Reducing Giveaway: Batching Algorithms Explained

What giveaway is

Giveaway is the product a packer delivers above the declared or contracted weight: actual net weight minus nominal weight, summed over every pack. It is product that has been caught, handled, chilled or frozen and then handed over without being paid for.

The arithmetic is simple and the effect is large. A 5 kg box packed to an average of 5.100 kg carries 100 g of giveaway, or 2 %. Over 1,000 boxes that is 100 kg of saleable fish. On a vessel that packs several thousand boxes per trip, giveaway is often the single largest controllable loss between the haul and the invoice.

Giveaway should not be confused with legitimate allowances. If fresh fish loses weight through drip between packing and delivery, a packer may deliberately add a margin so the box still meets its declared weight at the customer. That margin should be measured and set on purpose, not left to chance.

Why packs end up overweight

Packs end up overweight because fish come in discrete pieces of different sizes and a box is normally closed only once it has reached the target, so the last piece almost always overshoots.

Two sets of rules push packers towards the minimum-weight approach. The first is the buyer's contract, which frequently states a minimum net weight per box. The second is legal metrology for prepackages. Under the average system of Directive 76/211/EEC, a batch of e-marked prepackages must meet three rules: the average content must not be less than the nominal quantity; only a small proportion of packs (in general no more than 2.5 %) may be short by more than the tolerable negative error (TNE); and no pack may be short by more than twice the TNE.

Table values as summarised by the UK Business Companion guidance on packaged goods, based on the same average system. For a 5 kg e-marked pack the TNE is 75 g. Many trade boxes of fish are sold on actual (catch) weight or under contract terms rather than as e-marked prepackages, so check which regime applies before choosing a target. The legal metrology hub explains the background.

Sequential filling: the baseline

Sequential filling, adding pieces until the box reaches the target and then closing it, produces an average overweight of roughly half the average piece weight, and more when piece sizes vary widely.

The overshoot of the last piece grows with both the average size of the pieces and the spread between small and large ones. As an illustration, fish averaging 450 g with a standard deviation of 120 g leave an average overshoot of about 240 g per box. On a 5 kg box that is close to 5 % giveaway, before any operator adds "one more for safety".

This explains two field observations. Packing larger fish into the same box size raises giveaway almost proportionally. And mixed sizes raise it further, because a wide spread makes a large final piece more likely. Grading into weight classes first (see Weight Grading Fish) narrows the spread, but it cannot remove the overshoot of the final piece.

How batching algorithms reduce giveaway

Batching algorithms reduce giveaway by keeping several choices open, so that the piece that closes a pack is selected because it fits, not because it happens to arrive next.

Best-fit across open batches

A grader with several batching outlets keeps several packs open at the same time. Each weighed piece is sent to the outlet where it does the most good: to a pack it closes with the smallest overweight, or, if it closes none acceptably, to a pack that still has room so a later piece can close it. The more open batches, the more likely it is that an arriving piece fits somewhere well.

Combination selection

A combination system holds a number of weighed pieces or portions in buffers (pockets, hoppers or positions on a belt). For each pack it evaluates combinations of buffered pieces and releases the combination whose total reaches the target with the least overweight. The number of combinations grows quickly: 10 buffered pieces give 175 combinations of one, two or three pieces, which gives the algorithm many near-perfect options.

Bulk fill plus top-up

Many practical systems combine both: most of the pack is filled quickly and roughly, and the final part is chosen by best-fit or combination logic from weighed pieces. This keeps throughput high because only the top-up needs precise selection.

Worked example with hypothetical numbers

This is an illustrative simulation, not a measured result of any machine. It uses invented but plausible inputs to show the order of magnitude of the differences between strategies.

Inputs: target 5.000 kg per box, minimum-weight rule (no box below 5.000 kg), piece weights normally distributed with mean 450 g and standard deviation 120 g, limited to 200–900 g. About 11 pieces per box. Each strategy was simulated over thousands of boxes.

The best-fit result depends strongly on how much overweight the algorithm accepts before closing a pack: a tighter acceptance limit lowers giveaway but keeps packs open longer. In the simulation, accepting up to 30 g gave about 18 g average giveaway; accepting up to 100 g gave about 52 g. This trade-off between giveaway and packing rate is the main tuning decision on any batching line.

Practical limits on board

On a vessel, the achievable giveaway is limited less by mathematics than by handling, space, product quality and weighing accuracy.

• Weighing uncertainty. An algorithm cannot aim closer to the target than the scale can measure. If individual piece weights are uncertain by ±10 g and a pack contains 11 pieces, the uncertainty of the pack total must be allowed for in the target. On a moving deck this includes the effect of vessel motion, discussed on the weighing at sea hub.
• Piece count rules. Buyers may require a minimum or maximum number of fish per box, or uniform sizes within a box. These constraints remove combinations and raise giveaway.
• Handling and quality. Every extra movement of a fresh fish costs time and can cause bruising. Buffers must be designed for wash-down.
• Frozen product. Frozen blocks are made in fixed moulds, so batching happens before freezing; glaze must be accounted for separately because declared net weight excludes glaze.
• Drift of the target. Operators tend to raise the target "to be safe". Lock targets in presets and review changes in the audit trail.

Measuring giveaway per preset

Giveaway can only be managed if it is measured per product, per preset and per period, using every recorded pack weight rather than spot checks.

The minimum statistics per preset are: number of packs, total net weight, mean, standard deviation, minimum, maximum, number of packs below nominal, and average giveaway (mean minus nominal). Tracking these per shift shows whether a change of target, grader recipe or crew changes the result. The fields to record and export are described in Weighing Data Logging at Sea.

How WPL approaches this

For manual or semi-automatic packing on static scales such as the M2 Series, WeightControl logs every weighing automatically and shows statistics per preset and period, which makes giveaway visible per product and per shift. Where packs are built by an automatic grader or batching line, bring its pack weights into the same data system, so that manual and automatic packing are judged on the same basis. Achieved giveaway always depends on product, piece size distribution, targets and constraints, so it should be assessed on the actual product. More on the whole process chain is on the onboard processing hub.