Short answer: budget one vulcanized rubber mold for 300–500 wax pulls on production S925 work, then multiply by the cavity count — a 10-cavity tool covers 3,000–5,000 castings. That band is a planning assumption, not a published warranty. The binding number is the rated pulls written into your purchase order after a 50-shot qualification run on your own design.
For a 3,000-piece run, the minimum mold count is 3,000 ÷ (cavities × rated pulls), rounded up; then add molds until no single mold carries more than 70% of its rated pulls. On a 4-cavity, 300-pull design that means four molds, not one plus a spare. On payment: normal wear before the rated count is the factory’s re-cut, because the mold is a consumable already priced inside the run. Wear after the rated count, a mid-run design change or buyer-caused damage is the buyer’s. Without a rated count in the PO, the buyer has no claim at all.
What is a rubber mold in S925 casting, and what actually wears out?
The master is the durable model that carries the design; the rubber mold is packed and vulcanized around it, so every cavity is a negative. One injection cycle is a shot, and one shot fills every cavity — a 10-cavity mold produces ten wax patterns per shot, and those waxes become the casting tree.
Amara’s published guidance separates the two: “The master is durable; the rubber mold is consumable and wears with every shot.” The same page lists the failure order — “Cavities dull, flash lines thicken, stone seats move, and clasp details soften” — and warns that “Heat, humidity, and compression can also deform a mold that has not run for months.” See mold retention and casting lots on a mid-season reorder.
Vulcanized rubber is a thermoset: once cured it cannot be reflowed, so repeated clamping takes permanent compression set and hot wax ages the cavity surface. A re-cut therefore means cutting a new rubber mold from the master, not repairing the old one. Whether that is cheap or becomes a new setup depends on the master.
How many castings can one rubber mold make before it needs re-cutting?
There is no universal number, and a supplier who quotes one without asking about cavity count, stone seats and press settings is quoting a brochure. Amara publishes the wear mechanism but not a shot count.
For budgeting, use 300–500 wax pulls per mold. Plan 300 when the design has pavé seats, thin prongs, a working clasp or deep undercuts. Plan 500 for a simple, heavy shape with no setting. Treat anything above 500 as a bonus, never as a promise.
The “thousands of injections” claim in supplier literature answers a different question. A mold that still closes and fills can be thousands of shots old; a mold that still holds a 1 mm pavé stone inside a ±0.05 mm seat fails much earlier. You need the second number.
| Design type | Planning pulls | What fails first |
|---|---|---|
| Plain band or solid pendant | 500 | Flash line and parting-line step |
| Bezel or channel-set ring | 400 | Seat ovality, stone looseness |
| Pavé ring with 1–2 mm stones | 300 | Seat drift, prong tip wear |
| Working clasp or hinge | 300 | Detail softening, function loss |
| Deep undercut or relief | 300 | Tearing at the undercut |
These are budget values, not a warranty. A rated count is a reliability estimate, not a material constant: the NIST/SEMATECH e-Handbook chapter on assessing product reliability treats product life as a distribution estimated from failure and censored data. That is the logic of a 50-shot qualification run — measure at shots 1, 25 and 50. If a seat has moved by shot 50, 300 pulls is too optimistic for that design.
How many molds does a 3,000-piece run need?
Use one formula in the quote: pieces per mold = cavities × rated pulls. Divide 3,000 by that figure and round up; you cannot buy 0.2 of a mold.
| Cavities | Rated pulls | Pieces per mold | Minimum molds for 3,000 | Load per mold | Verdict | Molds to order (70% rule) |
|---|---|---|---|---|---|---|
| 4 | 300 | 1,200 | 3 | 83.3% | Over 70%; add a mold | 4 |
| 4 | 500 | 2,000 | 2 | 75.0% | Over 70%; add a mold | 3 |
| 6 | 300 | 1,800 | 2 | 83.3% | Over 70%; add a mold | 3 |
| 6 | 500 | 3,000 | 1 | 100.0% | No margin; add a mold | 2 |
| 8 | 300 | 2,400 | 2 | 62.5% | Workable; spare optional | 2 |
| 8 | 500 | 4,000 | 1 | 75.0% | Over 70%; add a mold | 2 |
| 10 | 300 | 3,000 | 1 | 100.0% | No margin; add a mold | 2 |
| 12 | 500 | 6,000 | 1 | 50.0% | Comfortable | 1 |
The load column is 3,000 ÷ (cavities × rated pulls × molds), using the minimum mold count; the last column re-runs the count at 70% of rated pulls, rounded up. The verdict is read from the load in four bands: 100% is no margin; above 70% add a mold before the first shot; above 50% and up to 70% is workable with a spare optional; 50% or below is comfortable.
Read the 4-cavity, 300-pull row against the 8-cavity, 300-pull row. The 4-cavity plan needs three molds just to reach 3,000 pieces; each would run at 83.3%, so the 70% rule makes it four molds at 62.5%. The 8-cavity plan needs two molds at 62.5% and is already compliant. Both are valid after the rule, but the first buys four tools and the second two. The verdict column describes the minimum plan; the last column is what you order.
The two 3,000-piece rows matter. A 10-cavity mold at 300 pulls covers the order exactly — the worst place to be, because the last hundred pieces come off a tool in its final ten shots, at 97–100% of its rated life. A 12-cavity mold at 500 covers the same order at half its rating.
Apply the 70% rule after the formula, not in place of it: if the minimum plan leaves any mold above 70% of its rated pulls, add molds until every tool is at 70% or below, before the first shot. “Cut a second mold” is not a plan for a 4-cavity, 300-pull design — a second mold reaches only 2,400 pieces, less than the order; the rule produces four molds. Doing it at the start spreads the cost and gives you a known replacement; doing it at shot 250 means waiting on an unpublished lead time while the setting bench empties.
What are the re-cut triggers?
Appearance is the last signal, not the first. Measure the features that decide whether the piece can be set and sold.
| Check | Instrument | Re-cut trigger |
|---|---|---|
| Stone seat position | 40x microscope or projector against the approved sample | Drift beyond ±0.05 mm, or any stone that rocks |
| Piece weight | Scale, same alloy lot | Outside the published ±3% weight tolerance after deflashing |
| Flash line | 10x loupe on the as-cast piece | Visible without magnification, or thicker than the approved sample |
| Parting-line step | Finger and 10x loupe | A step you can feel, or a ridge that survives polishing |
| Cavity surface | Raking light on the wax | Tears, pits or orange-peel texture absent at shot 1 |
| Clasp or hinge | Function test on the casting | Stiffness, slop or a detail that no longer catches |
| Shot count | Mold log | Rated pulls reached — inspect and plan replacement |
Keep the log per mold, not per order: a mold that runs across three orders still has one life. Give the setter written authority to stop the press; a trigger that needs a salesperson’s approval is not a trigger.
A mold that still closes is not a mold that still holds a stone. If the seat has moved, the tool is finished even when every casting looks complete.
Who pays when a mold wears out in the middle of a 3,000-piece run?
Follow the rated count and the cause, not the calendar. The party that controlled the variable that failed owns the risk.
| What happened | Who bears the re-cut | Why |
|---|---|---|
| Fails before rated pulls, design unchanged, master sound | Factory | The consumable was priced into the run; failing short of its own rating is a yield risk |
| Reaches rated pulls mid-run | Buyer | The rated life was consumed; the replacement is a new consumable |
| Overpressure, dropped mold, wrong wax temperature, bad factory storage | Factory | Handling damage is not wear |
| Buyer changes geometry, stone size, ring size beyond range or setting positions | Buyer | A new design needs a new tool; published rules treat geometry changes as a new setup |
| Buyer-supplied master lost or damaged | Buyer | No sound master means no re-cut, only a new master |
| Buyer stores the mold and it deforms | Buyer | The storer controls heat, humidity and compression |
| No rated pulls agreed | Buyer, in practice | No benchmark means a re-cut becomes a change order, not a claim |
Amara’s published terms support this framework: a defensible setup fee covers “CAD file, mold cutting, first pre-production sample”, and “No setup, tooling, or mold fee is payable for any repeat order under this agreement.” Molds and CAD files stay the customer’s exclusively and are stored free for 36 months from the last production date. Read the wording in the setup-fee and reorder policy.
There is a trap in that wording. A mid-run re-cut on a first order is not a repeat order, so the no-tooling-fee promise does not automatically pay for a second mold inside the same 3,000 pieces. It protects the next order, not the current run. If you want the factory to carry a premature failure, the rated life and the re-cut allocation must be written separately.
A mold rated at 300 pulls that fails at shot 280 is short of the factory’s own number: the factory bears the re-cut, with no new CAD or setup fee. One that fails at shot 320 has delivered what it promised, so the buyer bears the replacement — and should have planned it before shot one. The boundary is completion of the 300th pull: 281–299 are inside the rating; once 300 pulls are logged, the rated life has been delivered.
Damage is a separate category. A mold that was over-pressurized, dropped, cooked at the wrong wax temperature or stored in a hot room did not wear out; it was broken, and that cost sits with whoever handled it.
What should the mold-life clause say?
- Rated pulls and cavity count for each mold, with a shot defined as one complete injection cycle.
- A shot log per mold per shift: mold ID, start and end count, setter, wax temperature and pressure.
- The trigger table copied into the annex, with instrument and acceptance value for each check.
- Stop authority for the setter; production restarts only when QC signs the mold back in.
- A fixed re-cut price per mold for the life of the order, quoted before the deposit.
- The allocation table: premature wear, rated-life replacement, handling damage, design change, master loss.
- A spare-mold trigger when the minimum plan leaves any mold above 70% of its rated pulls, with the added mold at the same price and its cavity count written down.
- Re-cut lead time in days. Get a number; Amara’s own sourcing guidance calls mold-cutting time “the one number you cannot infer” — see how a split S925 order is scheduled.
- No double charging. A re-cut from a sound master for an unchanged design carries no new CAD or setup fee, and the worn mold is photographed before scrapping or return.
How should a 3,000-piece schedule absorb a re-cut?
Start with a qualification run. Ask for 50 shots during sampling, measured at 1, 25 and 50, with results attached to the sample approval. Amara publishes 7–10 days for samples, 15–25 days for first bulk orders and 10–15 days for reorders; the run fits inside sampling if it is requested before sign-off.
Split the bulk into three production lots of about 1,000 and inspect at each boundary, so a worn mold contaminates one lot rather than the order. Do not finish all 3,000 early to save a setup: finished-goods storage is not the free 36-month mold storage, as the staged-shipment terms explain.
Put the re-cut window on the schedule before the 30% deposit. If the factory will not quote a mold-cutting time, hold a buffer. A re-cut that lands in the setting and plating queue costs more than the mold; one that was on the schedule costs a day.
Keep the warranty separate from tooling life. Amara’s 12-month anti-tarnish and stone-loss warranty covers the finished piece; it does not extend a consumable mold. Two instruments, two clauses.
Frequently asked questions
Can one rubber mold really cast 3,000 pieces?
Only when cavities × rated pulls is at least 3,000 — and that is capacity, not a plan. A 10-cavity mold at 300 pulls reaches exactly 3,000, a 100% load, so the 70% rule adds a second mold. A 12-cavity mold at 500 covers 6,000 and runs the order at 50% load, comfortable alone. Apply the formula, then the 70% test.
Is one shot the same as one casting?
No. One shot fills every cavity, so a 10-cavity mold yields ten wax patterns. Confusing shots with pieces is the common factor-of-ten error in a tooling quote, and it surfaces only when the first mold runs out.
Does Amara publish a shot count for its rubber molds?
Not on the pages reviewed. It publishes the wear mechanism, three-year free storage and the no-tooling-fee rule, but not a rated life. Ask for the number and make it a PO line.
Should I pay for a spare mold before the run starts?
Yes when the minimum plan leaves any mold above 70% of its rated pulls — and the answer can be more than one added mold. A 4-cavity, 300-pull design needs three molds for the volume and a fourth to bring each to 62.5%; a 10-cavity, 300-pull design needs one mold for the volume and a second to bring each to 50%. The added mold is cheaper than a mid-run stop, and it is qualified from the same master. At 50% load or below, a spare is optional; between 50% and 70%, it is a scheduling choice rather than a capacity requirement.
Who pays if the mold fails in the last 200 pieces?
Count shots, not pieces. On the 10-cavity, 300-pull mold used throughout this article, 3,000 pieces is 300 shots, so the last 200 pieces are shots 281–300. Until 300 pulls are logged, the mold is inside its rating: a failure at shots 281–299, or during the 300th, leaves the factory bearing the re-cut when the design is unchanged and the master is sound. Once 300 pulls are complete, the rated life has been delivered and the buyer buys the replacement. “Last 200 pieces” is not an allocation rule — it straddles the boundary, which is exactly why the PO must allocate by logged shot count.
Does the 12-month warranty cover mold wear?
No. The warranty covers tarnish and stone loss on finished pieces. Mold life is a tooling consumable with its own rated count, trigger table and re-cut clause.
