Concept

Source hierarchy (literature / discovery / quote)

Overview

A source hierarchy is the ranking of the kinds of place a number can come from by how much evidential weight each can bear — from a direct quote or measurement for the specific case, through peer-reviewed literature and authoritative databases, down to discovered and promotional figures. It is what sets how strong a figure’s provenance is, and therefore how far a result built on it can be defended.

Body

What it ranks. The hierarchy ranks sources, not numbers — it says how much trust a value earns from where it came from, before anyone asks whether the value is right. Each input inherits the tier of its source, and the result inherits the tier of its weakest load-bearing input.

The tiers, strongest to weakest in TEA practice.

• Primary, case-specific — a vendor quote for your equipment, your own measured data, a direct expert estimate for your configuration. Highest weight because it matches the case being modeled, not just its class.
• Peer-reviewed literature & authoritative compilations — journal papers, standard reference texts, government or agency databases (physical properties, emission factors). Strong, but generic to a class rather than fitted to the specific case.
• Discovery — figures found by digging: industry reports, technical datasheets, conference papers, theses, patents, curated and uncurated databases. The broad working bucket of early TEA; quality varies enormously within the tier.
• Tertiary / promotional — press releases, marketing material, news, analyst targets. Weakest, because they are selectively reported and often a best case, a target, or an aspiration rather than an achieved value.

A stated assumption is its own tier. A declared modeling choice (“taking ~$X as a round anchor”) is a legitimate, traceable source — weaker than evidence but honest — and sits above an unstated guess, which falls below the hierarchy entirely because it cannot be evaluated at all.

Tier interacts with what is being sourced. The order is not absolute across quantities: for a physical constant, peer-reviewed literature is effectively primary; for an equipment cost, a vendor quote outranks any paper. The correct tier for a number depends on the kind of number it is.

Tier is separate from basis. Where a number came from and what basis it was measured on are independent axes. A top-tier figure on the wrong cost year, scale, or system boundary can be worth less than a lower-tier figure on the right basis. Both axes bound the figure.

Limits & typical error

• Tier is trust, not truth. A top-tier source can still carry a wrong or mis-stated value; the hierarchy ranks how defensible a number is, not whether it is correct (the provenance-is-not-correctness point in source form).
• Spread within a tier can exceed the gap between tiers. “Discovery” spans a curated cost database and a one-line press figure; treating the bucket as uniform mis-rates inputs by a wide margin, so the tier label alone under-determines the weight.
• Applying one fixed order to every quantity misranks. Using a generic tier ladder regardless of what is being sourced — trusting a journal paper over a vendor quote for a cost — inverts the real order for that quantity.
• Basis can dominate tier. A perfectly-sourced number on a mismatched scale, vintage, or boundary is attributed but not transferable; high tier does not rescue a wrong basis.
• Laundering promotes a number it shouldn’t. A guess cited to a secondary source that itself took it from another estimate acquires the appearance of a high tier while its citation chain dead-ends at nothing — the circular-source failure.
• Aggregation hides the governing tier. A blended figure assembled from mixed-tier inputs reads as a single tier; its real defensibility is set by the weakest contributing source, which the aggregate conceals.

See also

• Provenance & defensibility — the property a source tier feeds: how well the result withstands outside scrutiny.
• Data sourcing & validation — the build-time discipline that rates each number against this hierarchy and flags weak ones.
• Reference / comparable process — a common source whose own tier and basis this hierarchy rates.
• Uncertainty & accuracy class — the input band that a number’s source tier helps set.
• Engineering proxies & shorthands — deliberately low-tier-but-legitimate values, declared as such.
• Benchmarking against incumbents — comparison figures that inherit the source tier of the incumbent datum.
• Driver / key parameter — the high-leverage inputs whose source tier most determines the result’s defensibility.

Mini-example

Green ammonia’s $800/t headline rests on three load-bearing numbers, each at a different tier. The electricity intensity (10 MWh/t NH₃) traces to electrolyzer datasheets and peer-reviewed efficiency data — a strong tier for that quantity. The electricity price ($40/MWh, a stated round market anchor) is an assumption tier — honest and traceable, but defensible only as a declared choice with a range, not as a contract. The total capex ($1.0bn, order-of-magnitude) is a discovery-tier figure borrowed from comparable plants, awaiting a quote that would lift it to primary. The headline’s defensibility tracks the weakest of the three — the price and the capex — not the well-grounded intensity.

Separately, to show a number’s tier mis-read: taking a competitor’s ammonia cost from a press release (~tertiary) as a benchmark treats a selectively-reported best case as if it were a measured value. The same figure drawn from a curated cost-curve database would sit a full tier higher — the number can be identical while its weight is not.

See also

• Provenance & defensibility
• Data sourcing & validation
• Reference / comparable process
• Uncertainty & accuracy class (FEL-1/2, ±30%)
• Engineering proxies & shorthands
• Benchmarking against incumbents
• Driver / key parameter