Ask a production team how much material they lose, and you will usually receive a figure quite quickly. Ask what that figure consists of, and the conversation slows down. Some of it is in the reject bins, some went down a drain, some was sold as animal feed, some evaporated exactly as intended, and some nobody can account for.
Those categories call for very different responses. Treating them as one number leads plants to spend effort in the wrong places, sometimes attempting to reduce what is simply physics while overlooking a real loss that was hidden in the total. Good loss analysis begins by separating them.
Five things a missing kilogram might be
A practical classification for food processing separates mass into five groups, according to what happened to it and what evidence exists.
- Finished-product recovery: in-specification output, released for sale as the intended product.
- Saleable byproducts: streams with a market and a price, such as pomace, bran, peel sold for feed, or trim sold as ingredient.
- Moisture and process mass changes: mass that changes through the normal physics of the process, including evaporation, cooking loss, or water uptake.
- Documented losses: real losses with a known origin and a record, such as sorting rejects, QA rejections and clearance at changeover.
- Unresolved discrepancy: what remains once everything else has been accounted for.
Only the fourth group is loss in the everyday sense, and a part of the fifth may be. Yet they often make up the smallest shares of the total. An example run makes the point.
| Class | Mass | % of input | What it calls for |
|---|---|---|---|
| Finished product | 4,200 kg | 42.0% | The figure of record |
| Saleable byproduct (skins and seeds) | 600 kg | 6.0% | Review value, not just mass |
| Moisture removed in concentration | 4,650 kg | 46.5% | Verify against solids, not reduce |
| Documented losses | 400 kg | 4.0% | Reduce where practical |
| Unresolved discrepancy | 150 kg | 1.5% | Investigate |
Reported simply, this run has an apparent 58% loss. Reported properly, it has 4% documented loss and 1.5% unexplained. Those two descriptions lead to different meetings, and only the second can produce useful actions.
Water causes the most confusion
Water is the most common reason for misreading mass in food processing. Concentration, drying, cooking and freezing all change water content, and milling typically adds water through tempering before losing some again. In some operations output can exceed input on a raw mass basis at certain stages. That is not a good result or a bad one. It is simply water.
The solution is to track a conserved quantity where the process permits it. If puree enters an evaporator at 13°Brix and leaves at 16°Brix, soluble solids should be largely conserved while total mass falls. A solids balance can estimate how much water should have left, and comparing it with the actual outcome gives a far more meaningful view of the stage. Dry-matter measures do the same job in drying, ingredient and milling contexts.
One caution: these calculations are only as good as the samples behind them. A Brix reading taken once per batch may not represent a stream that varies, and a small error in concentration becomes a large error in calculated water. So a mass derived this way should be marked as calculated, rather than presented as equal in standing to a weighbridge reading.
Where losses tend to hide
Total-run balances show that a gap exists. Locating it requires looking at stages, and most plants weigh at very few of them. Even so, certain locations recur often enough to deserve attention.
- Intake and tare: gross-to-net errors, soil and foreign-matter allowances, and rounding on receipts set the denominator for every later calculation.
- Preparation: peel, trim and core losses move with raw-material size and condition, and with machine settings. They frequently go unweighed because the material leaves on a conveyor.
- Extraction and separation: material left in the residual stream is often the largest controllable recovery opportunity, and because it is sold as byproduct it may not register as loss at all.
- Tanks, pipework and hold-ups: product in transfer lines, tank heels and hoppers, which may be flushed at changeover and rarely measured.
- Filling: overfill against declared weight gives away product without any record of it. The filler count says 12 kg per drum, while the true average may differ.
- Cleaning: solids leaving through clean-in-place and washdown are genuine losses that tend to be invisible without effluent measurement.
Some losses exist only in the records
A portion of any reported loss may not be physical. It can be an artefact of how and when things were recorded, and this portion should be ruled out before anyone starts adjusting equipment.
Work-in-progress is the classic case. If a run closes with product in a tank and no level is recorded, that run appears to lose material and the next appears to gain it. Averaged over a month this cancels out, but at the level of an individual run it can produce misleading swings in yield. Rework behaves similarly. When material returns to the process, it can be counted as output in one batch, then fail to appear as input in the next, flattering the second.
Inventory adjustments introduce a timing problem, in which a correction for last month's error is booked this month. Scale drift creates a persistent bias that looks like a stable process characteristic. These are all fixable, and fixing them has a value beyond the immediate analysis, since every later calculation depends on them.
Treating the unresolved figure with respect
Many plants absorb the unexplained remainder into a general category, sometimes called invisible loss or unaccounted, and allow for it in the standard. This is understandable, and often unavoidable to a degree, since no balance closes perfectly. But the habit has a cost: once the remainder has a budget, nobody asks about it.
A better approach is to report it as its own line and watch how it behaves. A stable, small residual likely reflects measurement noise. A residual that moves with a certain line, shift, product or changeover pattern contains information. A sudden increase may point to a scale, a recording practice or a real physical issue. In each case the figure is a question to be pursued, and one should resist assigning it a cause merely because a cause would be convenient.
A practical way to begin
A plant does not need new instrumentation to start making progress. Choose one product on one line and build a balance for a recent sample of runs using existing records. Classify every stream and mark each as measured, calculated or assumed. The list of what could not be established then becomes a measurement plan, and it is usually short and inexpensive compared with what it reveals.
From there, add temporary weighing at one or two suspected stages, verify critical scales, and record tank levels at run close. Often the unexplained figure narrows quickly, and the items that remain are the ones worth an engineer's attention.
Conclusion: loss is a finding
The word loss implies a conclusion that has not always been earned. Much of the mass that leaves a food process is doing exactly what it should, and a good share may be earning revenue. What deserves pressure is the smaller part that is truly lost or truly unexplained. A plant that can see this part clearly can direct its limited engineering attention far more effectively than one working from a single large and undifferentiated number.