Operations · 5 min read

Nine seconds: inside a grader's decision

A trained grader spends about nine seconds on a platform and makes seven judgements. Here they are, in order, as they happen.

OperationsOctober 7, 2025Dale, repair lead
Request · quote · pickup

Question about this post?

Put it in the message field — technical questions go straight to the yard.

Letters only — no numbers.

As it appears on your paperwork.

A real inbox you read — throwaway addresses are rejected.

US or Canada only. Format: (###) ###-#### — it formats as you type.

Optional.

Optional.

Fills itself in from your ZIP.

US ZIP (23224) or Canadian (K1A 0B1). Checked against real ranges.

Whole number of units.

Pick one or type your own.

Routes your request to the right person.

Leave blank if you are unsure.

Plain text, up to 2,000 characters.

Every field is checked before it leaves your browser and again on our server — phone numbers against the North American numbering plan, ZIP codes against real USPS ranges. It keeps the bots out and means a real human reads every submission within one business day.

We tried a conveyor on the sort line once. It was slower than three people and a trained eye, and we sold it within a year. A grader averages nine seconds per platform including handling, and in those nine seconds makes a sequence of decisions in a fixed order. Here is the sequence, because customers often ask what they are paying for.

Second one to two: footprint

Before condition, before anything: what size is this? A worn 48 × 40 measuring 47½ is still a 48 × 40; a 40 × 48 is a different pallet. This is a glance, not a measurement, and it is the thing a new grader gets wrong most often in week one.

Second two to three: construction

Stringer or block. Notched or solid. Board count on the top deck, roughly. This determines which spec pile the platform could belong to, which narrows everything that follows.

Second three to five: the lead boards

Both ends, both decks. Cracks, splits, missing ends, crushing. This is forty-one percent of all failures and it is where the eye goes first because it is where the damage is. A splintered end tells a different story from a crushed one — the first is a tine drop, the second a dock plate step — and a good grader is reading the customer's dock as well as the pallet.

Second five to six: the stringers

Specifically at the notches. A break at the notch is the classic structural failure, because the notch concentrates stress. Then the ends, then along the length for splits that run with the grain. This decides repair-versus-dismantle more often than anything else.

Second six to seven: fasteners

Running a hand or an eye across the deck for anything standing proud. A protruding or backed-out nail is an automatic reject regardless of how good everything else looks — it tears film, cuts hands and fouls conveyors. Around nine percent of bench work exists for this alone.

Second seven to eight: moisture and staining

Not appearance — risk. Is this damp enough to bloom in a wrapped stack? Is that discolouration surface mould, blue stain or actual decay? The thumbnail test on a suspect board takes a second and separates cosmetic from finished.

Second eight to nine: the call

Grade A, Grade B, repair, or dismantle. Written or thrown, depending on the pile. And then the next one.

Why we have not automated it

Vision systems are good at measuring and poor at judgement. A camera can tell you a board is cracked; it cannot easily weigh a hairline check against the lane the platform is going to, or notice that this trailer load has a pattern of damage worth telephoning a customer about. The nine seconds contain context, and context is still the thing humans are cheap at.

That may change. For now, every platform leaving this yard has passed an eye belonging to somebody on our payroll, and that is a deliberate choice rather than a technology gap.

What a new grader gets wrong, in order

We train from zero — nobody arrives knowing how to do this — and the error sequence is remarkably consistent across every person we have trained.

  • Week one: footprint. Reading a worn 48 × 40 as undersized, or missing a 40 × 48 entirely. The fix is a tape measure for a few days until the eye calibrates.
  • Week two: over-rejecting. New graders reject anything with visible damage, because damage looks like failure. Most of it is repairable and the repair is cheap.
  • Week three: under-rejecting. The overcorrection. Hairline checks at a notch get passed because they look minor, and a notch is the one place where minor is not.
  • Month two: staining versus decay. Blue stain looks alarming and is structurally irrelevant; early decay looks mild and is fatal. The thumbnail test is learned here and never unlearned.
  • Month three: context. Realising that the same platform is A-grade for one customer's lane and over-specified for another's. This is the point at which somebody becomes a grader rather than an inspector.

Reading the customer from the pallet

A trailer load carries a pattern. Uniform crushing on top lead boards across two hundred platforms is a dock plate with a lip. Splintered bottom ends is a tine dropping into a gap. Damage concentrated at one end only is a habit or a single bad door. Soft bottom boards across the load is outdoor storage on concrete.

When a grader sees a pattern, somebody telephones the customer. Those calls are the most valuable thing this sort line produces, and they are the specific reason we have not replaced it with a camera — a vision system can grade a platform and cannot notice that a load is telling you something about a building three counties away.

The economics of the nine seconds

At nine seconds including handling, a grader processes around 350 platforms an hour and the three of them handle a full trailer in under half an hour. The conveyor system we trialled managed a nominal higher throughput on paper and a lower one in practice, because every judgement call it could not make became a manual exception, and exceptions queue.

We sold it within the year. We expect three people and three eyes to still be the arrangement in a decade, and if that changes it will be because somebody solved judgement rather than measurement.

Dale, repair lead · Pallets Recycle, 1113 Ingram Ave, Richmond VA · published October 7, 2025 · 904 words. Corrections welcome — tell us if we have this wrong.

Next step

Let's get wood moving

Whether you have 40 broken pallets behind the building or need 4,000 a month, the next step is the same: tell us the footprint and the frequency.

Start a request