Buying melee without getting burned

A parcel of melee lands on the bench in a small fold of paper, and the first thing anyone does with it is the one thing the invoice can’t do: tip it out under the light and start counting. The invoice says two carats, so many dollars a carat, done. What actually matters — how many of those stones sit cleanly in a head, match their neighbors, and are still round once the wheel is done with them — gets decided one stone at a time, hours after the money has already changed hands.

Melee is priced like a commodity and set like jewelry, and that gap is where a store’s margin quietly leaks out. Here’s what to actually check before the next parcel comes in, and the moment it does.

Weight is not yield

A parcel is priced by total weight. It’s set one stone at a time. Those are two different units, and the space between them is where the real cost of a parcel lives.

Say you buy a parcel of one-point melee — roughly a hundred stones to the carat — at $600 a carat. Ten carats, $6,000, a thousand stones: $6 a stone, if every one of them goes into a piece. None of them will.

Tip the parcel onto the sorting tray and some stones are out of the running before a bench jeweler has touched a single prong: a chipped culet, a stone barely holding together at the girdle, a few visibly included to the naked eye no matter what the invoice says the grade is. Say that’s a hundred and eighty stones — eighteen percent, as an example — gone before any work has started. Eight hundred twenty are left. $6,000 divided by 820 is $7.32 a stone, not $6.

Now say a second parcel costs ten percent more per carat — $660, or $6,600 for the same ten carats — because whoever sorted it actually pulled the rejects before it was weighed and priced. Say it runs six percent reject instead of eighteen: 940 stones survive. $6,600 divided by 940 is $7.02 a stone — cheaper than the “cheaper” parcel, before counting the hours spent sorting the first one down to something usable.

Cost per usable stone is the only number that tells you what a parcel actually cost. Cost per carat is what’s on the invoice. Ask for both.

What the sieve size doesn’t tell you

Small stones are sorted by sieve size — a stack of mesh screens, each one a fraction of a millimeter tighter than the last — and a parcel is sold “to size” for whichever sieve it’s supposed to sit inside. On paper this is the most exact measurement in the whole transaction. In practice, screens wear and calibration drifts, and a parcel that’s supposed to run tight to one size can spread across two or three.

That spread doesn’t show up looking at the parcel. It shows up one seat at a time. A stone a hair over size won’t drop cleanly into a head cut for that size — the bench either forces it, which stresses the prongs, or opens the seat, which is time nobody quoted for. A stone a hair under size drops in and rattles, or leaves a visible gap next to its neighbors. Multiply either problem by forty stones in an eternity band and it’s not a rounding error anymore. It’s an afternoon.

A parcel that runs a shade wide or narrow of its stated sieve isn’t a defect you catch by eye. It’s a defect you catch one stone, one seat, at a time — which is exactly why it’s worth checking before the whole lot goes to the bench, not after.

Matching hides in a parcel, and shows up in a row

A parcel is graded on average — color and clarity called for the lot as a whole, not stone by stone. That average can be entirely correct and still hide a real spread: a parcel called G to H might genuinely run anywhere from F to I once every stone is looked at on its own. Scattered loose on a tray, that spread is invisible. Nothing next to a stone tells you it’s off.

Set the same stones in a continuous line — a pavé row, a channel, an eternity band — and the spread stops being invisible. Under raking light, one stone half a grade warmer than its neighbors reads as a dead spot in an otherwise lively row, and a customer who couldn’t name a color grade will still see it, even if she can’t say what she’s seeing. Cut drifts the same way: melee isn’t cut to the tolerance a center stone is, and a parcel with uneven makes gives a row uneven sparkle, stone to stone, even when every one of them is technically within grade.

By the time either problem is visible, the stones are already set. The fix is a remake, not a return.

Buying a parcel is not the same purchase as buying a job

Stocking melee for the case is a different decision than buying for a special order, and it’s worth treating them as two separate purchases rather than one habit.

A parcel bought for stock is bought on the average — a good color range, an acceptable reject rate, a price that works across whatever it eventually goes into. That’s the right way to buy for a case you’re filling over months.

A specific job is not average. A client’s eternity band needs forty-two stones, matched to each other, not to a lot average — and a parcel bought the ordinary way won’t hand you forty-two matched stones out of a hundred without a fight. Buying to a job means over-buying on purpose, ordering enough that there’s a real run to cull down to, and pricing the job on what that matched set actually cost — not on the blended price of the parcel it came from. Treat the two the same and the job either loses money quietly or gets set with stones that don’t belong next to each other.

Before you buy, and the moment it lands

Before you agree to a price, even informally, ask for four things:

• The reject allowance. What percentage of this parcel you should expect to lose to chips, breaks, and outliers — and whether the price already accounts for it.
• The real sieve range. Not just the size it’s labeled — a parcel tight to a quarter-sieve is worth more than one that spreads across a full size, even at the same average weight.
• The color and clarity range. Not just the grade called for the lot as a whole.
• What happens if it doesn’t check out. A straight exchange, a credit, or nothing — find out before you need the answer.

Then, the moment it lands: weigh it against the invoice. Run a real sample — not a handful, a proper count — through a sieve and under a loupe before committing the whole lot to a job. Count what’s actually usable before counting what it cost. The parcel that looks cheapest on the invoice is, routinely, the one that costs the most once the bench has fought it stone by stone, and the only way to know which one you bought is to check before it’s already set.

This is exactly the part of the trade Gemercel takes on directly: it supplies the diamonds, melee included, so a store isn’t sorting a stranger’s parcel alone at the bench every time a shipment lands. If today’s arithmetic sounds familiar, put your store on the list.