Worked benchmark example
One large item, five small items, one carton.
This example uses Packmetry's lockedhighly-different-item-sizesbenchmark case. It shows what the benchmark expects, how canonical volume metrics are derived, and what the result does — and does not — prove.
Repository source
The benchmark case lives insrc/test/benchmarks/cases.ts. Its expected outcome is feasible packing with one carton, six placed item instances and zero unplaced instances.
Inputs
A deliberately uneven mix of item sizes.
The benchmark uses canonical millimetres, unrestricted rotation and the benchmark helper's fewest-cartons objective. It does not provide item weights, carton tare weight, external carton dimensions or DIM settings.
| Input | Dimensions | Quantity | Role |
|---|---|---|---|
| Large item | 80 × 80 × 80 mm | 1 | Requested item |
| Small item | 20 × 20 × 20 mm | 5 | Requested item |
| Carton | 100 × 100 × 100 mm | As needed by the plan | Internal packing space |
Locked benchmark expectation
All six item instances fit in one carton.
The benchmark expectation records a feasible result with one used carton, six placed instances and no unplaced items.
Requested item volume
Large item:80 × 80 × 80 = 512,000 mm³
Five small items:5 × 20 × 20 × 20 = 40,000 mm³
Total:552,000 mm³
Canonical carton metrics
Carton volume:100 × 100 × 100 = 1,000,000 mm³
Empty volume:1,000,000 - 552,000 = 448,000 mm³
Volume utilization:552,000 / 1,000,000 = 55.2%
What the example teaches
Volume is useful, but geometry still decides whether the plan is valid.
55.2% utilization does not prove fit.
The volume arithmetic tells us how much internal carton volume is occupied after a plan exists. It cannot prove that a collection of three-dimensional items can be placed without overlap or boundary violations.
The benchmark locks the outcome, not a universal placement recipe.
The case records the expected status, carton count and placed/unplaced item counts. Production planning still relies on solver-generated placements and the independent verifier rather than treating volume arithmetic as a substitute for placement geometry.
No weight or DIM conclusion belongs to this case.
The benchmark does not supply item weights, carton tare weight, external carton dimensions or dimensional-weight settings. Packmetry therefore should not invent gross-weight, DIM-weight or shipping-price conclusions for it.
One carton is the benchmark result, not a global-optimality proof.
The current baseline is deterministic and bounded. This example shows the locked expected outcome for this test case; it does not turn the baseline heuristic into an exhaustive proof of mathematical optimality.
Trust boundary
To understand how candidate generation, independent verification, objective ranking and canonical-plan construction fit together, read thepacking algorithm methodology.
Try your own case
Replace the benchmark dimensions with the items and cartons you actually have.
Personal is the simpler route for everyday packing. Business adds carton inventory, costs, weights and DIM settings while using the same core planning architecture.