How We Built This

One kilogram of copper produced by BHP is identical to one kilogram of copper produced by Freeport-McMoRan. One troy ounce of gold from Newmont is the same as one from Barrick. This physical homogeneity is rare across industries — a Tesla is not interchangeable with a Toyota — and it makes mining uniquely suited to standardized operational comparison.

Because the products are the same, the operational metrics are genuinely comparable: how many units were produced, what it cost per unit, what grade the ore was, what recovery rate the mill achieved. These metrics strip away the noise of geography, accounting policy, and corporate structure. They measure the physical reality of extracting minerals from the earth.

This is why we focus on operational data rather than financial data. Revenue depends on commodity prices and hedging strategy. EBITDA depends on accounting policy. But production volume, cost per unit, ore grade, and recovery rate are physical measurements that reflect how well the actual mining operation is performing.

Every metric falls into one of four families:

Each KPI has a commodity suffix: aisc_au always means gold AISC in $/oz. c1_cost_cu always means copper C1 cost in $/lb. The commodity is encoded in the ID so the system can enforce that every record of a given KPI uses the correct unit.

We defined 132 KPIs in the registry. 102 have actual data. The remaining 30 are defined but no company in our universe reported that specific metric in a way we could capture.

Our criteria: western-listed, actively producing, publicly reporting mining companies that produce at least one of our 12 commodities at meaningful commercial scale. This gives us 156 companies across 7 exchanges (TSX, ASX, NYSE, LSE, JSE, TSXV, NASDAQ).

Why western-listed: Western stock exchanges require standardized, publicly accessible filings in English. The NYSE requires SEC filings. TSX requires SEDAR+ filings. ASX requires continuous disclosure. All of it is public, all of it goes back years, and all of it is in English. Chinese and Russian exchanges were excluded because filings are often not in English and historical documents are not consistently accessible.

Why only producers: Royalty companies (Franco-Nevada, Wheaton Precious Metals) don't operate mines — they collect royalties on other companies' production. Our KPIs are meaningless for them. Development-stage companies have no commercial production to measure. Physical commodity funds hold metal in storage. Only active producers generate the operational data our system tracks.

Why these 12 commodities: We chose commodities that are physically homogeneous. One kilogram of copper is identical regardless of who produced it, so production volumes are genuinely comparable. We excluded rare earth elements (different elements are different products), specialty chemicals, and processed products where the “unit” varies by product type.

Every single number in ProveMines comes from reading the company's own filing. Not from a data feed. Not from a third-party aggregator. From the actual PDF document the company filed with their stock exchange.

For each company, for each year, we locate the primary source document (annual report, 10-K, 20-F, production report), find the relevant tables and commentary, extract the values, verify the units, check the definition methodology, and record a verbatim quote from the source page.

This means every one of 3,220 datapoints has a direct URL to the source document, an integer page number within that document, and an exact quote from that page containing the number. Anyone can independently verify any value by following the URL and turning to the page.

The source hierarchy is strict: the company's own website first, then SEC EDGAR, then SEDAR+, then ASX announcements, then LSE RNS. We never use wire services (GlobeNewsWire, PRNewswire) as sources unless the company itself uses the wire service as its press release platform. We never use aggregators, news sites, or forums.

A company that produced 100,000 ounces of gold is not comparable to one that produced 3,000,000 ounces. The big producer might be getting worse while the small producer is getting dramatically better. Ranking by production size tells you who is big, not who is improving.

Comparing absolute values across companies is even more problematic for costs. Newmont's AISC of $1,444/oz versus Centerra Gold's $1,100/oz does not tell you much — they operate in different countries, different cost environments, different ore types, potentially different definitions of AISC. Comparing how each company's AISC changed over the same time period is far more meaningful.

This is a deliberate product decision, not a limitation. Every view in the workspace shows percentage change. Absolute values are shown as context only — beside the percentage, never as the primary metric. This keeps the focus on trajectories and trends, which is what actually matters for understanding whether a mining operation is improving or deteriorating.

All-In Sustaining Cost (AISC) is the most widely tracked cost metric in gold mining. The World Gold Council published guidance in 2013 defining it and updated it in 2018. But “guidance” is not a standard. Companies interpret it differently.

The same acronym, reported by different companies, can mean: AISC including or excluding corporate G&A, AISC net of by-product credits (silver and copper credits reduce the gold AISC) or gross, AISC on a per-oz-sold or per-oz-produced basis. A company that nets out significant by-product credits will show a much lower AISC than a pure gold producer, and comparing them directly without knowing this is misleading.

Every datapoint in ProveMines has a definition_name field that describes how the company calculates the metric. When the definition changes between years — because the company changed their methodology — the workspace shows a break. Users can see both the old and new definition side by side.

Every datapoint has exactly 17 fields. The schema is intentionally tight — every field exists because it solves a specific data quality problem we encountered during collection.

Currency is always the company's native reporting currency — we never convert. If an Australian company reports costs in AUD, we store AUD. If they also provide a USD equivalent, that becomes a separate record. This preserves the original data as the company reported it.

Reporting basis is checked every year, not assumed. AngloGold Ashanti switched from attributable to consolidated reporting in 2022 — a real methodology change that affects every number. The system catches this because it checks reporting basis per year, not per company.

ProveMines tracks what mining companies actually did, not what they say they will do. We do not collect forward guidance, management targets, analyst estimates, or mineral resource statements. Only actual reported results from completed fiscal years.

Forward guidance is by nature speculative. Companies routinely miss their own targets. Analyst consensus estimates are opinions, not facts. Mineral reserves are geological estimates that depend on commodity price assumptions. None of these can be “proven” in the way that a reported annual production figure can.

Similarly, we exclude revenue, EBITDA, net income, free cash flow, and capex. These are financial metrics that depend on accounting policy, tax jurisdiction, hedging strategy, and corporate structure. They describe the company as a financial entity. Our KPIs describe the mining operation as a physical system: how much was extracted, at what cost, at what grade, with what recovery.

This is not a design feature. It is the foundational constraint the entire system was built around. If we cannot prove a number, we do not include it.

Every datapoint stores three proof fields: the URL of the source document (a direct link to the PDF or filing, not a search page), the integer page number within that document, and a verbatim quote from that page containing the number. The quote is exactly as it appears in the original document — no paraphrasing, no annotations, no formulas we constructed.

This means any user can independently verify any value: click the proof URL, go to the specified page, find the quoted text. No other mining data service provides this level of source transparency for every single data point.