Yard notes
How we grade: a walk down the sort line
Request pricing
Everything marked with a star is required. The rest helps us quote faster.
Visitors always want to see the grinder. The grinder is the least interesting machine we own. The interesting thing is a person with a hook standing at the end of a conveyor.
Everything about a pallet yard's economics is decided in about four seconds per unit at the sort line. Route a repairable pallet to the grinder and you have destroyed nine dollars. Route a cracked stringer to the repair line and you have wasted four dollars of labour and produced a pallet somebody might rack.
The line, in order
- 1
Gate and scale
Trailer weighed, load photographed, batch reference opened. Everything downstream ties to that number.
- 2
Unload and stage
Stacks broken down onto the infeed. Pooled pallets pulled here, not later.
- 3
Thaw or dry, if required
November to March, and after any wet load. Grading wet wood is grading blind.
- 4
Sort line
The four-second decision. One grader, one hook, four outfeed lanes.
- 5
Route
Reuse, repair, remanufacture, dismantle. Four physical destinations.
- 6
Verify
A second grader spot-checks 3% of the reuse lane. This is how we know our accuracy.
The four-second decision
It is four questions, asked in a fixed order because the order minimises the number that need asking.
- Is it sound? Look down both stringers. A crack anywhere structural ends the inspection — it goes to dismantle. About 12% exit here.
- Is it clean? Mould, chemical staining, oil, char. Rejection, isolated, not into any stream. About 1%.
- Is it complete? Missing boards and shiners. Missing board goes to repair; shiners alone stay in the reuse lane with a quick shear.
- Is it original? Companion stringers and plugs. Present means Grade B; absent with ≤2 replaced boards means Grade A.
Most pallets are answered by questions one and four. The whole thing is a decision tree that fits in somebody's hands.
How we know it is working
| Measure | Target | Last quarter |
|---|---|---|
| Grade A units correctly graded | ≥ 98% | 98.7% |
| Grade B units correctly graded | ≥ 96% | 97.1% |
| Repairable units wrongly sent to dismantle | ≤ 2% | 1.4% |
| Unsound units wrongly sent to reuse | 0% | 0.02% (3 units) |
| Customer grade disputes per 1,000 units | ≤ 0.5 | 0.31 |
The row that matters is the fourth. Three unsound units reaching the reuse lane in a quarter is three too many, and each one is investigated — which pallet, which grader, what time of shift. Two of the three last quarter were in the last forty minutes of a Friday.
Why we do not automate it
We have looked at vision systems twice. They are genuinely good at the things that are easy — missing boards, gross geometry, obvious damage. They are poor at the things that decide value: distinguishing a check from a crack, recognising a companion stringer under weathering, and smelling mould before seeing it.
A system that gets 92% of decisions right at four times the speed sounds attractive until you price the 8%. On our volume, an 8% error rate against a 1.3% human one is roughly $280,000 a year of misrouted value. When a vision system beats a trained grader on the hard cases we will buy one, and I expect that to happen. It has not yet.
The job
Grading is the best-paid non-management role in this yard, ahead of drivers and well ahead of the repair line. That is not generosity. It is the only role where a person's individual accuracy moves a six-figure number, and paying for it is the cheapest capital investment we have ever made.