Benchmark Spread Methodology

A Benchmark Spread is the platform's structured publication of where multiple public sources diverge on the same metric for an asset, the size and source of the divergence, and the most-defensible reading with reasoning. The discipline addresses what an institutional capital allocator most needs to know but rarely sees published: not the answer, but the confidence range around the answer.

The structural advantage. Incumbent institutional research platforms (Wood Mackenzie, S&P Global Capital IQ Mines, CRU, Benchmark Mineral Intelligence, Fastmarkets) produce single-point estimates from proprietary data models. Where their estimates diverge from each other or from operator disclosures is largely opaque to the buyer — the buyer subscribes to one platform and trusts its number. Publishing divergence transparently is structurally hostile to that subscription model. The platform's three-state Quality Standard (Sourced / Derived / Absent), field-level provenance discipline, and Counterparty Extension architecture make divergence-publication structurally feasible. Benchmark Spreads are the operational form that discipline takes for high-stakes capital-allocation metrics.

What a Benchmark Spread is

For an asset and a metric (NPV, mine life, AISC, CAPEX, resource estimate, royalty rate, recovery rate, etc.), a Benchmark Spread is the structured publication of:

Every credible public-source value for that metric — operator disclosure, regulatory filing, peer-reviewed analysis, comparable-asset reference, platform-derived calculation. Each value carries a date and a primary-source citation.
The numerical spread — the range, the median, and where the operator-disclosed value sits within that range.
The most-defensible reading — the platform's published view of which value should anchor a screening-stage decision, with the reasoning made visible.
The implied confidence range — what the spread tells the reader about how much weight to put on the central estimate.
What changed across time — when subsequent disclosures revised earlier numbers, the trajectory is published as part of the spread.

A Benchmark Spread is not an opinion masquerading as data. It is the structured surfacing of public-source divergence, with the platform's reasoning published as part of the artefact so a reader can replicate or contest it.

Why this is decision-aid utility

The institutional reader's question is rarely "what's the answer." It's "which answer should I trust, and how confident should I be?" A DFI investment officer screening a critical-minerals development asset cannot rely on operator-FS NPV alone — operator FS is sponsor-prepared and sponsor-presented. They cannot rely on a single research-house estimate alone — that estimate is one model from one institution. The credit memo work that follows requires the institution's own confidence range, which is bounded by the public-source divergence the institution can verify. Surfacing that divergence is high-value primary work that would otherwise require a senior analyst spending two-to-three days assembling per asset. The platform produces it once, publishes it openly, and refreshes it as new disclosures arrive.

Three-state Quality Standard mapping

Benchmark Spreads operate within the platform's three-state Quality Standard. Every spread carries one of three states explicitly:

Sourced spread

All values in the spread trace to named primary sources with dates. The spread itself is a structured surface of the public-source landscape. Most-defensible reading is platform editorial judgment based on source authority, recency, and methodological transparency.

Derived spread

Spread values include platform-derived calculations (e.g. cycle-1.5 reconciling cost from the DCF Test Battery; comparable-asset-cost benchmarks scaled to the subject asset). Each derived value's calculation is published; inputs are themselves Sourced or Derived. The state of each input is visible.

Absent

Where no credible public-source values exist for the metric, or where existing values cannot be reconciled to the public-source standard, the spread is recorded as Absent. No platform-fabricated estimates substitute for absence.

What makes a Benchmark Spread defensible

Five criteria the platform applies before publishing a Benchmark Spread:

Multiple credible sources exist. A spread requires at least two independent public-source values. A single-source datum is not a spread; it is an operator disclosure surfaced as Sourced data.
Sources are dated and citable. Each value carries a primary-source citation with date. Wikipedia summaries, undated industry reports, and aggregator estimates without provenance do not qualify.
Divergence is material. Spreads are published where the divergence affects screening-stage decisions. Trivial measurement-error divergences are not spreads; structural divergences (different methodologies, different time horizons, different scopes, different assumptions) are.
Most-defensible reading reasoning is visible. The platform's published view is reasoned in the artefact, not asserted. A reader who disagrees can identify exactly where the disagreement lies and apply their own judgment.
Counterparty Extension is preserved. No source is described as a commissioning party, customer, partner, or institutional relationship. Sources are public-source documents engaged with on their published merits.

What a Benchmark Spread does NOT do

It does not predict future values. A spread is a snapshot of public-source divergence at the date of publication. Forward-looking estimates of where the metric will go are explicitly outside scope.
It does not substitute for an Independent Engineer (IE) report. A real DFI engagement requires commissioned IE work that re-walks the underlying technical model. The spread surfaces public divergence; the IE produces internal validation.
It does not produce institutional credit memo work. Risk weighting, stress-test scenarios specific to the institution's portfolio, mandate-fit decisions — all sit with the institution.
It does not implicate any source in a relationship with the platform. Operator disclosures, regulatory filings, research-house publications are engaged with as public-source documents only.
It does not provide proprietary or non-public data. Where the most informative numbers are confidential to specific parties (lender models, internal cost curves, transaction-confidential information), the spread is bounded by what is publicly disclosed.

How a Benchmark Spread is constructed

The platform's standard construction sequence for a Benchmark Spread on a single metric:

Identify the metric. Specific enough to be measurable (e.g. "after-tax NPV at 8% real discount rate, mine-life basis"); not aggregated (not "asset value" generally).
Catalogue public sources. Operator disclosures over time (PEA → PFS → DFS → annual reports → quarterly updates); regulatory filings (NI 43-101 / JORC / S-K 1300 Technical Report Summaries); peer-reviewed academic analyses; industry-reference cost curves where applicable; comparable-asset reference points where structurally informative.
Tabulate values with dates and sources. Field-level provenance per the published Quality Standard.
Identify the trajectory. Where multiple operator disclosures span time, the trajectory itself is a benchmark signal — increasing resource estimates with infill drilling; CAPEX growth with detailed engineering; NPV revisions with operator-deck updates.
Surface the divergence drivers. Methodological differences (different price decks; different mine lives; different reserve-vs-resource scope); scope differences (initial development vs LoM-extended); time differences; assumption differences (recovery rates; tax structures; royalty applications).
Publish the most-defensible reading. Editorial judgment on which value should anchor screening-stage decisions, with reasoning visible.
Publish the implied confidence range. What the spread tells the reader about how much weight to put on the central estimate, and what would change the reading materially.
Audit-log the publication. Each Benchmark Spread is recorded in the public Audit Log at publication; updates are recorded as separate audit-log entries.

Update cadence

Benchmark Spreads are updated on three triggers:

New primary disclosure. Operator quarterly or annual filings; regulatory updates; new TRS publications. The platform's Daily Watchlist routine (published at /methodology/) surfaces material disclosures within the platform's Correction Velocity SLA.
Substantive new third-party analysis. A new peer-reviewed paper or research-house analysis with material divergence is incorporated into the spread.
Quarterly deep audit. Per the platform's published audit cadence (Quarterly: deep audit on three rotating provinces), Benchmark Spreads on assets in the audit cohort are reviewed comprehensively and updated where the underlying landscape has shifted.

Each update is recorded in the Audit Log with the trigger, the prior reading, the revised reading, and the change rationale.

Where the platform publishes Benchmark Spreads

Initial publication: Benchmark Spreads on the four published case-study assets — Kabanga Pre-FID, Manono Disputed Tenure, Loulo-Gounkoto Post-Settlement, and Lobito Corridor Infrastructure. Each spread is nested under the asset's case-study directory and linked from the parent dossier.

Roadmap: Benchmark Spreads extend to the platform's full intelligence-grade tier (18 IG dossiers) across calendar 2026, sequenced by where institutional reading interest indicates greatest demand. The discipline architecture is the same regardless of asset; only the specific metrics and source-documents change per asset.

Why this discipline matters now

African mineral assets are particularly subject to public-source divergence. Three structural reasons:

Operator-disclosure trajectory is often long. An asset under development for ten or fifteen years generates multiple operator disclosures (scoping → PEA → PFS → DFS → updated FS → operating quarterly reports), each with revised numbers. The trajectory itself is informative.
Multiple regulatory regimes apply. An ASX-listed sponsor with a DRC asset might file under JORC; a NYSE-listed sponsor with a Tanzanian asset files under S-K 1300; a TSX-listed sponsor files under NI 43-101. Cross-regime reconciliation is non-trivial and surfaces as divergence.
Geopolitical and contested-tenure layers add divergence. Assets where ownership is contested (e.g. Manono), where settlements have been negotiated (e.g. Loulo-Gounkoto), or where multi-jurisdictional infrastructure is in play (e.g. Lobito Corridor) generate divergence between competing public claims that institutional readers must reconcile.

The platform's coverage focus on Africa and on critical minerals means these divergence patterns are concentrated in the platform's data layer. Publishing them as structured Benchmark Spreads converts the platform's editorial discipline from a quality bar (the three-state Quality Standard) into a published decision-aid output (the spread itself). The same data discipline produces both.

← Methodology · Quality Standard · Case Studies →