Where Our Data Comes From
Sourced to the original document, not to someone who read it. One layer of transcription introduces one layer of potential error. We accept zero.
Primary Sources
The source hierarchy
Every number comes from one of these source types, in strict order of preference. We always use the highest-ranked available source.
Company website PDF
Annual reports, integrated reports, quarterly activities reports, production reports. Published directly by the company on their investor relations page. This is our preferred source for the vast majority of records.
SEC EDGAR filings
10-K (annual report for US companies), 20-F (annual report for foreign private issuers cross-listed in the US), 10-Q, 6-K exhibits. The SEC hosts filings going back decades. When a company cross-lists in the US, their 20-F is often more detailed than their home-market filing.
SEDAR+ filings
Annual Information Form (AIF) and Management Discussion & Analysis (MD&A) for TSX and TSXV companies. Canada's equivalent of SEC EDGAR.
ASX announcement PDF
Direct from announcements.asx.com.au for Australian companies. These are the exact documents lodged with the Australian Securities Exchange as continuous disclosure.
London RNS via Investegate
Regulatory News Service announcements for LSE-listed companies. Glencore, Anglo American, Antofagasta, Ferrexpo, and other UK-listed miners publish results through RNS.
JSE SENS
Stock Exchange News Service for Johannesburg Stock Exchange companies. Gold Fields, AngloGold Ashanti, Impala Platinum, Sibanye-Stillwater, and other South African miners.
Wayback Machine
Used only when a company URL is confirmed dead. Gold Road Resources (GOR) was acquired and goldroad.com.au went offline. We found all annual reports archived on the Wayback Machine and verified each URL worked.
Document mirrors (kalkik.com, minedocs.com)
Last resort only, used when both the company URL and Wayback Machine fail. Used for a handful of older Evolution Mining documents when their domain migration broke all historical URLs.
Rejected Sources
What we do not use — and why
Wire services
GlobeNewsWire, PRNewswire, BusinessWire, Cision
Wire services distribute press releases but they are not the primary source. The exception: a small number of companies (Wesdome Gold, Allied Gold, G Mining Ventures) use GlobeNewsWire as their actual press release platform — their IR pages link directly to GNW. In those cases, GNW is the canonical source. For everyone else, we find the company-hosted version.
Forums and aggregators
HotCopper, MarketScreener, Voxmarkets, Stockhead
HotCopper is an Australian stock forum that reposts ASX announcements. The original is at announcements.asx.com.au — we always use that. MarketScreener and similar aggregators scrape and republish company data with no original value added.
News and financial media
Reuters, Bloomberg, Yahoo Finance, Seeking Alpha, Mining.com
These sites cite the number but are not the source. A Reuters article saying “Newmont reported AISC of $1,444/oz” is a journalist quoting the annual report. The annual report is the source. One layer of transcription introduces one layer of potential error.
Exchange Coverage
7 exchanges across 4 continents
| Exchange | Location | Companies | Filing System | Key Filings |
|---|---|---|---|---|
| TSX | Toronto, Canada | 57 | SEDAR+ | AIF, MD&A, NI 43-101 |
| ASX | Sydney, Australia | 39 | ASX Market Announcements | Annual Report, Quarterly Activities |
| NYSE | New York, USA | 22 | SEC EDGAR | 10-K, 20-F, 8-K |
| TSXV | Toronto, Canada | 12 | SEDAR+ | AIF, MD&A |
| LSE | London, UK | 11 | RNS / Investegate | Annual Report, Final Results |
| JSE | Johannesburg, SA | 10 | SENS | Integrated Report, Annual Results |
| NASDAQ | New York, USA | 5 | SEC EDGAR | 10-K, 20-F |
Document Volume
600+ source documents
156 companies across up to 6 years of data means reading hundreds of primary source documents. Each document is a PDF annual report, SEC filing, or exchange announcement ranging from 20 to over 300 pages. The relevant operational data is typically found in production summary tables, operational review sections, and financial highlights.
For each document, we identify the correct tables, extract the values we track, verify the units, check for definition changes, and record the proof fields (URL, page, verbatim quote). This process is repeated for every company, every year, across all 7 exchanges.
Extraction Process
How we go from PDF to datapoint
Mining company annual reports are complex documents. A typical gold miner's annual report includes mine-by-mine breakdowns, multiple cost definitions, multiple currencies, and production data split between equity and consolidated bases.
Our extraction process for each company follows a consistent pattern:
Locate the primary source document for each fiscal year
Find the consolidated production summary table (usually in the first 10 pages or operational review section)
Extract production volumes, verifying units match the KPI registry
Find cost tables (AISC, C1, cash cost) and record the exact definition the company uses
Find realized price data (usually in a financial summary or commodity price table)
Find efficiency metrics (grade, recovery, throughput) in the operational review
Record a verbatim quote from the page for each extracted value
Compare reporting_basis, definition_name, currency, and unit against the prior year to detect breaks
Why This Matters
The difference between sourced and cited
Most data services cite a number. They tell you the value and maybe the company it came from. ProveMines sources every number: the exact document, the exact page, the exact text.
This distinction matters because errors compound. If a data aggregator misreads a number from a news article that misquoted an annual report, the error is invisible. There is no way to trace it back. With ProveMines, you can follow the URL to the PDF, turn to page 47, and find the exact sentence. If we got it wrong, you will know immediately.
This is what “every number sourced” means. Not referenced. Not attributed. Sourced — with a verifiable chain from the number in our database to the text on the page of the original filing.
Check our sources
Click any value in the workspace to see its source document, page number, and verbatim quote.
Open Workspace