Chain Operations
Dimensional Tolerances for Tableware: How Close Is Close Enough
Ceramics are fired, not machined — every dimension carries a tolerance, and every interface (lid, stack, rack, tray) needs to know it. Setting tolerances that match process reality.
A dinner plate specification that says "27 cm" is not a specification — it is a wish. Ceramics emerge from kilns, not machine tools: clay shrinks as it dries, shrinks again as it vitrifies, and shrinks slightly differently by position in the kiln and moisture in the body. Every dimension on every piece of tableware carries a tolerance, whether anyone wrote it down or not. Programs that write it down negotiate the tolerance; programs that don't inherit the factory's default — and meet it at the lid that doesn't seat, the stack that rocks, the rack that won't close.
This guide covers tolerance-setting for tableware: what the process can honestly hold, where tight matters, and how to write the sheet.
The Process Reality First
Fired ceramics' dimensional variation comes from layered sources:
Shrinkage scale. Bodies shrink from wet to fired by double-digit percentages — meaning a 1% variation in shrinkage moves a 27 cm plate by nearly 3 mm. The factory's shrinkage-control discipline (body preparation moisture, forming consistency, firing) is the tolerance's foundation; the buyer's tolerance has to sit above what the controlled process actually delivers.
Kiln position. The thermal-history map our kiln guide describes has dimensional consequences: pieces from different load zones drift measurably in both size and flatness within one lot.
Format asymmetry. Non-round formats (the rectangular trays, the oval fish plates) hold tolerances differently than rounds — corners and long spans amplify warp and shrinkage-direction effects. The tolerance sheet knows its geometries.
The reference points buyers anchor on — commercial ceramics' working tolerances for well-controlled production run in the neighborhood of ±0.5–1.5% on diameters and lengths (a 27 cm plate's honest band measuring a few millimeters), with flatness and height bands of similar scale. The exact capability differs by body, format and factory — which is why the number that matters is the one your factory's own process data supports, agreed at the first article (our FAI guide's measured bench).
Where Tight Tolerance Actually Matters
Tolerance budgets are not uniform — they concentrate where pieces interface:
Interface | The tolerance that matters | Failure when loose |
|---|---|---|
Lid-to-bowl | Diameter + rim profile fit | Lids rock, wobble, spill heat |
Stack seating | Foot-to-well seating | Tilted stacks, grinding (our footring guide) |
Rack fit | Height + diameter classes | Pieces jam, racks won't roll |
Tray insets (set meals) | Footprint within the tray grid | Compositions don't sit flat |
Divider walls (compartments) | Wall height | Sauce migration (the bento lesson) |
The discipline: interfaces get the tight bands; free dimensions (a plate's well diameter that touches nothing) run the looser bands the process holds naturally. Tolerance budgets spent everywhere buy cost and rejects; spent at interfaces, they buy the fleet's function.
The warpage dimension. Flatness is the quiet interface in every stack: a plate that rocks on the glass plate at first article (the FAI checklist's wobble test) will rock in every stack it joins for its service life. Flatness tolerance belongs on the sheet with its test method — the flat-surface wobble check, defined.
The class system. Chain programs often specify dimensional classes rather than absolute tolerances: stack-height classes (racks and shelving), diameter classes (racks, tray grids, lid families). The class logic turns individual variation into interchangeable function — the same standardization reasoning our par and kit guides run on.
Writing the Tolerance Sheet
The working format, per SKU:
Nominal dimension + tolerance band per dimension type: diameters, lengths, heights — with the interface dimensions flagged for their tighter bands
The measurement points: where the caliper lands (rim outside? well inside?) — dimensions without measurement points measure opinions
Flatness/rock test: the method and the pass standard
Gauge points for wall/gauge minimums: plate rims, flatware necks (the gauge disciplines our flatware guide sets)
The acceptance tie: first-article measurement (the FAI bench), lot sampling per the AQL plan (our sampling guide) — tolerances enforce through the inspection system, or they are prose
One page per SKU family; the sheet versions with the program (the change-log discipline our FAI guide runs).
Tolerances and Cost, Honestly
Tighter tolerance is bought, not wished for: extra forming control, sorting labor at the factory, higher reject rates — the price curve steepens past the process's natural capability. The buyer's leverage is knowing where tolerance is worth buying (interfaces, lid families, the classes that make fleets interchangeable) and where the natural band serves perfectly (free dimensions no piece ever touches). The specification that spends its budget at the interfaces runs both cheaper and truer than the one that demands precision uniformly.
Frequently Asked Questions
Why can't ceramics hold machining tolerances? The material transforms — shrinkage through drying and vitrification, with kiln-map variation on top. The process controls the shrinkage tightly; it does not delete it. Tolerance-setting starts from that physics.
What tolerance do lids actually need? Enough to seat without rock across the batch pairing — lid-to-bowl families tolerance as pairs (the pairing test at first article: lid roulette across the sample set), not as independent pieces.
How do we handle non-round formats? Diagonals and corner-squareness join the dimension list; warp tolerance tightens on long spans. Round-format habits imported onto a long rectangular tray meet their failure at the first stack.
Who measures — us or the factory? Both, to the same sheet: the factory's lot data (logged per the traceability discipline) and your verification at first article and audit. The shared numbers keep the tolerance a contract instead of a dispute.
What is the most common tolerance-sheet mistake? Single-number specs without measurement points — "27 cm" with no band, no caliper position, no flatness method. The second most common: uniform tightness everywhere, paying process prices for dimensions nothing touches.
