No ISO Standard for Loose Stones: A Stone-Security QC Spec and Test Record Template for Your S925 Moissanite Factory

No. No ISO, national, or industry standard defines a mandatory pull test or ultrasonic test for stone security on S925 moissanite rings before shipment. Every factory writes its own internal claw-check routine, which means the same phrase — “pull-tested” — can describe very different checks. What you should ask your factory to record is a signed per-lot stone-security test record built on four defined checks: a probe test on every stone of every sampled piece with the result zero movement; a 10× visual inspection; one ultrasonic cycle with stated frequency and duration and zero stones lost; and a destructive pull test on agreed samples that records peak force in newtons, pull direction, and failure mode. This article gives you the spec language, the AQL sampling row, and the record template to paste into your purchase order.

Is There a Standard Pull or Ultrasonic Test for Stones Before Shipment?

No. Neither ISO nor any major national standards body publishes a mandatory stone-retention test for set jewelry. What exists publicly is either generic — gemological-care guidance, sampling mathematics like ISO 2859-1 — or factory-specific. The force values and cleaner settings that circulate in buyer forums are references, not pass/fail criteria that a factory can be audited against. In practice, “we pull-test every stone” means whatever the factory decided it means: a fingertip flick for one bench, a calibrated gauge for another.

Amara Jeweiry is open about this. Our homepage states that every piece is 100% claw-checked and pull-tested, that micro-pavé work is set under 40× magnification, and that final inspection runs to an AQL 2.5 plan — but those are our internal definitions, backed by a 12-month stone-loss warranty. Until the definitions are copied into your own QC protocol, “pull-tested” in an email is a claim, not a specification you can enforce or claim against.

That is the gap this article closes. Below is a stone-security acceptance specification and test record template written at the level of detail a purchase contract needs: the probe tool, force, and direction; the 10× inspection points; the ultrasonic cycle and its prohibited cases; the AQL sampling row; the destructive pull test fields; and the “zero movement” acceptance wording.

What Must a Stone-Security Spec Define?

Four checks, each with a tool, a parameter, a direction, and a recorded result. Any check you leave undefined defaults to the factory’s habit, not your requirement. Freeze the bracketed values in this section against a pre-production sample before mass production — the same discipline you already apply to the pre-production sample itself.

Check 1 — The Probe Test: Tool, Force, Direction

• Tool: a pick with a rounded tip of roughly 1–1.5 mm in wood or hard plastic. A steel needle is only acceptable under magnification on melee, because a carelessly placed steel tip can score plating and create a claim of its own.
• Force: start at 1–2 newtons (about 100–200 grams-force) — firm, controlled pressure. Verify once on a small jewelry scale so the whole bench calibrates to the same feel.
• Direction: push each stone tangentially around the girdle at three to four positions — this is the direction a prong actually lifts — plus one outward lift attempt under the table edge on prong settings. Two full circuits per stone.
• Recorded result: pass means zero perceptible movement. A rock, rattle, or click on any stone marks the piece as a critical defect, not a repairable major.

Check 2 — The 10× Visual Inspection Points

The probe tells you whether a stone moves; the loupe tells you why it might move later. At 10×, before the probe, the inspector confirms:

• every prong tip sits fully over the girdle with even contact and no visible gap when the piece is tilted to the light;
• seats are free of polishing residue that would let a stone settle after wear;
• no glue anywhere — adhesive as a setting aid is an automatic reject on moissanite rings sold as mechanically set;
• channel walls are flush and bezel rims are continuous, with the stone seated level;
• prongs are straight and undamaged — polishing machines are a known cause of bent prongs, and the distinction between a setting flaw and a plating flaw changes who reworks the piece, as covered in our bent-prongs article.

Check 3 — Ultrasonic Screening: Parameters and Prohibited Cases

Define the cycle in the spec so “ultrasonically cleaned” is reproducible: a stated frequency (40 kHz is the common default), a stated duration (3–5 minutes), a mild alkaline jewelry solution at warm temperature, pieces in a basket — never loose in the tank — and the batch identified on the record. Acceptance result: zero stones lost.

Prohibit the cycle for: any glue-set stone; pieces with resin, enamel, or lacquer components; mixed-material designs containing pearls, opals, coral, amber, or other organic or treated stones; and any piece that already failed the probe test (it goes to the bench, not the tank).

One sentence belongs in your contract and your head: a stone that survives ultrasonic cleaning is not proven tight. Ultrasonic is a screen that evicts already-marginal stones before your customer’s sink does. The acceptance evidence is the probe and pull-test record, not the fact that a bath happened.

Check 4 — The Destructive Pull Test: Force, Direction, Failure Mode

• Samples: agreed pieces per style per production lot — three is a workable starting point — pulled from QC-passed stock, so you are measuring what would have shipped.
• Equipment: a digital force gauge, calibrated within the last twelve months, with its ID and calibration date written on the record. Grip the stone with soft nylon jaws or a sling so the reading reflects the setting, not jaw slip.
• Direction: pull at roughly 45° to the girdle plane, toward the weakest prong — the catch direction. Slow and steady; a jerk measures your wrist, not the setting.
• Force values: these are yours to negotiate, not law. Draft starting points for the first trial round: ≥20 N (≈2 kgf) peak before release for a 5–6.5 mm solitaire, ≥10 N for a bezel, ≥5 N for pavé melee. Raise or lower them after you see the first destructive results on your own design.
• Failure mode to record: prong tip opened, prong bent, bezel rim deformed, channel wall broken, seat solder failed, or stone fractured. The preferred outcome is that metal yields before the stone releases — a stone shattering at low force tells you as much about stone quality as about the setting.

Which AQL Sampling Row Should Stone Security Use?

Stone security is an attribute check — a stone moves or it does not — so the sampling framework is ISO 2859-1, Sampling procedures for inspection by attributes — Part 1: Sampling schemes indexed by acceptance quality limit (AQL) for lot-by-lot inspection, reissued as the January 2026 edition and available through national standards bodies such as DIN Media. Use General Inspection Level II as the default severity.

Classify the defect classes like this in your protocol:

The accept/reject numbers above are the standard’s published table values for that sample size, but confirm them against the current edition’s table for your actual lot size before freezing the row. For where this gate sits in the sequence — pre-plating versus post-plating — see our two-gate AQL plan for S925; the stone-security gate belongs at the end, after plating, before packing.

The Per-Lot Test Record: What Result Should the Factory Write Down?

This is the actual answer to “what result should I ask the factory to record”: numbers and enumerated outcomes, signed per lot. Adjectives — “tight”, “secure”, “good quality” — are not results. The template:

Require the signed scan to travel with the pre-shipment evidence pack, alongside the batch video taken before balance payment — the video shows that stones are present and seated; only the record shows what force they survived.

The Acceptance Clause: “Zero Movement” Wording for Your PO

Stone security acceptance — zero movement. Before packing, the Supplier shall probe-test every stone on each sampled piece per the attached Stone-Security QC Specification, using a [wood/plastic] pick of [1–1.5 mm] tip at [1–2] N, applied tangentially around the girdle and in an outward lift direction. Any stone exhibiting perceptible movement is a critical defect; acceptance requires zero moving stones in the sample. The Supplier shall run one ultrasonic cycle at [40] kHz for [3–5] minutes on sampled pieces with zero stone loss, and shall pull-test [3] pieces per style per production lot, recording peak force in newtons, pull direction, and failure mode per specimen. A signed Stone-Security Test Record shall be delivered for each production lot before the balance payment is released.

Fill every bracket at the pre-production sample stage, after the first destructive trial — a threshold neither side has measured is a dispute neither side can win. Factories that already run 100% claw-check and pull-test routines, as we state for our own production, adapt to a written spec quickly; the record simply gives you their result in your units.

Soalan Lazim (FAQ)

Is a batch video enough instead of a test record?

No. Video proves which stones were present and roughly how they were seated at a moment in time. It cannot prove tightness under load, and it carries no force values, no direction, no failure modes. Ask for both: the signed record for the numbers, the video for the visual state of that exact lot.

The factory says pull-testing is not standard practice. Now what?

Ask them to run the first destructive trials on your pre-production samples and share the readings, then freeze thresholds together. That is a one-time cost measured in a few rings. A factory that refuses any measured pull test is telling you it has no stone-retention data for its own process — which is exactly the situation that ends in a chargeback instead.

How many pieces per lot must be destroyed for the pull test?

Three per style per production lot is a defensible starting point; five for a new mold or a new setting supplier. Pieces are drawn from QC-passed units. Against the cost of one consumer stone-loss claim — and the review that follows it — three rings per lot is the cheapest insurance in the agreement.

Does ultrasonic cleaning itself loosen stones?

On a sound setting, properly set, it should not — which is precisely why it works as a screen: it evicts stones that were already marginal. That is also why the record must state zero stones lost, and why a stone found in the tank stops the lot. Keep the prohibited list firm: glue-set stones, resin or enamel components, and organic or treated stones never enter the cycle.

Should stone security be tested before or after plating?

After. Plating baths, rinsing, and tumbling all move marginal stones, so a probe test that passes pre-plating proves nothing about the piece that ships. Run the 10× and probe gate after plating and before packing, as the final gate in your two-gate inspection plan.