5 Resource Estimation Mistakes That Cost Mining Companies Millions

I’ve sat on both sides of the table in due-diligence reviews. The buy-side reviewer asking the awkward questions, and the sell-side geologist trying to defend a model that was rushed because the deal team needed numbers two weeks ago. The pattern is depressingly consistent. The same five mistakes show up in project after project, and each one has cost real money to real companies.

This post is the senior-to-junior version of that conversation. None of these are exotic. All of them are preventable with discipline that takes hours, not weeks. The reason they keep happening is that they look fine on screen and only break under audit pressure.

If you’re estimating a resource right now, run this list against your workflow before you sign anything.

Mistake 1: Composite length doesn’t match mining selectivity

Symptom: 1m raw assays composited to 1m for kriging because “shorter is more accurate.” Or 5m composites used on a deposit that will be mined on 2.5m benches with selective shovels.

Why it matters: The composite length defines the support of your samples. The block size defines the support of your estimate. The mining method defines the support that actually matters. If those three aren’t reconciled, your model either over-promises selectivity (small blocks fed by short composites that mining can’t actually pick up) or under-promises (long composites smear high grades into waste).

Real-world consequence: A Sumatran gold project I reviewed had 1m raw assays composited at 1m, kriged into 2.5m x 2.5m x 2.5m blocks. The mining contract specified 5m benches with conventional truck-and-shovel. The reconciliation after first 18 months of mining showed grade reductions of about 22% versus the resource model. The selective high-grade pods that justified the small block size couldn’t actually be mined separately at 5m benches. The model was technically correct and operationally fiction.

How to avoid it:

• Composite length should be a multiple of the smallest meaningful sample interval and aligned with bench height.
• Block size should be 30 to 50% of the average drillhole spacing, never smaller.
• Sub-blocking is for shape, not for grade resolution. If you sub-block to 1m for grade, you’re inventing data.
• Run a reconciliation simulation: re-block your model at the actual SMU (selective mining unit) size and report grade-tonnage at SMU support. That’s the number mining will actually deliver.

If your resource report shows grade-tonnage curves at the kriging block size only, ask why. Decision-grade reporting needs SMU-support curves.

Mistake 2: Top-cut decisions made without statistical justification

Symptom: “We capped Au at 30 g/t because the consultant did that on the last project.”

Why it matters: Top-cuts (capping high grades) are one of the most consequential decisions in resource estimation. Cap too aggressively and you destroy genuine high-grade tonnage that mining can actually recover. Cap too loosely and a few outlier samples drive your global mean and the entire estimate becomes unstable. Either way, if you can’t defend the number to an auditor with numbers and plots, you’re going to fail.

Real-world consequence: A bankable feasibility study for a Central Sulawesi epithermal Au project capped at the 99.5th percentile (about 22 g/t on a deposit with a true high-grade vein system reaching 80+ g/t). Internal review missed it. External technical advisor flagged it. Re-estimate without aggressive capping increased the contained ounces by approximately 18%, but more importantly, exposed that the original CP couldn’t justify the 22 g/t number. The CP was replaced. The financing was delayed by six months.

How to avoid it:

• Top-cut analysis runs three methods minimum: decile analysis, log-probability plot, and mean+nσ.
• Domain first. A single top-cut applied across geological domains is wrong. Vein domains often need no cap or a much higher cap than disseminated halo material.
• Document the decision: what method, what value, what percentage of the metal removed, and why you chose this over alternatives.
• A top-cut that removes more than about 5% of the contained metal needs an extra paragraph of justification, not a sign-off.

See the dedicated top-cut analysis walkthrough for the full method.

Mistake 3: Search ellipsoid copy-pasted from another deposit

Symptom: Search parameters in the kriging file are identical to the project the analyst worked on six months ago. Different deposit type, different drilling density, same numbers.

Why it matters: The search ellipsoid is one of the few user choices in kriging that significantly changes the estimate. The major-axis range, the anisotropy ratio, the orientation, the minimum and maximum samples, the octant restrictions. All of these need to come from the variogram and the geological model of this deposit, not the last one.

Real-world consequence: I reviewed a copper porphyry resource where the search ellipsoid had been imported from a Cordilleran porphyry template. Anisotropy 1:1:0.5 with a 250m major range. The actual variography on the project showed 1.5:1:0.6 with a 110m major range, and the major axis was rotated 35° relative to the template orientation. The kriging was pulling samples from across structural boundaries the geologist had carefully mapped. After re-running with the correct search, the Indicated category shrank by about 30% and the Inferred category grew. Net resource ounces unchanged. Classification confidence severely impacted. The project’s investment-grade rating was re-assessed.

How to avoid it:

• Search ellipsoid major axis aligns with the variogram major axis, which aligns with the geology. Three things, one direction.
• Anisotropy ratio comes directly from the variogram range ratios. If your variogram says 2:1:0.5, your search says 2:1:0.5.
• Major-axis range is typically 1 to 1.5 times the variogram range. Going beyond 2x is rarely defensible.
• Minimum samples and maximum samples are calibrated to drillhole spacing. Default of min=4, max=24 is a starting point, not a rule.

If you can’t articulate why your search ellipsoid has the orientation it does in three sentences referencing the geology, the variogram, and the drillhole spacing, you’re using someone else’s parameters.

Mistake 4: Validation skipped because the deadline was tight

Symptom: Cross-validation, swath plots, and visual inspection of the block model didn’t make it into the report. The estimate ran, the numbers came out, and the report was filed.

Why it matters: Estimation validation is what separates a defensible resource from a press-release resource. Without it, you have no internal evidence the model is consistent with the input data. Auditors notice immediately, and the technical advisor representing the buyer or financier will absolutely notice.

Real-world consequence: A West Java polymetallic project had its resource statement issued without swath plot validation. The report passed internal QA because the executive summary was polished. During buy-side due diligence, the technical advisor ran swath plots and found that the kriged block grades along strike showed a systematic 12% high bias relative to drillhole composites in the first 200m of the deposit. The estimate had over-extrapolated grades from a clustered sub-area. The deal valuation was reduced by about 25% on technical risk alone, and a re-estimate was required before close. Three months of lost time.

How to avoid it:

• Cross-validation (leave-one-out) on at least 5 to 10% of composites, with mean error and slope of regression reported.
• Swath plots in three orthogonal directions, comparing kriged blocks against declustered composites and nearest-neighbor estimates.
• Visual inspection: at least 5 cross-sections through the block model with composites overlaid, looked at by human eyes, not just rendered to a PDF page.
• Global statistics check: kriged mean grade should be within a few percent of declustered composite mean grade. Big differences mean you’ve over-smoothed or under-extrapolated.

If you’re skipping any of these because the deadline is tight, the deadline is the actual problem. Push back.

Mistake 5: Classification driven by tonnage targets, not data confidence

Symptom: The Indicated boundary is drawn so that total Indicated tonnage hits the threshold the press release wants. Drillhole spacing and kriging variance considerations are reverse-engineered to fit.

Why it matters: Classification is the single most important output of a resource estimate from an investor’s perspective. Measured, Indicated, and Inferred determine financability, what can convert to reserves, and what can be reported in a feasibility study. Classification driven by anything other than data confidence is misrepresentation, full stop. JORC and KCMI both treat it as a CP responsibility, and a CP who signs a tonnage-targeted classification is exposing themselves personally.

Real-world consequence: A nickel laterite project in Halmahera reported a Measured + Indicated tonnage that triggered the next financing tranche. Audit found the Measured boundary had been drawn at 80m drillhole spacing, with no defensible kriging variance threshold. The CP had inherited the project late and signed off on classification criteria that were never properly documented. The Measured category was downgraded to Indicated, the Indicated category partially downgraded to Inferred, and the financing tranche was rescinded. The CP’s professional standing took a public hit. The project was eventually re-estimated by a different team.

How to avoid it:

• Classification criteria are written before the estimate runs, not after the tonnage is known.
• Criteria reference: drillhole spacing, kriging variance or kriging efficiency, number of informing samples, geological continuity confidence.
• Numbers go on paper. Example: Measured = ≤25m drillhole spacing AND kriging variance ≤0.4 sill AND ≥3 informing holes within 25m.
• Classification gets reviewed by someone who isn’t on the deal team.

The CP signing the report has personal liability for the classification. If the classification was decided by anyone other than the CP, that CP shouldn’t sign.

The pattern underneath all five

Every mistake on this list is a discipline failure, not a technical skill failure. The geologists who make these mistakes know better. They know what compositing should look like. They know top-cuts need analysis. They know the search ellipsoid matters. They know validation is mandatory. They know classification has rules.

What they’re under is pressure. Deadline pressure. Deal team pressure. “Just get the number” pressure. And the work that protects against all five mistakes is the unglamorous, repetitive, paper-trail-generating work that doesn’t make slides look better. Statistical justification for the top-cut. Variogram-derived search parameters. Cross-validation plots. Classification rules in writing.

The senior geologist’s job, more than anything else, is to refuse to skip this work when the deadline says to skip it. The cost of saying no this week is a few uncomfortable conversations. The cost of saying yes is the project examples above.

How Orebit Geotools helps

The Geologist Toolkit phases are built specifically to make the boring discipline less boring:

• Phase 02 (Drilling EDA) runs decile analysis and log-probability plots automatically and forces you to record a top-cut rationale before composites generate.
• Phase 03 (Resource Estimation) auto-fits variograms with directional analysis and pulls the search ellipsoid from the variogram parameters by default. Override is available, but the default is the defensible one.
• Cross-validation, swath plots, and global mean checks run as part of the standard estimation output. You get the validation pack as a PDF whether you ask for it or not.
• Classification criteria are inputs you have to type in before classification runs. The block model can’t get classified without rules on paper.

For the JORC vs KCMI reporting requirements, the toolkit outputs Table 1 sections aligned to the standard, with the validation evidence attached.

See the toolkit → · Buy on Lynk.id → (Bundle IDR 99K, single modules IDR 49K, lifetime access)

Bottom line

These five mistakes have cost mining companies hundreds of millions in delayed financings, downgraded classifications, and re-estimated resources. None of them require advanced training to avoid. They require senior geologists who push back on deadlines and do the work properly the first time.

If any of the five symptoms describe a project you’re working on right now, fix it before someone else finds it. The conversation is much cheaper internally than it is at the audit.

Got a war story you can share (anonymously) for a future post? Email hello@orebit.id. Junior geologists learn from the failures we admit to.