Practical writing on modern resource practice.

In every audit conversation we have with Competent Persons, the same worry surfaces: ‘Can I re-run the numbers that support this signed statement, exactly as they were, two years from now?’ For most teams the honest answer is no — the CSV moved on the drive, the software version was patched, the spreadsheet tab was renamed. The signed statement is defensible today; the workings are not.

SRES treats reproducibility as a first-class engineering promise. Every kriging run, every pit optimisation, every CPR export is snapshotted with the exact inputs that produced it — assay tables, variogram parameters, density map, cut-off strategy — and pinned inside the tenant so any past run can be re-downloaded byte-for-byte, months or years later.

The consequence is subtle but powerful. Because reproducibility is cheap, teams stop treating a signed resource statement as a fragile artefact. They iterate more, they domain more thoroughly, they explore more cut-off sweeps — because there is no operational cost to producing a defensible new snapshot. Rigour compounds.

For teams graduating from inverse-distance estimation, Ordinary Kriging often feels like a step off a cliff. The literature is dense, the parameters are unfamiliar, and the software historically buried the workings behind opaque dialog boxes. In practice, defensible OK is much less mysterious than it looks.

The heart of the discipline is the variogram — a summary of how similar two samples are as a function of the distance between them. SRES computes an experimental variogram from your assays, then fits a spherical, exponential or Gaussian model with a small set of geologically-interpretable parameters: nugget, sill and range. Every fit is a chart you can eyeball, share, and defend.

Composite lengths, domaining and search neighbourhoods complete the picture. SRES exposes each choice as an explicit knob — not a hidden default — and records what you picked into the CPR narrative that ships with your final PDF. There is nowhere for a bad assumption to hide.

Deterministic NPV numbers are neat but dangerous. They anchor the board on a single figure that ignores the ranges of commodity price, mining cost, processing recovery and dilution that drive real project economics. Monte-Carlo NPV — with even a modest 5,000-trial run — restores the honest picture.

SRES defaults to triangular distributions because they are transparent to non-statistical readers: a minimum, a most-likely and a maximum. Every simulation captures the full distribution, the P10/P50/P90 percentiles, and a histogram you can drop straight into the CPR.

The trick is not the maths — the trick is the discipline of disclosure. SRES pins the seed, the trial count and the input distributions into the report so a reviewer can replay the simulation, byte-for-byte, and verify the numbers were not cherry-picked.

The client — a mid-cap iron ore developer — arrived with 112 drillholes, four CSVs (Collar, Survey, Assay, Geology) and a deadline: a JORC-aligned CPR PDF for the following morning's board. Their previous provider needed three weeks and a two-week onboarding.

In SRES, the ingest step is measured in minutes. Column alias auto-mapping picked up BHID, easting, from/to and the Fe grade column without manual configuration. Validation flagged three geometry overlaps and a single out-of-hole assay; all resolved before lunch.

By late afternoon the block model was kriged, the pit shell optimised at three cut-off scenarios, and the CPR PDF assembled with narrative, figures and the CP's signature block. The signed statement went to the board with the reproducible snapshot pinned to the workspace — and the client kept it.