Stage one
Aggregation and cleaning
Collar, survey, lithology, and assay tables are joined and checked for overlaps, gaps, and impossible values. The cleaned database is the only input the estimate is allowed to see.
Technical work
Metrics that hold up underground
We do not stop at a spreadsheet average. The useful signal in a deposit is usually the variation, and that is what the model has to carry.
See the workflow
How the numbers are handled
Raw drill-hole files are rarely consistent. Units, detection limits, missing intervals, and duplicate assays are reconciled before any estimate is run. Averages that ignore that work look tidy and fail in reconciliation.
The same dataset then moves into a multi-domain model. Domains follow geology first. Statistics are used to test the domains, not to invent them.
Assay validation
Geostatistics
Block modeling
From data to a model you can defend
Three stages. Each one has a written check before the next begins.
Statistical checks
Declustering, capping, and variography are documented. Kriging or another suitable estimator is chosen because it fits the data, and the choice is written down so an auditor can repeat it.
Yield scenarios
Grade and tonnage are tested against cutoff and selectivity assumptions before a mine plan depends on them. The output is a range, with the assumptions attached, not a single unsupported number.
Reviewed by the people who use the models
The validation steps on our copper dataset are now the internal standard. The bias we had been carrying in the composites was obvious once it was plotted.
Raushon J.
Grade reconciliation had drifted for months. The audit of the block model found a domain boundary our site team had stopped questioning.
Arshell Bosie
Senior Resource Geologist
The data trail survived an external audit without a missing table. That is the reason we kept the same workflow for the next study.
Mouse Alidaei