Mining Feasibility Study
PEA, PFS, and BFS Explained

The incentive structure behind a PEA is the first thing to understand, ahead of any technical definition.

A junior mining company has spent $10M on exploration drilling. The board needs to show the market that the deposit has economic potential. A consulting firm is engaged. The firm knows the client's situation. The client needs a study that makes the next equity raise possible. If the consulting firm delivers a conservative PEA that produces a disappointing NPV, the client's share price drops, the financing round fails, and the consulting firm does not get hired for the PFS. Nobody discusses this dynamic in the room where the study parameters are set. Nobody has to.

So the commodity price assumption drifts toward the optimistic end of the consensus forecast. Recovery rates get benchmarked against analogous operations that happened to perform well. Operating costs reference the lower quartile of comparable mines. Each individual assumption is defensible. The cumulative effect pushes the NPV higher than a neutral set of assumptions would produce.

There are maybe a dozen consulting firms that write the majority of NI 43-101 technical reports for TSX and TSXV-listed juniors. Every one of those reports is filed on SEDAR+. It would be possible to build a database: pull every PEA a given firm has published over ten years, match each to the PFS or BFS for the same project three to five years later, calculate the average capital cost escalation. A firm-specific optimism coefficient. Nobody has built this database. Not sell-side analysts, not institutional mining funds, not retail investors. The work is tedious and nobody gets paid for it. Fifteen years of SEDAR filings sitting there, and the most basic empirical test of PEA reliability at the firm level has never been performed.

Now the technical dimension. A PEA is the only study stage that can include Inferred Mineral Resources in the economic analysis. Inferred Resources are the lowest confidence category: estimated from widely spaced drilling, with geological and grade continuity assumed rather than demonstrated. The NI 43-101 rules require the PEA to disclose that Inferred tonnes are speculative. The disclosure sits in the report. The press release says "$500M NPV at 30% IRR."

The accuracy range on capital costs is ±35% to ±50%.

These two PEAs can produce the same headline NPV. The probability that the project survives the next stage is completely different, and the resource category split is the number that determines this. When that breakdown is absent from the executive summary, that omission is informative.

The PFS is where most of the interesting things happen, and where articles about mining feasibility studies tend to spend the least ink.

Measured and Indicated Resources only. Inferred tonnes get reclassified as waste. A 15-year mine life at PEA can shrink to 8 at PFS. The plant costs roughly the same to build whether it runs for 8 years or 15.

Engineering specificity jumps from "conventional truck and shovel" to Cat 793F haul trucks, fleet of 12, two Liebherr R9800 excavators. Equipment selection drives haul road widths, fuel consumption, maintenance facility sizing, power demand. The water balance transitions from a spreadsheet row to the constraint that dictates plant location and may force a water treatment facility that nobody budgeted for.

Metallurgy deserves the most attention of anything in this article, because metallurgical risk at the PFS stage has destroyed more project economics than any other single factor, and the mechanism is the same every time.

A PEA tests four composite samples from the best-looking drill intercepts. Bottle-roll and column leach. 92% gold recovery. Fine. The PFS variability program tests 35 samples across geological domains, oxidation states, and depths. Clean oxide: 91%. Transitional zone: 83%. Primary sulphide with high arsenic: 72%. Weighted average across the mine plan drops to the low 80s.

The plant was sized for 92%. The revenue model assumed 92%. The mine sequence was optimized for 92%. An eleven-point drop in recovery requires a larger plant to compensate, which raises capex, which pushes the project past the economic threshold that the PEA defined. The geology did not change between PEA and PFS. The PEA tested the easy part of the deposit. The PFS tested what the mine will actually process.

Simple gold oxide deposits amenable to heap leach or CIL hold their recovery numbers from PEA to PFS, sometimes improve them. Everything else degrades. The industry has known this for decades. PEAs with complex metallurgy continue to be published with recovery rates from clean composites because the incentive structure described in the PEA section above has not changed and will not change.

The frustrating thing about this is that variability testing could be done at PEA stage. There is no regulatory prohibition. It costs more, it takes longer, and the junior company funding the PEA usually does not have the money or the patience. A junior that spent $2M on PEA-stage variability met testing and published a PEA with a realistic weighted recovery would produce a study with a lower NPV, a lower IRR, and a dramatically higher probability of surviving the PFS transition intact. The market would punish the lower headline number. The project would be better understood. This trade-off is available to every junior and almost none of them take it, because the share price responds to the headline, not to the underlying confidence in the headline.

Permitting. A PEA can treat permitting as an abstraction. A PFS has to engage with specific agencies, specific environmental regulations, specific communities. What the PFS team finds can end the project regardless of what the engineering and economics show.

Pebble in Alaska. Copper-gold-molybdenum, Bristol Bay watershed, salmon fishery. Over fifteen years of opposition from commercial fishing interests, Alaska Native communities, environmental groups. Multiple presidential administrations. No permit. Northern Dynasty has spent hundreds of millions. The deposit is geologically excellent.

Mining executives talk about Pebble with a mixture of frustration and resignation. The frustration is that a world-class copper deposit in a stable jurisdiction cannot be permitted while projects in countries with weak environmental regulation advance rapidly. The resignation is that Bristol Bay salmon are a multi-billion-dollar renewable resource and the opposition will never stop. Both of these reactions miss something. The comparison between permitting in Alaska and permitting in the DRC is not a comparison between "slow" and "fast." The Alaskan process is trying to measure cumulative watershed impact, indigenous rights, and intergenerational ecological risk. The DRC process is trying to measure whether the government wants the project to proceed. The documents both produce are called permits. They are measuring different things.

The environmental baseline timing issue connects to this. A company choosing when to start its one-year environmental baseline is making a decision with economic consequences. Start during a drought year: lower water tables, fewer wetlands, less aquatic habitat in the dataset. Start during a wet year: the opposite. Regulators have caught on and are requiring supplemental baselines for projects where the original data was collected during anomalous conditions.

Streaming and royalty companies at PFS stage. The technical teams at Wheaton Precious Metals, Franco-Nevada, Royal Gold evaluate hundreds of projects a year. Their pattern recognition across deposit types, jurisdictions, and study quality is built at a scale that individual investors cannot replicate. When they take an early position on a PFS-stage project, that decision encodes a level of technical evaluation that is not available from any other public signal.

When they pass, nothing happens. No press release. No disclosure. The deal that did not happen is invisible.

Rubicon Minerals deserves more space than any other case in an article about mining feasibility studies, because the failure mechanism goes to the foundation of what feasibility studies can and cannot do.

Start from the end. Late 2015. First gold poured at the Phoenix Gold Project in Red Lake, Ontario. Operations suspended within weeks. The grades encountered in underground development headings had no relationship to the resource model. Share price from over $3 to pennies. Creditor protection.

Red Lake is one of the best gold camps in Canada. Goldcorp ran a high-grade underground mine in the same district for years. The geology is well-characterized in academic literature and in decades of operational history from neighboring mines. The camp produces spectacular gold grades with extreme variability over short distances. This is known. This has been published. Anyone who has worked Red Lake understands that narrow-vein deposits in this camp demand tightly spaced underground drilling to build a resource model that captures the grade distribution. Wide spacing will produce a model that connects high-grade intercepts with smooth interpolation, creating an illusion of continuity that the geology does not support.

The QP who signed the resource estimate under NI 43-101 made a professional judgment call about drill spacing relative to geological complexity. NI 43-101 allows this: the code provides guidelines for resource classification, and the QP applies professional judgment on whether the data supports the classification. The judgment was wrong.

The process plant was engineered competently. The mine plan was well-constructed for the resource model it was given. The financial model was properly built. None of this was relevant. When the development headings penetrated the ore zones and the company could see the geology in three dimensions, the grade continuity that the resource model assumed did not exist.

After the collapse, a question circulated privately among geologists and analysts who followed the story: what was the QP's specific experience with Red Lake geology? NI 43-101 requires "relevant experience." A QP with twenty years in porphyry copper deposits has relevant experience in mining, broadly defined. That same person may have no relevant experience for a narrow-vein high-grade gold deposit with nugget-effect grade distributions in a camp where every geologist who has worked there will tell you about the grade variability over ten-meter intervals. The QP's qualifications are listed at the back of every NI 43-101 technical report. This section gets less attention from investors than any other part of the document, which is exactly backward given what happened at Rubicon.

There is no engineering solution to this. More engineering, more detailed flowsheets, more sophisticated financial modeling: none of it can detect or correct a resource model that is wrong. The only solution is geological: more drilling, tighter spacing, better understanding of the deposit's spatial variability. And the decision about whether the drilling is sufficient is a judgment call made by a QP whose professional background and specific deposit-type experience may or may not be adequate for the geological challenge at hand.

Rubicon was restructured. New management. New resource model based on tighter drill spacing. The project was eventually renamed and continued under different ownership. Red Lake is still a good camp. The deposit had gold in it. The original resource model just did not know where.

The standard narrative about Pascua-Lama focuses on engineering: 5,000 meters elevation, international border, water management, scope creep, capex from $2.8 billion to $8 billion+. All of that is accurate. What gets less attention, and what makes Pascua-Lama interesting for reasons beyond the engineering failure, is the management story.

Aaron Regent was CEO of Barrick Gold when the project was sanctioned. He was fired in June 2012 as costs escalated. Jamie Sokalsky replaced him, inheriting a construction project that was already in trouble. Sokalsky spent two years trying to manage costs down and eventually suspended the project. John Thornton, a former president of Goldman Sachs with no mining operating experience, was brought in as executive chairman by a board that had apparently concluded the problem at Barrick was not operational but strategic, or perhaps cultural, or perhaps something that a Wall Street executive could see that mining executives could not. Under Thornton's chairmanship, Barrick lurched through a period of strategic confusion that the mining press covered extensively and that the company's shareholders endured expensively. Mark Bristow, who had built Randgold Resources into one of the most successful mid-tier gold companies in the world, became CEO through the Barrick-Randgold merger in 2019. Under Bristow, the full Pascua-Lama write-down was taken, the project was mothballed, and Barrick's capital allocation strategy was overhauled.

The sequence from Regent to Sokalsky to Thornton to Bristow tracks a corporation trying successive theories about what went wrong. Operational failure, then cost management, then the idea that an outsider perspective was needed, then finally a return to someone who had built and operated mines successfully in difficult environments. The Thornton era is the most telling. Bringing in a Goldman Sachs executive to fix a mining company's capital allocation problem is a decision that makes sense in a boardroom and makes no sense underground. Thornton's tenure at Barrick was marked by executive departures, strategic drift, and a share price that did not recover until Bristow arrived.

The ±15% capital cost accuracy that a BFS promises assumes competent execution by a stable management team. The assumption is often wrong, and the feasibility framework has no mechanism for flagging it.

The BFS targets ±10% to ±15% capital cost accuracy and is prepared for institutional lenders. The lenders hire independent technical advisors whose role is adversarial: find the ways the project could fail to service its debt.

Contingency: 10% on a remote greenfield is too low. Lenders who have done this for more than five years know it. Owner's Costs below 5% of direct capital means the company has not realistically costed its own construction management. Sensitivity tables in Chapter 22 show the breaking points. Compare the base case commodity price to incentive price estimates from Wood Mackenzie or CRU: a BFS that needs a price 30% above incentive to generate a positive NPV is a cycle bet, not a project.

The Resource-to-Reserve conversion ratio measures how much of the Resource survives modification for real-world mining constraints. Above 70% from Indicated to Probable means clean geometry. Below 50% means complexity.

Whabouchi project in Quebec. Large spodumene pegmatite. The mining side was straightforward. The BFS proposed an electrochemical plant to convert spodumene concentrate into battery-grade lithium hydroxide. The technology worked at pilot scale. At commercial scale in the proposed configuration, it had not been proven.

Construction started. Costs overran. Creditor protection, late 2019.

The BFS priced a scale-up challenge as an engineering challenge. Engineering risk responds to better drawings. Scale-up risk responds to building a demonstration plant and running it. Nemaska did not build a demonstration plant. The gap between pilot results and commercial reality opened during construction and swallowed the project.

Restructured under new ownership. New study. Higher capex. The lithium stayed in the ground. The cost estimate moved.

Briefly. Rio Tinto, one of the two or three most technically capable mining companies on earth. The Hugo North underground expansion, a block cave, has seen capital estimates revised from below $6 billion to over $7 billion after geotechnical challenges emerged that the feasibility work did not capture. Block caving depends on the rock mass fragmenting and flowing under gravity. When the rock does not fragment as modeled, the production concept is at risk. Add a long-running dispute with the Mongolian government over cost overruns and tax treatment.

Infill drilling between PEA and PFS often costs more than the PFS. If the results are disappointing, the PFS may never happen. The press release will say "evaluating strategic options."

Between PFS and BFS: detailed engineering, permitting, geotechnical drilling, pilot met testing if warranted, offtake and infrastructure negotiations.

Some companies hire a separate engineering firm to peer-review the PFS before committing to a BFS. $200K to $500K, two to three months. A firm that did not work on the PFS. Companies that do this never announce it. Companies that skip it are overrepresented among those that later publish "updated BFS results" with completely different numbers.

The PFS-to-BFS timeline is a signal. Ten months for a complex project means corners were probably cut.

PEA: resource category breakdown, commodity price vs. spot and forward, AISC vs. industry cost curve.

PFS: how much did capex move from PEA, did recoveries hold across all ore types (not just the clean ones), did the mine life shrink. Tailings facility design and water balance produce more post-BFS cost blowouts than any other items.

BFS: construction schedule, contingency, ramp-up (nameplate in month three is a fantasy), closure costs. QP credentials at the back.

Market reaction on release day. Stock falls on a positive PEA headline: institutional investors found something in the report. PFS worse than PEA and stock holds: the market expected deterioration and considers the surviving economics credible.

Insider trading data on SEDI (Canada) or EDGAR (US). CEO buying in the open market six months after a positive PEA signals conviction. Insiders selling while press releases tout advancement signals a gap. Fifteen minutes per company.

Economics only work at elevated commodity prices. Massive infill drilling needed and the company cannot raise capital. Strong economics in a hostile permitting environment. The PEA was the exit strategy: establish the deposit's scale, get acquired by a major, let the major fund the PFS and BFS with its own balance sheet and team.

Donlin Gold, Barrick and Novagold. BFS completed 2011. Over 39 million ounces in Reserves. Not built. Over $6 billion capex. Western Alaska. No roads. A 500-km gas pipeline needed. Contested EIS. Novagold's entire market cap is built on a non-producing asset.

Kamoa-Kakula in the DRC was the rare success that advanced methodically through every study stage and delivered on the BFS promises. Extraordinary copper grades, 3% to 6% Cu, provided a margin of error that most projects do not have. Attributing the success to study discipline alone ignores the grade, the geometry, Zijin's construction capabilities, and the DRC's faster permitting environment.

NI 43-101 (Canada), JORC (Australia), SAMREC (South Africa), S-K 1300 (US). S-K 1300, effective 2021, allowed US-listed miners to disclose Mineral Resources for the first time, opening the capital markets to earlier-stage projects and bringing in generalist investors who evaluate mining companies through the same frameworks they use for SaaS businesses and who read Resource estimates without understanding the confidence hierarchy.