The Order Is Not the Lot: How to Divide Ceramic Production Before You Inspect It

Every inspection plan arrives with a page of agreed numbers: sample size, acceptance level, defect classes, the drawing that fixes the reference. None of them states the number the others depend on: how many pieces count as one lot.

Products in this guide: Vintage Couple Illustration Ceramic Mug 365ml · Colorful Mini Ceramic Bowls

Twenty thousand mugs made across six weeks are one purchase order and, in production terms, four or five populations: separate clay deliveries, a glaze vat remixed, a decal reprinted in week five, pieces fired in different cycles. The order is a commercial unit; the lot is the unit of judgement, and only the first appears on the purchase order.

What follows is where to place the cuts, what each direction costs, and how to write the rule so an inspector has something to stand on.

See the full ceramic sourcing process map.

The lot size box is not your order quantity

A sampling table takes one input before it tells you anything: the lot size. Enter twenty thousand and the table returns a sample, an acceptance number and a rejection number. The arithmetic is exact, and it answers a narrower question than it appears to: how many pieces to draw from one population produced under essentially the same conditions.

That assumption is the whole argument. A sampling plan was built for goods that share a history. When the order behind the number does not share one, the table is not being misread; the input is. What follows is a confident verdict on a population that does not exist.

What makes fired ceramics homogeneous

Five variables decide whether two pieces belong to the same population, and each can move without a line changing on the purchase order.

Body. Clay arrives by the tonne and its mineral content shifts between deliveries, so a new delivery is a new starting point.

Glaze. A vat is mixed, consumed and mixed again. The second mix is close rather than identical, and the gap appears as a surface that matches under a lamp and drifts by the third firing.

Decoration. Decal runs carry their own ink batches and registration. A mid-order reprint is a different decoration even when the artwork file never changed.

Firing. Kiln position and cycle profile move shrinkage, colour and glaze surface together. Two cycles are two conditions.

Forming. A plaster mould wears across its working life, so pieces from the first week and the last week of a long run come from tools at different stages.

A change in any one of these creates a second population; whether it is acceptable is a separate question, and can only be asked once the first is separated.

Why one merged lot cannot see a drift

The instinct is to keep it simple: one order, one lot, one verdict. A larger lot does draw a larger sample, so the arithmetic looks generous. Two failures follow, running in opposite directions.

The first is scale. A merged lot produces a single verdict for everything. If one firing out of six ran wrong, the verdict lands on all six and the remedy is applied to goods that were never in question. Rejecting a container to correct a fraction of it is how a quality decision becomes a schedule loss.

The second is subtler. Sampling finds pieces that differ from a reference; it cannot find a reference that moved. A glaze mix that shifts the whole run by a small uniform amount produces no defective piece, because every unit sits inside the tolerance and the tolerance is what changed. In one lot there is no comparison group inside the sample, so the drift passes inspection and appears months later, when the reorder does not match the first delivery. The master sample guide explains how sealed references keep batches identical.

What splitting costs

Cutting an order into eight lots is not free. Each lot carries its own minimum sample, and sample sizes do not shrink indefinitely with lot size, so inspection load per piece climbs as lots get small: eight lots of two thousand five hundred pieces call for more pieces drawn in total than one lot of twenty thousand.

Each lot also carries an independent verdict. If one lot has a ninety-seven percent chance of passing, eight have roughly a seventy-eight percent chance of all passing. The same process, judged in more pieces, returns more failures.

Warehouse obligations come last: a lot has to be marked, stored apart and shipped as a unit, and a warehouse that consolidates two of them silently undoes the structure.

A default rule, and how to write it

Most programmes do not need a bespoke scheme:

• One lot is one combination of body delivery, glaze mix, decoration version and firing cycle.
• An order made in a single continuous run and shipped in one container is one lot.
• Production spanning more than about two weeks is cut by week or by firing, whichever changes first.
• Colour-critical programmes are cut by glaze mix rather than by calendar day.
• Any change to decoration, artwork version or packing starts a new lot.
• Runs under roughly five hundred pieces stay together even when they mixed two cycles, with the mix recorded.
• No lot should be larger than what one inspector can physically sample in a shift.

Three things have to travel with the rule. The definition, naming the variables that must stay constant inside a lot and stating that a change starts a new one. The marking, so the lot is visible on the carton and the unit label instead of living in a spreadsheet. The consequence, written per lot rather than per order: each lot is judged on its own, a failed lot is held and dispositioned on its own, and the passing lots ship. That sentence keeps a local problem from becoming a container-wide dispute. Revision control belongs beside it, because an artwork or packing change creates a new lot. The spec freeze guide explains how artwork and packing files are locked between orders.

The question that decides it

One test is worth applying before the first inspection. If this entire lot failed, what would you do?

If the honest answer is reject the container and produce again, the lot is too large, because you have just described a decision nobody in the room would take. If the honest answer is ship the rest and deal with the bad part, you have already agreed there is more than one lot, and the remaining task is writing down where the line falls.

Fired ceramics do not fail in averages. They fail in runs: one firing, one glaze mix, one week of a press. A lot definition is the cheapest instrument a buyer owns for finding that run in the factory rather than in a container.