NI 43-101 Technical Report
Interpretation and Breakdown

The Ontario Securities Commission convened the Mining Standards Task Force afterward. NI 43-101 entered force in 2001. Every section of the report traces back to a gap that Bre-X exposed: the QP requirement to the absence of independent oversight, the QA/QC disclosure to the fact that tampered samples passed through an analytical pipeline without triggering alarms, the data verification section to the fact that nobody outside the company checked anything.

The report has 27 prescribed sections. Most of them can be skimmed in minutes. Sections 2 through 5 cover property description, accessibility, physiography, climate, infrastructure. Compliance material. Section 8 is deposit type classification, usually a page. Section 23, adjacent properties, is usually filler. Section 27 is the reference list. These exist because the regulation says they must. They contribute almost nothing to interpretation.

The sections that matter are 6, 7, 9 through 14, and 25/26. The economic sections (15 through 22) matter more at the PFS and FS stages than at the PEA stage. Section 1, the summary, is a special problem discussed further down.

Top-cutting gets addressed first in this article because it is, by a wide margin, the single most consequential technical decision in most gold resource estimates and the one that the fewest readers of NI 43-101 reports have ever encountered. The term does not appear in press releases. It rarely comes up in analyst reports. The mechanics are buried in Section 14, sometimes in a single paragraph, sometimes in a table without commentary.

Gold grade distributions are positively skewed. Many low-grade samples, a tail of increasingly high values stretching to the right. Before interpolation, the QP has to decide what to do with the extreme high-grade assays. Uncapped, those outliers get smeared across surrounding blocks by the interpolation algorithm, inflating the estimate. Top-cutting sets a ceiling: every assay above the cap value gets reduced to the cap before it enters the block model.

The statistical tools for selecting the cap are well established. Log probability plots show the grade population on a cumulative probability scale; a break in slope at the upper end suggests where the main population ends and the outlier tail begins. Histograms and coefficient of variation analysis provide additional context. Cumulative metal contribution curves show what percentage of total contained metal sits above any given threshold, which reveals how sensitive the estimate is to the treatment of the upper tail. In a rigorous report, these analyses are presented with plots and the QP walks through the reasoning. The inflection point between "main population" and "outlier tail" is often ambiguous, covering a zone where a cap anywhere from, say, 28 g/t to 65 g/t could be defended. The QP picks a number somewhere in that zone.

Some reports state "a top-cut of 40 g/t was applied" and that is it. No plots, no analysis, no stated rationale. The reader is left with no way to evaluate whether the cap was derived analytically, drawn from professional experience in similar deposit types, or arrived at through a conversation between the QP and the company about what cap value would produce a resource figure that supports the project's narrative. That last scenario sounds like an accusation. It is a description of a dynamic that the regulation's disclosure framework does not prevent, because the regulation does not mandate disclosure of the statistical basis for the top-cut.

There is a further complication. Whether the cap is applied before or after compositing produces a numerically different result. Compositing, the process of combining irregular assay intervals into regular lengths before interpolation, already averages out some of the extreme variation. Capping after compositing removes less metal from the estimate than capping the raw assays first and then compositing the capped values. The order of operations should be stated. Often it is not.

Section 11. Sample preparation, analyses, and security. The section that gets skipped.

CRM results are the clearest window into assay accuracy. Certified reference materials have known grade values, typically certified by round-robin programs involving multiple laboratories. They are inserted into the sample batches going to the analytical lab. If the lab returns a value outside an acceptable window, usually ±2 standard deviations from the certified value, the batch has a potential accuracy problem. Failure rates of 3% to 6% are normal and expected in well-run programs. A rate of zero over hundreds of insertions is anomalous because laboratories drift, instruments require recalibration, and human error in sample preparation introduces occasional excursions. The way the QP handles the failures matters more than the failure rate itself. Re-assay of the affected batches, investigation of the cause, exclusion or correction of affected samples: those steps, when described, show the program functioning. A report that states "all CRM results were satisfactory" at a summary level does not give the reader enough to evaluate anything.

Blanks detect carry-over contamination. Barren material inserted between samples in the preparation sequence. If the lab's crushing or pulverizing equipment retains traces of a high-grade sample and deposits them into the next sample in line, the blank will show anomalous values. In a gold environment, even small amounts of carry-over from a 100+ g/t sample can produce detectable contamination in a subsequent blank. The contamination travels through the entire subsequent sample stream until the equipment is cleaned, so a single blank failure can implicate a run of samples, not just one.

Field duplicates test precision. Two subsamples from the same interval should return similar grades. In deposits where gold is fine-grained and evenly disseminated, duplicate variance should be low. In coarse-gold deposits, duplicate pairs can produce wildly divergent results because each subsample captures a different random selection of coarse particles. High duplicate variance in a coarse-gold deposit is a geological fact, not a lab error. High duplicate variance in a fine-grained deposit is a data quality flag. Distinguishing between these two scenarios requires cross-referencing with the mineralization description in Section 7.

Fire assay with a 30-gram charge is the default analytical method for gold. Thirty grams is roughly two tablespoons of pulverized rock. In a coarse-gold deposit, where individual gold particles can weigh several milligrams, a single particle landing in or out of that 30-gram subsample can swing the assay from 2 g/t to 150 g/t on the same interval. The variance is a sampling problem, not an analytical one: the lab is measuring the 30 grams accurately, but the 30 grams is not representative of the interval. Screen metallic fire assay was developed specifically for this situation. A much larger sample, typically 500g to 1kg or sometimes more, is screened to separate coarse and fine fractions. Each fraction is assayed independently, and a weighted average is computed. The procedure reduces the nugget effect at the sampling scale. When a technical report for a deposit with known visible gold uses standard 30-gram fire assay without discussing the sampling implications, the grade database carries a noise level that propagates into every subsequent step of the estimation.

Inter-laboratory check programs go beyond minimum regulatory requirements. Sample splits sent to a second commercial lab test whether the primary lab carries systematic bias. The differences are typically a few percent, and they can be consistently in one direction. These programs are uncommon in early-stage exploration reports. When they appear and the comparative results are discussed, the assay data has an additional validation layer.

Section 14 converts geological data into a number. The section is technically dense and contains multiple points where the QP's judgment determines the outcome.

Assay intervals from drill core come in varying lengths, depending on how the geologist logged and sampled the core. Before interpolation, the QP combines these into uniform-length composites. The choice of composite length interacts with the deposit geometry and the proposed mining method in ways that are easy to miss.

A narrow high-grade zone 1.2 meters wide, composited into 2-meter intervals, gets diluted by the surrounding lower-grade material included in each composite. The block model then reflects composite grades that are lower than the actual in-situ grade of the high-grade structure. Whether this matters depends entirely on whether the mine can selectively extract the narrow zone or has to take a wider cut. Some narrow-vein underground operations achieve stoping widths of 1.5 meters. Others have minimum stoping widths of 3 meters due to equipment constraints. If the composite length does not match the actual mining selectivity, the grade in the block model does not match what the mill will receive, in one direction or the other.

The composite length should be stated in the report along with its relationship to the minimum mining width or selective mining unit. Some reports do this explicitly. Others state the composite length without connecting it to the mining parameters described in a different section of the report.

Ordinary kriging is the dominant interpolation method for deposits with reasonable geological continuity. The variogram, the mathematical model describing how grade correlation changes with distance and direction, is the backbone of the kriging estimate. The variogram's parameters encode the QP's interpretation of grade continuity in three-dimensional space.

The nugget value represents grade variability at distances shorter than the drill spacing can resolve. In a coarse-gold deposit, the nugget can represent 40% or more of the total variance, meaning that nearly half the grade variability is essentially random at the scale of observation. A high nugget means the estimate carries large uncertainty at the individual-block level even where drill spacing is relatively tight. The sill is the total variance. The range is the distance beyond which samples provide no useful information about grade at a given location. Anisotropy ratios describe how the range differs by direction, typically longest along the strike of mineralization and shortest perpendicular to it.

These parameters directly affect resource classification. If the variogram range along strike is 45 meters and the drill spacing along strike is 40 meters, there is spatial correlation supporting the estimate. If the spacing is 60 meters, the samples are too far apart for the variogram to help, and the estimate is essentially extrapolating beyond demonstrated continuity. The report should present the variogram parameters. Many do, in a table or a figure. Some present only the fitted variogram plot without the underlying parameters, which makes evaluation harder.

Inferred, Indicated, Measured. There is no universal drill-spacing standard for these categories. The CIM Definition Standards describe them in terms of geological and grade continuity, and the QP translates those descriptions into spatial criteria for the specific deposit. Two QPs working on the same dataset can draw the classification boundaries differently, and both can produce defensible reports. The classification is a professional judgment informed by the variogram, the geological model, and the QP's experience with similar deposit types.

The conversion ratio from Inferred to Indicated across the industry is inconsistent. Some projects lose a quarter of their Inferred tonnage upon infill drilling. Others gain. The Inferred category communicates uncertainty about the deposit's size and grade distribution, and treating the Inferred number as a preliminary version of the Indicated number misreads what the classification means.

Open-pit resources must demonstrate "reasonable prospects for eventual economic extraction" under the CIM Definition Standards. The standard demonstration is a pit optimization using the Lerchs-Grossmann algorithm or the Whittle algorithm, producing a conceptual pit shell. Material inside the shell gets reported. Material outside does not.

The shell is defined by its inputs: metal price, mining cost per tonne of rock moved, processing cost per tonne of ore processed, metallurgical recovery, and pit slope angles. Every input is an assumption. The pit expands when the metal price goes up, when costs go down, when recovery improves, or when steeper slopes are assumed (steeper slopes mean less waste stripping). "Reasonable prospects" is not defined with numerical precision. A QP using $2,200/oz gold, bottom-quartile mining costs, and a 50-degree slope will generate a shell that encloses substantially more material than one using $1,800/oz, median costs, and 45 degrees.

Pit optimization inputs are sometimes in an appendix, sometimes in a footnote in Section 14, sometimes absent. Comparing them to recent commodity trading ranges and to published operating cost data for similar operations provides a check on whether the shell is conservative or aggressive.

Tonnage = volume × grade × density. The density dataset is almost always the smallest of the three. Tens of thousands of assays behind the grade estimate. A three-dimensional geological model behind the volume. And sometimes 40 or 50 bulk density measurements behind the tonnage calculation for millions of tonnes of reported resource.

A 10% error in bulk density propagates directly into a 10% error in tonnage. Whether different density values were assigned to different rock types and weathering horizons matters when the deposit spans an oxide-transition-sulphide profile. Oxide saprolite can have a bulk density of 1.8 to 2.2 g/cm³. Fresh sulphide rock underneath might be 2.8 to 3.2 g/cm³. Applying a single average density across that profile introduces an error whose magnitude depends on the relative proportions of each zone. The check takes thirty seconds: how many density measurements, differentiated by rock type or not.

Sections 6 and 7. The quality of these sections varies across reports to a degree that would surprise anyone who has only read a few of them.

Section 6 at its best reads like a geological argument built from property-level data: measured vein orientations, identified alteration minerals from petrography or XRD analysis, structural features mapped at surface and correlated to drill-hole intersections underground, cross-sections that integrate multiple data sources into a coherent spatial model. Section 6 at its worst reads like a digest of published government geological survey maps with some general statements about the regional tectonic setting and a sentence or two of speculation about how the mineralization might relate to known geological events. The first kind gives the reader confidence that the geological model beneath the resource estimate is grounded in specific observations. The second kind tells the reader that the geological model is still largely conceptual, and that the resource estimate built on it inherits that conceptual foundation.

Section 7 describes the mineralization itself: its geometry, its continuity, the relationship between grade and geological features, the mineral assemblages that carry the economic metals. The character of the mineralization determines the difficulty of estimation. A large, thick, disseminated porphyry system with grade that changes gradually across tens of meters is statistically well-behaved. Standard estimation methods work. Moderate drill spacings produce reliable results. A narrow-vein gold deposit where grade swings from 80 g/t to 0.3 g/t across 1.5 meters is a different problem entirely.

Oxide-sulphide descriptions in Section 7 have consequences in the metallurgy and economics sections. Oxide ore and sulphide ore frequently require different processing. A heap leach circuit for oxide material, a flotation or milling circuit for sulphide. If Section 7 describes a significant oxide zone and the economic model assumes one processing route, there is an inconsistency that the reader can check by flipping between sections.

Section 10. Drill-hole orientation relative to the mineralization geometry determines whether reported intercept lengths represent the deposit's true width or overstate it. Vertical holes into a vein dipping at 65° produce intercepts that are geometrically longer than the true perpendicular width by a factor of 1/cos(25°), roughly 1.1x. That correction factor grows as the angle between the drill hole and the normal to the mineralization increases. At a 40° angle between the hole and the vein normal, the drilled length overstates true width by about 1.3x. The report should present estimated true widths. Press releases often present drilled lengths without correction.

Core recovery matters in weak rock. In competent massive sulphide, recovery runs above 98% and is irrelevant to interpretation. In deeply weathered laterite, in clay-altered fault zones, in fractured ore near surface, core recovery drops. The lost material is not random. It is the weakest, most friable rock, and its grade may be higher, lower, or similar to the recovered material depending on the deposit-specific relationship between alteration and mineralization. Low core recovery in a mineralized interval introduces a sampling bias whose direction is uncertain without additional information.

Drill spacing and variogram ranges were discussed under classification. The point is that classification is not automatic. It involves the QP's judgment about whether the data density, in the context of the demonstrated grade continuity, supports the assigned category.

Section 12 describes the QP's independent checks. Site visits, collection of verification samples for independent re-assay, re-logging of selected core intervals, GPS checks on drill collar locations, comparison of original lab certificates to the digital database. The length and detail of this section roughly track the extent of independent work the QP performed.

Section 13. PEA-stage metallurgical programs are typically small: a handful of composite samples tested by bottle-roll, column leach, or locked-cycle flotation. The composites are usually drawn from the best-mineralized intervals in the central part of the deposit. By the Feasibility Study stage, the program should have expanded to cover geometallurgical domains representing the full range of ore variability: different lithologies, different alteration intensities, oxide versus transitional versus sulphide material, different gangue mineral assemblages that affect processing behavior.

The difference in projected recovery between PEA and FS for the same deposit can be eight to twelve percentage points, and that difference flows directly into revenue and NPV.

A specific cross-check between sections: the grind size (reported as P80, the sieve size through which 80% of the ground material passes) in the metallurgical test work versus the grind size assumed in the process design and economics. Finer grinding improves recovery and costs more in energy, steel consumption for grinding media, and capital for the grinding circuit. If the recovery projection in the economic model comes from test work at P80 75 microns and the process design specifies a grinding circuit producing P80 106 microns, those numbers belong to different flowsheets.

PEAs can include Inferred resources in the mine plan. PFS and FS cannot. This is the single regulatory distinction that most affects the reported economics, because the PEA models material whose existence has not been confirmed to Indicated confidence, which inflates both the tonnage and the mine life relative to what a PFS would show for the same project.

PEAs also model processing and mining configurations at a conceptual level. A PEA might assume heap leach because the capital cost is low. The PFS metallurgical program, with more samples across more ore types, might reveal that heap leach recovery on the sulphide-dominant ore is inadequate and the project needs a conventional grinding and flotation circuit, at four to five times the capital cost. The mining method and processing route in a PEA are concepts under exploration, not engineering commitments.

NPV depends on the discount rate, the metal price assumption, capital costs, operating costs, recovery, mine life, and throughput. Moving the discount rate from 5% to 8% on a 15-year mine life reduces NPV by roughly a third. Metal price has at least as much leverage.

IRR does not tell the reader when the money comes back. A project with an IRR of 22% requiring four years of construction and two years of ramp-up before reaching steady state has capital exposed for six years before payback begins. A smaller project with an IRR of 18% that pays back in 14 months has a fundamentally different risk profile.

When comparing NPV figures across projects, the discount rates must be normalized before the comparison has any meaning.

Sensitivity tables show how NPV responds to movement in each input. A project whose NPV goes to zero with a 15% metal price decline has almost no buffer. A project that stays positive through a 25% decline has structural margin.

Resources report contained metal: total estimated metal in the ground. Mining dilution reduces head grade by mixing waste into the ore stream. Mining recovery leaves some ore unmined in pillars, stope walls, or pit boundaries. Metallurgical recovery extracts a fraction of the metal fed to the plant. The cumulative reduction from contained to recoverable can range from about 20% in favorable conditions to 40-45% when mining selectivity is poor and metallurgy is complex. Resource headlines report contained ounces. Revenue comes from recovered ounces.

The summary is drafted last and read first. It compresses the report into a few pages. Press releases are built from it. The compression involves editorial choices about emphasis, and those choices systematically favor the most attractive numbers. A summary that leads with the NPV and the best intercepts and mentions the Inferred resource proportion in a subordinate clause halfway down the page is not inaccurate. It is arranged.

Section 25, the QP's interpretation and conclusions, is where risk language appears. "Additional drilling is recommended to confirm the continuity of the high-grade zones" means the QP does not consider continuity confirmed. "Metallurgical test work should be expanded to include variability samples" means the current program is not representative. Each phrase points back to a specific technical gap elsewhere in the report.

The language the QP uses about the resource estimate compared to the language about the economics usually shows an asymmetry. The geology tends to get firmer endorsement. The economics get more qualification. The asymmetry reflects the relative maturity of the technical work.

Section 26 states what the QP recommends next and the estimated budget. Drilling recommended means the geological model needs more data. PFS recommended means the current work is considered sufficient to advance. The budget tells the reader how much additional capital the project needs to reach the next milestone.

Classification, top-cut values, interpolation parameters, treatment of problematic drill holes: all judgment calls where qualified professionals can reach different conclusions from the same data. When a company engages a consulting firm and the resulting estimate falls below the company's expectations, the engagement can end. A different firm can be retained. The report that reaches SEDAR is the report signed by the QP who was willing to sign it. Prior engagements that produced less favorable results are not disclosed.

Pretium/Strathcona is the publicly documented case. Strathcona withdrew and published its reasoning. That is rare. Most consultant transitions happen without public explanation.

Large consulting firms operate internal peer review and carry institutional reputational weight beyond any individual QP. Sole practitioners carry the same professional liability without the institutional review layer.

Does the oxide description in Section 7 match the flowsheet in the processing section? Does the composite length match the minimum mining width? Is the metal price for cut-off grade consistent with the metal price in the economic model? Do drill spacings match variogram ranges? Is the grind size in the met tests the same as in the process design?

Each check takes minutes. Together they show whether the report was built as an integrated technical argument or assembled from independently authored sections that nobody reconciled.

A Feasibility Study on a well-drilled deposit from a company with production history needs attention on the economic and engineering sections. The geology has been tested. The open questions are about costs, recovery, and execution.

An exploration-stage PEA with 30 drill holes and an Inferred resource needs attention on Sections 6, 7, 9, 10, 11, and 12. Those sections contain whatever evidence exists that the deposit can be estimated at all. The economic sections at that stage are conceptual and should be read with full awareness of that.

Producing mines sometimes file updated technical reports that include reconciliation data: block model predicted grades versus mill head grades over months or years of operation. That reconciliation data, when it exists, shows whether the estimate survived contact with the ore body. The Rainy River reconciliation problems are an example of what that data looks like when it goes wrong. When reconciliation data shows the model tracking actual production within a few percent over sustained periods, the estimate has been validated in a way that no amount of statistical analysis at the pre-production stage can match. Early-stage reports cannot contain reconciliation data because there is no mine to reconcile against, which is another reason the geological and data quality sections carry more weight at earlier project stages.

Title page. The effective date is when the technical conclusions were finalized. The filing date is when the report was posted to SEDAR or SEDAR+. The gap can be months. Commodity prices move during that period. Drill programs generate new results. Permits advance or stall. Check the effective date before reading anything else. A report effective twelve or eighteen months ago describes the project as it was then.