New assay technologies rarely stall on the instrument. They stall on confidence. For the people who estimate and sign off on resources and reserves, replacing a century-old method raises the questions that matter most: is the data at least as reliable, how is that demonstrated, and what does the change mean for the estimate?
This presentation tells the story of how OceanaGold's Macraes operation in New Zealand replaced fire assay with on-site Chrysos PhotonAssay, through interviews with four members of the site team: the Geology Superintendent, the Geology Manager, the Continuous Improvement Manager and an Exploration Geologist. Each saw the change from a different seat (production grade control, resource strategy, the business case, and day-to-day handling of results), and together they give a candid account of the full adoption journey, from early scepticism to "no regrets."
Macraes is a high-tonnage, very low-grade gold system: reserve grade near 0.25 g/t, cut-offs of 0.2–0.3 g/t and tails near 0.1 g/t. At these grades, ore and waste are separated by a few tenths of a gram, so sampling error and assay precision bear directly on classification and on the reserve. PhotonAssay analyses a 500 g charge reached through fewer volume-reduction steps than the 30–50 g fire-assay aliquot, reducing the fundamental sampling error that dominates at low grade. The team describes how validation was designed: side-by-side testing focused on the sub-0.3 g/t interval, open treatment of a roughly 10% discrepancy at tails grades, and, decisively, three to four months of stable mine-to-mill reconciliation before fire assay was decommissioned. They also recount a negative reconciliation that coincided with the switchover, and the discipline it took not to blame the newest change when several variables had moved at once.
The operational effects feed straight back into resource work. Grade-control turnaround fell from a 48-hour target that often ran to 50–60 hours to under 24 hours. Exploration results that once took weeks to months now return in about four days, easing pressure on resource-model deadlines. Refires and lost or disposed samples have disappeared from the QA/QC workflow, with standards checked before results reach the database. Sampling has risen from about 20,000 to 28,000 samples per month, increasing the data density available for estimation.
The talk closes on what estimators and QPs elsewhere can take from Macraes: how to build a validation programme that will stand up to review, why reconciliation should be the final arbiter, what evidence a QP needs to support a change of assay method in public reporting, and where new options (smaller drill diameters, multi-cycle precision and multi-element data) may change how resources are drilled and estimated.