Metallurgical Recovery Explained: The Assumption That Quietly Sinks Projects
Recovery converts contained metal into saleable metal. It is derived from testwork on a handful of samples and applied to an entire orebody. When it is wrong, no amount of grade saves the economics.
Metallurgical Recovery Explained: The Assumption That Quietly Sinks Projects
Summary box
- Recovery is the percentage of contained metal that actually reports to a saleable product. Contained metal is geology; recovered metal is revenue.
- It is estimated from testwork on a small number of samples and then applied to an orebody of millions of tonnes.
- Composite testwork on a favourable blend flatters recovery. Variability testwork on spatially representative samples is the honest version, and it is often absent at early stages.
- Recovery and payability are different things. A 92% mill recovery followed by a 96% payable smelter term nets 88% of contained metal.
- A five-point recovery miss is frequently larger in dollar terms than a five-point grade miss, because it applies to every tonne for the life of the mine.
Where recovery sits in the chain
A resource statement reports metal contained in the ground. Getting from there to cash involves several haircuts:
contained metal
x mining recovery / dilution -> metal delivered to the mill
x metallurgical recovery -> metal in concentrate or doré
x payability (smelter/refinery) -> metal you are actually paid for
- treatment & refining charges, penalties
= revenueMetallurgical recovery is the largest single step in that chain for most operations, and the one most dependent on an assumption rather than a measurement.
How the number is produced
Recovery comes from Item 13 of an NI 43-101 technical report — Mineral Processing and Metallurgical Testing — and the equivalent section under JORC and S-K 1300.
The sequence typically runs:
- Bench-scale tests on drill core: bottle rolls, flotation tests, leach tests.
- Composite samples blended from several intervals to represent an "average" feed.
- Locked-cycle tests simulating recirculating loads in a real flotation circuit.
- Variability testwork on samples drawn from across the deposit — different zones, depths, alteration types, grades.
- Pilot plant work, at feasibility level on larger or more complex projects.
The critical distinction is between step 2 and step 4.
Composite testwork answers: what does the average feed do? It is cheap, early, and produces a single flattering number.
Variability testwork answers: what does the worst feed do, and how often will we see it? It is expensive, later, and produces a distribution.
A mine does not process the average. It processes whatever the pit or stope delivers that week. A project whose recovery assumption rests on composites has not tested the risk it will actually run.
What destroys recovery
Refractory mineralogy. In gold, fine gold locked in sulphides — pyrite, arsenopyrite — will not respond to conventional cyanide leaching. Recovery can fall from 90%+ to 40–60% without pre-treatment. Fixing it means pressure oxidation, roasting or bio-oxidation: large capital, high operating cost, and a different project entirely.
Preg-robbing. Carbonaceous material in the ore adsorbs dissolved gold back out of solution before it can be recovered.
Clays and fines. Degrade heap-leach percolation and complicate flotation. Recovery drops and reagent consumption rises.
Deleterious elements. Arsenic, antimony, mercury, bismuth, fluorine, chlorine, magnesium. These may not lower recovery much, but they attract smelter penalties or make concentrate unsaleable. A clean 25% copper concentrate and a penalised one are different products.
Grind size sensitivity. Finer grinding lifts recovery and consumes more power. The optimum is an economic choice, not a technical constant, and it shifts with power price.
Ore hardness variability. Harder ore than designed reduces mill throughput. Throughput is not recovery, but the two interact: pushing tonnage through a constrained mill often means coarser grind and lower recovery.
Recovery is not one number
Better disclosure gives recovery by ore type, by grade range, or as a function of head grade. Weaker disclosure gives a single life-of-mine percentage.
Recovery commonly rises with head grade — low-grade material recovers worse. A project mining above reserve grade early will therefore report better-than-modelled recovery early and worse later, entirely from sequencing.
For polymetallic deposits there is a recovery per payable metal, and they trade off against each other. Pushing copper recovery can depress the gold and silver reporting to the same concentrate. A single blended "recovery" figure for a polymetallic project is not a meaningful number.
Doing the arithmetic
A deposit with 50 Mt at 1.10 g/t gold contains roughly 1.77 Moz.
| Recovery | Payability | Saleable oz | Δ vs 92% case |
|---|---|---|---|
| 92% | 99.9% (doré) | 1.63 Moz | — |
| 87% | 99.9% | 1.54 Moz | −89,000 oz |
| 80% | 99.9% | 1.42 Moz | −211,000 oz |
| 65% (refractory, untreated) | 99.9% | 1.15 Moz | −478,000 oz |
At $2,400/oz, the gap between the 92% and 80% cases is roughly $506 million of gross revenue over the life of the mine. No grade improvement realistically closes that.
This is why Item 13 deserves more attention than it usually gets. Analysts scrutinise grade to two decimal places and accept recovery as given.
Auditing Item 13 in ten minutes
- How many samples were tested? Single digits at feasibility stage is thin.
- Are they spatially representative? Look for a map or table of sample locations against the deposit and the mine plan. Samples all drawn from one high-grade zone do not represent the feed.
- Was variability testwork done, or only composites? If only composites, the recovery is a central estimate with no stated spread.
- Does the recovery used in the economic model match the testwork? Studies sometimes model a number above the best test result on the basis of "expected optimisation". That is a forecast, not a result.
- Were deleterious elements assayed? Absence of an arsenic or mercury assay is not evidence of absence.
- Is there a grind-size versus recovery curve? If so, does the modelled grind match the modelled power cost?
- For producers: how does actual recovery compare with the feasibility assumption? Reported quarterly. This is the single best reality check available, and it is free.
That last point is the highest-value check in the list. A producer consistently delivering below its feasibility recovery is telling you the study was wrong — and telling you about every other project by the same consultants and the same management.
Red flags
- Recovery in the economic model exceeding every individual test result.
- No variability testwork at pre-feasibility or feasibility stage.
- A single recovery figure for a polymetallic deposit.
- Recovery unchanged from PEA to feasibility despite substantially more testwork.
- Concentrate grade and impurity levels not disclosed for a base-metal project.
- Heap-leach recovery quoted without a column-test duration; ultimate recovery over 180 days is a different proposition from recovery over 1,000.
How Mining Terminal handles recovery
Recovery assumptions are extracted from technical reports with the process route, ore type and study stage attached, and traced back to the source document. Where a report gives recovery by ore type or grade band, those are preserved as separate records rather than averaged into one figure — an averaged recovery for a variable orebody is precisely the number that misleads.
Where a filing states an economic model recovery that differs from its own testwork results, both are retained. The gap between them is information.
To compare recovery assumptions against realised producer performance across a peer set, get in touch or ask Nara.
Related reading
Sources
- National Instrument 43-101 and Form 43-101F1, Item 13 Mineral Processing and Metallurgical Testing — Ontario Securities Commission
- CIM Definition Standards, 10 May 2014 — BCSC-hosted copy
- The JORC Code, 2012 Edition — jorc.org
This article is educational and is not investment advice. Mining Terminal is a data platform, not a broker, dealer or investment adviser.