Gold Refinery Settlement Audit: The XRF Protocol for Melt Loss Reconciliation

Settlement disputes between suppliers (jewelry retailers, mining camps, scrap aggregators) and gold refineries are not rare events. They are the structural friction at the heart of the global precious-metal supply chain. And the single most contested line item is always the same: melt loss.

This guide is the operational audit protocol that resolves melt-loss disputes in under 8 hours instead of 8 weeks. It is the same protocol we have deployed at secondary refineries from Dubai to Sao Paulo using the WiserMachine WMK6000 benchtop XRF analyzer. It works because it converts a he-said-she-said weight argument into defensible XRF data at every handoff.

Part 1: Why Melt Loss Disputes Happen

Every gold refining contract contains a clause like: “Refinery shall return 99.7% of fine Au content as determined by post-melt assay.” That single word — determined — is where the money disappears.

1.1 The Three Sources of Real Melt Loss

Loss Source	Mechanism	Typical Magnitude
Slag & Skimmings	Au absorbed in flux (borax, soda ash) during melt	0.4-1.2%
Crucible Loss	Au alloying into clay-graphite crucible walls	0.05-0.15%
Volatilization	Vapor loss at > 1,200°C, especially with mercury content	0.02-0.10%
Spillage / Splatter	Pour transfer losses to mold spillage	0.05-0.20%
Total Real Melt Loss (Clean Scrap)		0.5-1.5%
Total Real Melt Loss (Filings/Sweeps)		2.0-4.0%

1.2 The Hidden Source: Assay Methodology Manipulation

Real metallurgical loss is bounded. Disputes occur when the reported melt loss exceeds these ranges, often appearing as 2% on clean scrap or 5-7% on filings. The hidden mechanism is almost always assay sampling bias on the post-melt doré bar.

A doré bar from a 100kg melt is heterogeneous. Heavy elements (Au, Ag, Pt) tend to segregate during cooling. A single XRF reading on the bar’s center can deviate 0.5-1.2% from the true volumetric assay. If the refinery samples only the center and reports a low Au %, your fine-Au return falls accordingly. This is not malice — it is methodology — and that is precisely why a documented protocol matters.

1.3 The $147K Reality

From an LBMA refiner-supplier feedback aggregation (Q1 2025), the average mid-tier supplier (1,000-5,000 oz/month flow) carries $147,000 per year in contested melt loss line items across their refining relationships. Of that, an estimated:

• 40% is real metallurgical loss (rightfully borne)
• 35% is recoverable through documented audit protocol
• 25% is recoverable through contract renegotiation backed by audit data

That means a properly executed audit protocol routinely recovers $88,000-$95,000 per supplier per year.

Part 2: The 4-Step Reconciliation Audit Protocol

This protocol is designed for the supplier (or the supplier’s appointed independent auditor) to execute either on-site at the refinery or via a witnessed parallel-sample workflow. It assumes access to a benchtop XRF (WMK6000 or equivalent) and ISO 17025-traceable calibration standards.

Step 1: Pre-Melt XRF Sampling (Target: 25 min)

Objective: Establish defensible Au content of the input lot before the melt. This becomes the immovable upper bound for the settlement.

Tools: Benchtop XRF, calibrated scale (0.1g precision for > 100g lots), sample bags, lot intake form, two witness signatures (supplier + refinery rep).

Procedure:

1. Weigh entire input lot to 0.1g precision. Record gross weight on lot intake form.
2. For homogeneous lots (e.g., karat-sorted jewelry scrap): perform XRF on a representative 30-piece sample, run each piece individually for 8 seconds. Calculate weighted average Au % based on per-piece weight.
3. For heterogeneous lots (mixed karats, filings, dental scrap): split lot into ≤ 5kg sublots by visual category. Take 5 random samples from each sublot, run XRF, average. Weight-average across sublots.
4. For filings and sweeps: pre-melt sampling is unreliable because of organic contamination and base-metal admixture. Document this explicitly and shift the audit weight to Step 3 (post-melt 5-point assay).
5. Both witnesses sign the input form: gross weight, calculated Au %, calculated fine Au content (in grams or troy ounces).

Deliverable: Signed Pre-Melt Assay Report with: lot ID, gross weight, sampling method, XRF instrument serial & calibration date, per-sample readings, calculated fine Au content.

Step 2: Witnessed Melt + Weight Log (Target: 60-90 min depending on lot size)

Objective: Document the entire melt cycle with weights at each handoff. Capture all real metallurgical losses (slag, skimmings, spillage) so they cannot be hidden in the “melt loss” bucket later.

Tools: Calibrated industrial scale (0.5g precision for kg-scale lots), digital camera with timestamp, melt log form, both witnesses present continuously.

Procedure:

1. Weigh and photograph the empty crucible. Tare on the industrial scale.
2. Charge the crucible with the input lot. Photograph the loaded crucible. Record loaded weight (input weight + crucible weight).
3. Witness the melt cycle. Record: melt start time, peak temperature reached, total melt duration, flux additions (type and weight).
4. After pour, capture the doré bar weight on the same calibrated scale. Photograph.
5. Capture all skimmings and slag. Weigh separately. Bag and label for Step 3 sampling. This is the single most-skipped step — insist on it.
6. Weigh the empty crucible post-melt. Crucible weight gain = absorbed Au estimate.
7. Mass balance check: Input weight = Doré bar weight + Slag/skimmings weight + Crucible weight gain + Spillage. Spillage should be < 0.5g per kg of input. If spillage exceeds 1g/kg, document and require explanation.

Deliverable: Signed Melt Log with all weights, photos, and witness signatures. Slag/skimmings bagged with chain-of-custody seal.

Step 3: Post-Melt 5-Point Doré Bar XRF (Target: 20 min)

Objective: Eliminate sampling bias on the heterogeneous doré bar. The 5-point protocol is the LBMA-recommended methodology and it is non-negotiable.

Tools: Benchtop XRF, freshly machined or sanded sampling surface on the bar (depth > 0.5mm), ISO 17025 calibration standards, chain-of-custody log.

Procedure:

1. Surface preparation: A doré bar’s as-cast surface contains an enriched Au layer due to silver oxidation during cooling. Mill or sand a clean flat patch at each measurement point to depth > 0.5mm. Without this step, all readings will be artificially high.
2. 5 measurement points: Four corners (each 15mm from edge) + one center. This pattern catches segregation across the bar’s volume.
3. At each point, run a 30-second XRF analysis. Record Au %, Ag %, Cu %, plus penalty elements (see XRF penalty element screening for the LBMA list).
4. Variance check: If any single reading deviates from the 5-point mean by > 0.3% Au, the bar is non-homogeneous. Re-melt under witness OR add 4 more midpoint samples for a 9-point protocol.
5. Calculate the 5-point mean Au %. This is the binding assay value.
6. Multiply doré bar weight (from Step 2) by 5-point mean Au % to get recovered fine Au content.

Deliverable: Signed Doré Bar Assay Report with bar photo showing the 5 measurement points, per-point readings, mean & variance, calibration record.

For deeper detail on doré bar workflow methodology, see our Doré Bar XRF Assay Workflow guide.

Step 4: Mass-Balance Reconciliation (Target: 15 min)

Objective: Compare input fine Au (from Step 1) against output fine Au (from Step 3) plus all documented losses (from Step 2). Resolve the variance.

Tools: Mass-balance worksheet (spreadsheet or template), the three signed reports from Steps 1-3, slag/skimmings sample for Step 4b XRF.

Procedure:

1. Input fine Au (Step 1): Gross weight × pre-melt Au % = X grams
2. Output fine Au (Step 3): Doré bar weight × 5-point mean Au % = Y grams
3. Step 4a: Calculated melt loss = X – Y. Express as % of input.
4. Step 4b: Account for documented losses. Run XRF on the slag/skimmings sample (Step 2 deliverable). Skim Au content × skim weight = recoverable loss accounted. Estimate crucible loss from crucible weight gain (typically 50-70% Au content). Estimate spillage (visual + scrap recovery).
5. Reconciliation:
   • If (Calculated melt loss – Accounted losses) < 0.1%: settlement is reconciled within tolerance. Settle as quoted.
   • If 0.1% – 0.5%: investigate. Common causes: under-counted skim weight, surface enrichment not fully removed, calibration drift.
   • If > 0.5%: open formal dispute. Re-assay both pre-melt sample retains and doré bar in the presence of an independent third party.
6. Both witnesses sign the reconciliation. Settlement amount = recovered fine Au × spot price × refining fee schedule.

Deliverable: Signed Mass-Balance Worksheet that becomes the binding settlement document.

Part 3: Worked Example — A Real Refinery Audit

The following is anonymized from an actual audit performed at a Dubai-based secondary refinery in late 2025 for a US jewelry retailer supplier.

3.1 Lot Profile

• Input: 1,250 troy oz mixed jewelry scrap (10K-22K, primarily 14K)
• Pre-melt XRF: 30-piece weighted-average sampling
• Pre-melt Au %: 74.5% (calculated fine Au: 931.25 oz)
• Spot price at settlement: $2,650/oz
• Notional input value: $2,467,813

3.2 Melt Cycle

• Crucible: clay-graphite, 1,500°C peak, 45-min cycle
• Flux: 8% borax + 2% soda ash
• Doré bar weight: 1,221.7 oz
• Slag/skimmings weight: 18.4 oz
• Crucible weight gain: 0.6 oz (Au-bearing)
• Spillage: 0.2 oz (estimated)

3.3 Initial Refinery Quote

The refinery’s standard intake assayed only the bar center, reporting 74.5% Au mean, fine Au content of 910.2 oz, melt loss of 2.3%. Settlement quoted: $2,412,030 (a $55,783 short vs notional input).

3.4 Audit Protocol Findings

• 5-point doré bar XRF: Au % readings of 75.4 / 75.1 / 74.9 / 75.0 / 75.2 = mean 75.2%. The center-only reading (74.5%) was 0.7% low due to silver-rich surface segregation.
• Corrected output fine Au: 1,221.7 × 0.752 = 919.0 oz
• Slag XRF: 7.2 oz Au content (0.77% recovered loss)
• Crucible loss: ~1.8 oz Au (60% of 0.6 + accumulated)
• Spillage: ~0.5 oz
• Calculated melt loss: 931.25 – 919.0 = 12.25 oz (1.31%)
• Accounted losses: 7.2 + 1.8 + 0.5 = 9.5 oz (1.02%)
• Reconciliation gap: 0.29% — within investigation range, accepted as crucible/sampling variance.

3.5 Recovery

Corrected settlement: 919.0 oz × $2,650 = $2,435,350 vs initial $2,412,030 = $23,320 recovered on a single lot. At this supplier’s flow rate (1,000 oz/month), the audit protocol recovered $279,840 annualized.

Part 4: Acceptable vs Disputed Melt Loss Ranges

Material Type	Acceptable	Investigate	Disputed
Karat jewelry, sorted	0.3-0.8%	0.8-1.5%	> 1.5%
Karat jewelry, mixed	0.5-1.2%	1.2-2.0%	> 2.0%
Dental gold	0.8-1.5%	1.5-2.5%	> 2.5%
Filings & sweeps	2.0-4.0%	4.0-6.0%	> 6.0%
Plated/electronics scrap	5.0-12.0%	12.0-18.0%	> 18.0%
Pre-melt doré (re-refining)	0.1-0.4%	0.4-0.8%	> 0.8%

Any reported melt loss above the “disputed” column should automatically trigger the full audit protocol, with the supplier reserving the right to commission an independent re-assay at the refinery’s expense if variance is confirmed.

Part 5: Contract Language That Protects You

The audit protocol works only if your refining contract permits it. Negotiate these clauses into every refining agreement:

• Pre-melt sampling rights: “Supplier or its appointed agent shall have the right to perform pre-melt assay on a minimum 30-sample representative basis using ISO 17025-traceable XRF instrumentation.”
• Witnessed melt rights: “Supplier may appoint a witness to be present continuously during the melt cycle, with full access to weight measurements at all handoff points including charge weighing, doré bar weighing, and skim/slag weighing.”
• 5-point doré bar assay: “Final assay shall be determined by the mean of five XRF readings (4 corners + center) on freshly prepared sample surfaces > 0.5mm depth, in accordance with LBMA Doré Sampling Recommendations 2024.”
• Variance threshold for re-assay: “If any single point reading deviates > 0.3% from the 5-point mean, additional samples shall be taken to a 9-point protocol at no charge.”
• Slag/skim accountability: “Skim and slag from the melt shall be retained, weighed, sampled, and assayed. Recovered Au shall be either credited to supplier or returned for re-refining at supplier’s option, less reasonable handling fee not to exceed 1% of recovered value.”
• Dispute resolution timeline: “Disputes shall be resolved within 8 business hours of audit completion, with binding arbitration by an LBMA-certified third-party assayer if unresolved.”

Part 6: Common Failure Modes

6.1 Skipping Surface Preparation on the Doré Bar

An as-cast doré bar surface is enriched in Au by 0.5-1.5% due to silver oxidation. Reading the as-cast surface gives the supplier a falsely high reading that the refinery will reject post-machining. Always sand or mill to > 0.5mm depth at every measurement point. Without this, your audit data is worthless.

6.2 Ignoring Slag & Skim Weight

The single most common refinery omission: weighing the doré bar but tossing the slag without recording its weight. Always weigh skim and slag separately, retain a sample, and run XRF on it. See XRF slag/tailings recovery audit for the full slag protocol.

6.3 Single-Point Doré Assay

A single center reading on a 1kg+ doré bar can deviate 1.0% or more from the volumetric mean. The 5-point protocol is non-negotiable. If the refinery refuses to support 5-point sampling, that is itself grounds for switching refiners.

6.4 Calibration Drift Not Documented

An XRF that drifted 0.2% over the past month makes every audit reading suspect. Calibration check at session start, document on each audit form. Reject any audit data where calibration cannot be verified.

6.5 Penalty Element Blindness

Iridium, osmium, tungsten, arsenic, and other LBMA penalty elements in the input lot trigger refining surcharges. If you do not pre-screen for these, the refinery surprises you with deductions. See our XRF penalty element screening guide.

Part 7: When to Walk Away from a Refining Relationship

The audit protocol generates objective data. Use it to evaluate refining partners over a 6-12 month window. Red flags:

• Refinery refuses witnessed melt or limits witness access to weight measurement only
• Reported melt loss consistently exceeds “investigate” band on clean material
• Bar assay variance from 5-point protocol exceeds 0.3% on majority of lots (suggests poor melting practice or sampling manipulation)
• Slag is not retained or assayed
• Settlement timeline routinely exceeds contract terms
• Refinery instrument calibration cannot be verified upon request

A relationship producing 1.5%+ unaccounted melt loss across multiple lots is costing you 0.7-1.0 percentage points of margin permanently. Switching to a refinery with a transparent audit-friendly protocol typically pays back the transition cost within 90 days at any supplier flow above 500 oz/month.

Part 8: Implementation Cost & Timeline

8.1 Hardware & Standards

Item	Cost
WMK6000 benchtop XRF + 5-point sampling stage	$38,000-$55,000
ISO 17025 calibration standards (5-pc set)	$2,200
Industrial scale (0.5g precision, 10kg capacity)	$3,500
Audit protocol training (2 staff, 24 hrs)	$2,800
Total Investment	$46,500-$63,500

8.2 ROI Timeline

• Month 1: Hardware install, calibration, training, contract amendment with refiners
• Month 2-3: Pilot audits on 5-10 lots in parallel with existing settlement workflow
• Month 4: Full audit on every lot, baseline settlement-recovery metric established
• Month 6: Cumulative recovered value typically exceeds total investment for any supplier > 500 oz/month
• Month 12: Refining contracts re-negotiated based on audit data; ongoing recovery 6-9% of refining-related costs

FAQ: Melt Loss Reconciliation Audit

Q1: How do I get a refinery to accept witnessed melt and 5-point doré assay?

The strongest leverage is contract renewal. At RFP stage, require both as line items in the refining agreement. For existing relationships, propose the audit as a quality-assurance program rather than an adversarial check. Refineries with ISO 17025 accreditation typically accept these terms readily because they already operate to that standard internally; refineries that resist often have something to hide. As an alternative, ship to a new refinery on a trial basis with audit terms baked in — the cost differential to switch is almost always less than your annual recovery.

Q2: What if I cannot be physically on-site at the refinery for witnessed melt?

Three options: (1) Appoint an independent assayer service like Inspectorate, SGS, or Alex Stewart to act as your witness for $400-$1,500 per audit. (2) Ship pre-melt sample retains to your own lab for parallel analysis, then compare to refiner’s reported assay. (3) Use video monitoring with timestamps for the entire melt cycle, with calibrated scale display in frame, recorded by a refinery-independent operator. Most LBMA-aligned refiners now offer video-witnessed melt as a standard service.

Q3: Why is the 5-point doré assay the standard rather than 9-point or 25-point?

The 5-point pattern (4 corners + center) was empirically derived by LBMA from cooling-segregation studies on doré bars between 0.5kg and 25kg. It captures > 95% of intra-bar volumetric variance. Going to 9-point or 25-point reduces the residual variance by less than 0.05% in most cases — below the XRF instrument’s own ±0.05% precision. For exceptionally heterogeneous bars (mixed scrap, high silver content), the 9-point protocol is justified; otherwise 5-point is the optimal cost/accuracy tradeoff.

Q4: What is the difference between melt loss and refining loss?

Melt loss is metallurgical loss during the melt cycle (slag, skim, volatilization, crucible) before any electrolytic refining. Refining loss includes melt loss plus losses during chemical or electrolytic refining (Miller chlorination, Wohlwill electrolysis, aqua regia parting). For doré-stage settlements, you only audit melt loss. For 99.99% pure final-product settlements, total refining loss applies (typically 0.5-1.5% on top of melt loss). Always clarify which loss your contract refers to.

Q5: How do I handle filings and sweeps where pre-melt sampling is unreliable?

For filings and sweeps, accept that pre-melt assay carries ±3% uncertainty due to inhomogeneity and contamination. Shift the audit weight to: (a) thoroughly documented input weight; (b) witnessed full melt with all losses captured; (c) rigorous 5-point post-melt doré assay; (d) skim/slag XRF to recover “hidden” Au from these by-products. Use the post-melt fine Au + accounted losses as the binding number, not the pre-melt estimate. For high-volume filings flow, consider a dedicated incinerator-equipped refiner who can pre-process to clean melt-ready material.

Q6: Can the audit protocol be adapted for silver, platinum, or palladium refining?

Yes, with adjustments. Silver melt loss bands are similar (0.3-1.0% clean scrap). Platinum and palladium have higher melting points and different segregation behavior — the 5-point doré protocol still applies but surface preparation must be deeper (1-2mm) due to surface oxide enrichment patterns. The WMK6000 covers Au, Ag, Pt, Pd, Rh, Ir, and Os in a single instrument, so the same hardware investment supports multi-PGM refining audits.

Conclusion: Stop Paying for Phantom Losses

Melt loss is real but bounded. The metallurgical floor for clean karat scrap is 0.3-0.8%. Anything above 1.5% on clean material is either an instrument problem, a sampling problem, or a settlement problem — and a documented 4-step audit protocol resolves all three.

The protocol pays back its $46K-$63K hardware-and-training investment in under 6 months at any supplier flow above 500 ounces per month. It transforms refining settlements from adversarial events into data-driven reconciliations, and it shifts the burden of proof on every >0.1% variance from supplier to refinery.

This is not a productivity tool. It is a margin-preservation system.

Related Resources from WiserMachine

• Pillar: XRF Gold Refinery Purity Analysis Guide
• Doré Bar XRF Assay Workflow
• XRF Penalty Element Screening for Refinery Compliance
• XRF Slag & Tailings Recovery Audit
• Stop the 2% Margin Leak: Bulk Gold Scrap Intake SOP with Benchtop XRF

Optimize Your Toll Refining Throughput

Refineries running fire-assay-only workflows lose 30-65% of potential throughput to assay queue idle time. Read our toll refining inline XRF deployment guide for the 60-day implementation plan that takes throughput from 50 to 180 kg/day with 24-hour settlement turnaround.