Custom Jewelry Manufacturing Tolerances: How to Define Dimensions and Weight Requirements

A CAD file can specify a ring band at exactly 3.00 mm wide and a finished weight of 4.50 grams. The cast, polished, stone-set piece that arrives from the factory will almost never hit both numbers to the decimal—and that is not a defect. It is the normal consequence of translating a digital solid into a physical object through casting, filing, polishing, and assembly. What separates a well-managed custom jewelry program from a dispute-ridden one is not the elimination of variation but the deliberate definition of how much variation is acceptable, on which features, and measured how. Manufacturing tolerances are the language that makes that definition explicit.

What a Tolerance Is—and What It Is Not

A manufacturing tolerance is the acceptable range of variation around a specified nominal value. A finding listed as 10.0 mm ± 1.0 mm may measure anywhere from 9.0 mm to 11.0 mm and still conform. A tolerance does not give a manufacturer license to produce carelessly; it recognizes that jewelry is made through processes—casting, polishing, stone setting, assembly—each of which introduces small, predictable shifts. The goal is to bound those shifts so they do not compromise fit, function, durability, or appearance, and to ensure buyer and factory share the same definition of “acceptable”.

This distinction matters because custom jewelry is held to a different scale of scrutiny than industrial products. A dimensional deviation that would be invisible on a machined bracket can be immediately noticeable on a 2 mm-wide band or a pavé setting. Tolerances must therefore be feature-specific. No single blanket figure should be assumed for every dimension on a piece.

Dimensional Tolerances: From CAD to Finished Metal

Casting Shrinkage and the Compensation Problem

The first major source of dimensional variation is casting shrinkage. Metal contracts as it cools, and the wax pattern and investment mold contribute their own dimensional changes before the metal even solidifies. Reported shrinkage figures vary widely because they depend on alloy, model-making method, mold material, and casting technique. Lost-wax molding shrinkage has been described in ranges from roughly 5–10% for pattern-making allowances, while metal shrinkage alone in investment casting is often cited at 1–3% depending on the alloy. One technical discussion of historical casting technology estimated a combined linear shrinkage of 2.5 ± 0.5% for wax/mold and metal together. These numbers are not interchangeable; they describe different stages of the process.

The practical consequence is that CAD models are not simply printed at final size. Professional CAD applies a metal-specific shrinkage factor so that after cooling the cast piece lands on the intended dimensions. Different metals require different factors—silver may shrink roughly 2–3%, brass or bronze approximately 1.5–2%, with gold alloys requiring their own compensation. If the caster uses a custom alloy, the shrinkage factor must be confirmed with that caster rather than assumed from a generic table.

For the brand or designer defining requirements, this means shrinkage is not a tolerance that can be inspected away after the fact; it is a process variable that must be agreed before the CAD is finalized. The manufacturing order should state which alloy is being cast, who is responsible for applying the shrinkage compensation, and whether the CAD file delivered to the factory is already scaled or is a final-dimension model from which the factory will scale.

Which Dimensions Deserve Tight Tolerances

Once a piece is cast, dimensional control should be concentrated on features where variation has real consequences. The first category is fit-critical dimensions: ring inner diameter, clasp and hinge interfaces, and any surface that must mate with another component. Ring sizing is the clearest example. ISO 8653 defines ring sizes in terms of inner circumference and specifies diameters and tolerances for ring sticks used during manufacturing measurement. A ring that is a quarter size large is not merely a cosmetic variation; it changes the wearing experience and the commercial size SKU.

A common industry benchmark for ring size tolerance in volume production with hand finishing is on the order of ±0.3 US size, corresponding to roughly ±0.4 mm on the inner diameter, though tighter control is achievable when the sizing is treated as a critical feature rather than a general dimension. For channel-set or bezel-set pieces, the seat diameter and bearing depth are similarly critical: a 1.75 mm seat cut for a 1.75 mm stone may carry only about 0.05 mm of working tolerance in the setting itself, because even a tenth of a millimeter changes whether a stone sits flush or proud.

The second category is structural and durability-critical dimensions: band thickness at the bottom of a ring, gallery rail thickness, prong tip diameter, and hinge wall thickness. These are the dimensions most vulnerable to being “polished away” during finishing. A finished weight that comes in well under the expected range usually signals that wall thickness was cut down too far during finishing rather than a benign material loss. Specifying a minimum—not only a nominal—on structural thickness is therefore more protective than specifying weight alone, because weight is a lagging indicator of a structural problem that has already occurred.

The third category is appearance-critical dimensions: band width consistency, stone position in a pavé row, the height of a setting above the finger, and the symmetry of matched earrings. These features may have generous functional tolerance but tight perceptual tolerance. A pavé stone sitting 0.2 mm higher than its neighbors is a dimensional variation that reads immediately as a quality failure.

A practical way to assign tolerances is to rank every dimension on the technical drawing by consequence. Critical features that affect fit, stone security, or structural integrity should carry the tightest range the process can reliably hold; cosmetic features can carry a looser range; and non-functional dimensions can be left to general tolerances. The more important a measurement is to the product’s function, the more carefully it should be controlled.

Stone Dimensions and Setting Clearances

Gemstones introduce their own dimensional reality that the metal tolerances must accommodate. Calibrated stones are not identical. As a general working guideline, gems under one carat can generally fall within about 0.1 mm of the setting size; stones over one carat may accommodate a 0.2 mm variance, and sometimes more. Other industry guidance places the working variance closer to 0.2 mm for stones under a carat and up to 0.3 mm for larger stones. The discrepancy between these figures is itself instructive: there is no universal stone tolerance, and the acceptable clearance depends on the setting type, the stone shape, and the setter’s technique.

For custom work, the more reliable practice is to specify the setting from the actual stone measurement in millimeters rather than from a nominal carat weight or a standard size label. A carat weight describes mass; a setting grips a diameter. Two stones of the same carat weight can differ in millimeter spread depending on cut proportions and depth percentage, and the millimeter measurement is what the setter must fit. The manufacturing specification should therefore include the measured millimeter dimensions of every principal stone and a stated clearance range—how much larger the seat should be than the stone, or how much interference is acceptable in a bezel—rather than leaving the relationship to be inferred from a carat figure.

Weight Tolerances: Why the Scale Never Matches the CAD File

The Systematic Gap Between CAD Weight and Finished Weight

CAD software reports weight by multiplying the volume of a solid model by the density of the chosen alloy. That number is exact for the model. It is not a prediction of the finished object. Between the approved file and the finished piece, metal is removed at sprue removal, filing, seat cutting, rubber-wheel work, and polishing. No step adds metal back. The result is a systematic downward bias. One production-focused source reports a working rule that finished pieces land at roughly 90% of the CAD weight, with documented examples at 91.8% and 89.0% for two 18K white gold pieces.

This gap is not a tolerance in the metrological sense; it is a process offset. If the manufacturing order treats the CAD weight as the acceptance target, every well-made piece will appear to fail. The correct approach is to establish a finished-weight target separately from the CAD estimate, either by weighing an approved sample or by applying a documented finishing-loss factor for that construction type, alloy, and finishing standard. The order should distinguish CAD-estimated mass, target finished mass, and the acceptable finished range as three separate figures.

How Wide Should the Weight Range Be?

There is no single legally mandated weight tolerance for custom jewelry across all markets, but the range of published practice is informative. For small findings and components, gram weights are often averages with a variation of approximately ±5%. For finished precious-metal jewelry, a ±10% range around the stated weight is commonly used to accommodate production variation. Broader allowances appear in some contexts—around 15% for stone carat weight and up to 20% for gold weight in fashion-oriented product disclosures—but these are commercial disclosure tolerances rather than precision manufacturing targets. In technical-standard terms, one intergovernmental standard for precious-metal articles specifies that actual mass should not deviate from the mass calculated from nominal dimensions by more than ±5%.

The choice of range should reflect what the weight is being used to control. If weight is a cost-allocation and customer-expectation metric, a moderate range such as ±7–10% is workable and consistent with common industry practice. If weight is being used as a proxy for structural integrity—to catch pieces where finishing has thinned the walls—a one-sided floor is more appropriate than a symmetric band: the finished piece should not fall below a minimum weight that corresponds to the minimum acceptable wall thickness. Weight can be useful as a check, but it cannot replace dimensional or structural inspection; a piece can be within weight while a critical prong is undersized, or outside weight while every structural dimension is sound.

Weight and Precious-Metal Cost

The commercial stakes of weight tolerance are highest when the piece is made in a precious alloy whose cost is tied to metal content. A 10% weight variance on a 4-gram 18K piece is roughly 0.4 grams of gold—small in absolute terms, but material across a production run and a recurring source of friction between buyer and factory when the expectation was not written down. The manufacturing order should therefore state whether the weight figure is metal-only or total finished mass including stones, and whether it is measured before or after setting. A total-mass figure that includes stones will move when the stone carat weights move, even if the metal is dimensionally perfect; the two variables must be toleranced separately.

Writing Tolerances That Survive Quotation, Sampling, and Inspection

A tolerance is only useful if it is measurable, agreed, and inspectable. The following practices turn tolerance decisions into a specification that a factory can quote against and a quality inspector can enforce.

Define the measurement method alongside the nominal value. A nominal dimension without a method is incomplete. “Band width 3.0 mm” can mean caliper measurement at the top of the shank, at the bottom, or at the widest point after polishing. The specification should identify the measurement point on a marked drawing, the instrument class (caliper, micrometer, pin gauge, optical comparator, or vision system), and whether the measurement is taken before or after setting and finishing. For ring sizing, the measurement should reference the sizing standard and the type of ring stick or gauge used.

Tie tolerances to a drawing, not a paragraph. A CAD file shows shape but does not by itself communicate which dimensions are critical. The technical drawing should flag critical dimensions with their nominal, upper and lower limits, and the measurement method. Features left unflagged should fall under a stated general tolerance so there is no ambiguity about whether an untoleranced dimension is critical or incidental.

Specify stone fit as a relationship, not a label. For every principal stone, record the measured millimeter dimensions, the intended seat or bearing dimension, and the acceptable clearance. For pavé and channel work, specify the maximum allowable height variation between adjacent stones and the prong tip dimensional range. These are the dimensions that determine whether the piece reads as properly set or as a near miss.

Separate CAD mass, target finished mass, and acceptance range. Do not ask a factory to “match the CAD weight.” State the alloy and construction, the agreed finishing-loss assumption or the measured weight of the approved sample, and the acceptable finished range in grams. If the piece contains stones, state whether the range applies to metal only or to total mass.

Agree on shrinkage responsibility before the CAD is finalized. Identify the casting alloy, the shrinkage compensation factor or source, and whether the delivered CAD is at final dimension or pre-scaled. A dimensional failure traced to an uncompensated or miscompensated model is a process failure, not a tolerance failure, and it is far cheaper to prevent at the file stage than to discover at inspection.

Use a pre-production sample to calibrate the ranges. The first approved sample is the empirical evidence that the chosen tolerances are achievable with that factory, that alloy, and that construction. Measure it, record the actual values, and use them to confirm or adjust the written ranges before bulk production. A tolerance that has never been validated against a physical piece is a hypothesis, not a specification.

Conclusion

Custom jewelry tolerances are not an admission that manufacturing is imprecise. They are the mechanism by which precision is defined in a process that is inherently dimensional—casting shrinkage, polishing loss, stone variation—and made manageable. The dimensional side requires attention to casting compensation, to the small set of features where fit and structure matter, and to the millimeter reality of stones rather than the carat labels attached to them. The weight side requires accepting that CAD mass and finished mass are different quantities, that finishing removes metal systematically, and that a finished-weight range should be set from evidence rather than from the file. When those definitions are written into the manufacturing order with a measurement method attached, the buyer and the factory stop negotiating over what “close enough” means and start inspecting against a shared standard. That is the difference between tolerances that cause disputes and tolerances that prevent them.