nothing on the floor punishes it. Every package ships two percent, three percent, sometimes five percent more product than the label promises. That extra product is not free to make. It costs raw material, cook time, and labor, and then you give it away.

Here is what makes it invisible: giveaway ships. It never lands in the scrap bin. Scrap is honest; it accumulates in a container someone has to empty, and it posts as a yield loss on a report. Giveaway leaves the building as a sellable case of product, indistinguishable from a case filled to target. The loss is real, it is systematic, and it compounds across every unit, but it never generates a single data point that says the word loss. This is the same shape as the broken OEE feed. It is a number that does not self-report, except giveaway is worse, because the number that would flag it was never wrong. It was never taken.

The same logic runs through cutting, portioning, and forming. A portioning step that averages a gram or two heavy per piece feels like nothing at the piece level and turns into real tonnage across a year of volume. None of it shows up as scrap. All of it ships.

How to find the ounce you are giving away

Start with a scale and ten packages. Pull them at random off the end of a running line, weigh each one, and subtract the labeled net weight. If the average overage is more than a point or two, you have found it. Then look at the shape of the distribution on the checkweigher, not just the pass rate. A line running at target has weights centered on the target with a tight spread. A line funding a giveaway has a distribution shoved to the heavy side, with a long right tail nobody prices.

Fixing it is a control problem, not a willpower problem. You cannot tell operators to run leaner and expect it to hold, because the incentive that pushed the target up is still there. You tighten the target toward the minimum and you buy back the safety margin with better control: a checkweigher tied to feedback, a portioning or loading step with less variation, the auto loaders that a line-optimization model tends to justify anyway. The margin you were paying in free product becomes the margin you now hold in tighter tolerance.

This is exactly where a capacity-constrained plant should look first. A chicken sausage maker doubling its plant, with meatball lines already at capacity running two shifts seven days, does not have a spare ounce anywhere. When you are that tight, every ounce of giveaway is also an ounce of throughput you could have sold to someone else. The giveaway and the capacity ceiling are the same wall, seen from two sides.

And it is why a digital twin matters here beyond the throughput headline. A model that matches the floor to 98 percent is precisely the instrument that surfaces giveaway, because giveaway is invisible to the P&L and visible to the model. The floor cannot feel two percent. The model can.

What a floor without a giveaway looks like

Average net weight sits within one to two percent of target, and the target sits just above the label minimum, not a comfortable buffer above it. The checkweigher distribution is centered, not skewed high, and the right tail is short. Somebody owns the number in the gap between QA, whose job is to never ship light, and cost accounting, whose job is to never ship heavy, and that person reconciles the two every week rather than letting the fear of underfill quietly win by default.

The scrap bin tells you what you lost. Giveaway is the loss that walks out the door looking exactly like a sale, so the only way you ever see it is to go weigh it yourself.