The Top-Cut — Ep. 3 show notes

Special — Ramp-up Curves

Both hosts are synthetic voices and the script is AI-generated. Every paper cited was verified against a primary source (DOI or arXiv ID) before publication, but the commentary is not that of practising researchers.

The Top-Cut: Show Notes

Special: Ramp-up Curves, or How Fast New Mines and Plants Really Reach Nameplate

Traci and Joao take the reference-class idea from the last special into ramp-up. A feasibility study's ramp-up schedule is the inside view. The McNulty curves and RAND's process-plant studies are the outside view. The episode follows three narrative spines:

the Kumtor gold mill at 4,000 m;
the three Western Australian nickel laterite plants (Cawse, Bulong, Murrin Murrin);
Cerro Verde 2 against Constancia, Century and Red Chris.

Along the way the hosts argue over whether the McNulty curve is a law of nature or a budgeting convention. They finish by testing ten rules of thumb against Côté Gold, Red Chris, Las Chispas and Platreef.

About the McNulty-curve update: McNulty's 2024 and 2025 Springer conference chapters are not in hand. Statements about the modern update are attributed to Hatch's blog series The Ramp-up Series: Beyond the Curves. Parts 1 and 5 of that series are by Britt MacKinnon, Terry McNulty and Phillip Mackey.

Every quoted number was checked against the source PDF, or for actual outcomes against the company's own release or filing, before rendering. The fact log has page references (fact_check_log.md).

Both hosts are synthetic voices and the script is AI-generated. The commentary is not that of practising metallurgists or mining engineers. The rules below are a starting reference class for a reviewer, not engineering or investment advice.

The rules of thumb

Each rule gives the benchmark, then the sources behind it, then the conditions and how strong the evidence is.

1. Conventional greenfield concentrator: about a year to sustained nameplate, and 75–90% in year one

Benchmark: Plan for roughly 12 months to sustained nameplate and a first-year average of about 75–90% of design. A schedule at 100% in year one is an inside view.
Sources:
McNulty 1998: Series 1 was about 98% by month 12 and was achieved by 23 of 41 plants (56%). 11 of 13 sulfide flotation concentrators were Series 1.
Goode et al. 1991: "common to assume production of 75% to 90% of full capacity in the initial year" for simpler mills; underground mines often take longer to catch up.
Red Chris 2012 TR: Year 1 mill feed 10,952 kt, which is the full 10.95 Mt/a rate. Actual: 25,698 t/d from mill start to the end of 2015 and 26,371 t/d in 2016, against 30,000 t/d design (Imperial Metals production releases). That's about 86% and 88%.
Mount Milligan: production September 2013, nameplate 60,000 t/d in December 2015 (Yue 2017).
Conditions: Mature flowsheet, piloted or with good analogues, and an experienced owner. Near-clones of an existing operation can beat it: Cerro Verde 2 reached nameplate in 5.1 months. Hard or variable early ore can lose it: Mount Milligan.

2. Novel, hydrometallurgical or raw-solids pioneer: start at Series 3

Benchmark: Assume about 60% of design in year 1, about 75% in year 2 and about 82% in year 3 (Mackey & Nesset's reading of McNulty Series 3). Make the project argue its way up. Add about 10–11 weeks of start-up per commercially unproven process step.
Sources:
RAND 1986 (Myers et al.):
53 plants: median start-up 4 months and mean 8 (capped at 30). This is start-up to steady operation, not to design.
Pioneer plants on raw solid feed averaged 21.6 months, against 2.2 for standard plants.
"None of the 11 solid-feed pioneer plants … met their startup schedules."
Regression: about 10–11 weeks per unproven step, about 4 months for raw solid feed, and 5+ months without an integrated R&D, engineering and operations team.
RAND 1981 (Merrow et al.):
Months 7–12: 22 of 43 pioneer plants were below 85% of design.
15 solids plants averaged 49%, and only 2 of them reached 85%. Gas and liquid plants averaged 84%.
Wasmund et al. 2011 (Hatch): of 23 new-technology pyro and hydro plants, 7 succeeded (design within 3 years), 9 had difficult start-ups (more than 3 years) and 7 failed (abandoned).
WA HPAL trio:
Murrin Murrin made practically no product in year 1 and reached about 70% after 3.5 years (McNulty 2014).
Cawse took about 20 months from commissioning to design (Dobson, Young & O'Sullivan 2001).
Conditions: "Novel" means unproven steps and design data not taken from commercial plants, not merely new branding. McNulty (1998, 2014) and Hatch Part 5 both stress that well-piloted new technology can ramp up fast. Falcondo went from 59% to 99.7% in year 2 with a completely new process.

3. Price the ramp-up as money

Benchmark:
Over the first three years, Series 2 instead of Series 1 costs about 10% of capex, and Series 3 about 20%.
Start-up cost itself has a median of about 3.6% of capital and a mean of 5.5%. First-of-a-kind plants average about 9%; plants the firm has built before, under 2%.
Sources:
Campbell, McConaghy & Vardill 2004: a hypothetical laterite project at a 2%/quarter discount rate. A Series 4 project ends year 3 with more than $100M of operating losses on top of unpaid capital.
RAND 1986: Table 14 (n=51), Table 29 (first-of-a-kind 8.8%), Table 30 (built similar before 1.8%, not 7.1%).
Mackey & Nesset 2003: even a Type 1 ramp-up forgoes about a third of a year's production. Type 2 forgoes about three-quarters.
Second-hand, via McNulty 2004 and 2014: Halbe 2003, and Kennedy et al. 2011.
Conditions: The capex percentages come from hypothetical models. Use them as orders of magnitude.

4. Read three definitions before reading the curve

Benchmark: Establish four things before comparing a project with any curve:
What nameplate is based on: which ore, which grind and how much design margin.
What time zero is.
What the company means by "commercial production".
Over what window "nameplate achieved" was measured.
Sources:
Commercial-production thresholds in the cases read:
60% of design: New Afton 2012 (O'Hara et al.) and Côté Gold 2024 (IAMGOLD, 2 Aug 2024).
30 days at 75% plus 48 h at 100%: Carmen de Andacollo 2010 (Rondestvedt et al. 2012).
Two months at 80% plus over 85% recovery: Las Chispas 2022 (SilverCrest, 31 Jan 2023).
Constancia: design 76,000 t/d, but the grinding circuit could treat up to 86,000 t/d on the softer supergene ore of years 1–5, and it ran at 95,000 t/d with a coarser grind (Klohn et al. 2016).
McNulty's time zero moved from the end of construction (1998) to 6 months after the end of construction (2004), where commissioning couldn't be distinguished.
Conditions: Applies always. A fast ramp-up to a soft nameplate is weak evidence.

5. Find the slowest ramp-up: mine, plant or people

Benchmark: The binding ramp-up is usually the mine, especially an underground one, or the workforce. It is rarely the plant. A planned stockpile buffer is a good sign. A plant waiting on an unfinished shaft will run at whatever the hoist allows.
Sources:
Goode et al. 1991.
Las Chispas:
Q4 2022 plant throughput was 1,135 t/d against the FS plan of 1,143. December was 1,318 t/d, against 1,250 nameplate.
The mine ran 700 t/d against 750 in the original guidance.
The stockpile was 261,600 t against the FS estimate of 311,200.
Platreef: Phase 1 plant (0.77 Mtpa) produced first concentrate in November 2025, on the FS schedule, but was campaigned on development ore while hoisting was constrained. Q2 2026 milled 37,296 t, about a fifth of quarterly capacity by our arithmetic. Commercial production is now expected in Q4 2026 (Ivanhoe Q2 2026 MD&A).
McCarthy 2003, via Ward & McCarthy 1999: of nine Australian underground base-metal mines, half reached design by year 3 and a quarter never did.
Conditions: Most important for underground mines and for phased projects.

6. Discount the honeymoon

Benchmark: A sprint above design is not sustained capacity. Ask for 30- and 90-day averages, then ask again a year later.
Sources:
Ward & McCarthy 1999, via McCarthy 2003: most mines could run three months at 15% above design, but averaged about 92% of design over the longer term.
Hatch Part 9 (2026): Series 1-looking starts can drift to a lower curve after handover.
Collahuasi: concentrate output fell back below 100% after an early Type 1 start (Mackey & Nesset 2003).
Conditions: General.

7. Look for the people

Benchmark:
Operational readiness funded at about 0.5–1% of plant capex, starting about 18 months before first ore.
Experienced operators drawn from an analogous plant.
30–50% turnover in the first six months built into the plan.
A workforce estimate built from a genuine analogue.
Sources:
Tostengard 1988: 0.5–1% of plant capital; "good" 95% versus "fair" 83% first-year production in his scenarios.
Cerro Verde 2: readiness started 18 months before first ore; 50% of experienced C1 staff transferred; four months' training on C1 (Vanderbeek et al. 2017).
Shillabeer & Gypton 2003: turnover of 30–50% in the first six months and up to 20% over the next 12.
Tropicana: 14 weeks to recruit a professional plus about 3 months to become effective. The engineering group was expected to be fully effective only 17 months after mining started (Chow et al. 2013).
Cawse: the FS workforce of 135 was "based on a standard gold plant model due to a lack of like-project". It recruited 153 and later added 26 positions (Dobson et al. 2001).
RAND 1986: pioneer plants averaged 7.5 months start-up with a representative team, against 17.0 without.
Conditions: Strongest evidence in RAND's regression. The rest is case histories.

8. Look at the contract and the owner

Benchmark: A lump-sum, fixed-price EPC on new technology with no owner engineers in the design office is a Series 3–4 flag. A strong owner's team that includes future operators is a Series 1–2 flag.
Sources:
Nice 2003, on the WA nickel trio:
Murrin Murrin was built under EPC with no owner technical staff in the engineer's offices, part of the process not finalised at award, and commissioning about 7 months late.
Cawse was EPCM with a strong owner's PM. It was the only one of the three to reach practical completion on the contracted date, at a final cost below its definitive estimate.
Wasmund et al. 2011: fixed-price contracts on new technology lead to disputes over who pays for inevitable post-start-up modifications.
Twigge-Molecey 2003, via McNulty 2004.
Hatch Part 7, on the accountability gap.
Conditions: Evidence is mostly case-based.

9. Make sure the balance sheet survives the slow case

Benchmark: Finance to the slow case, with a separate overrun facility.
Sources:
Nice 2003: Cawse reached design (May 2000), and its owner entered receivership in March 2001. Bulong was taken over by bondholders and then went into receivership. Capex overruns against the FS were 17–28%.
AMC magnesium: 11 years and about $200M of testing; McNulty rated its ramp-up Series I; abandoned before start-up after ownership change and an EPCM-to-lump-sum conversion (Wasmund et al. 2011).
Castle 1985: 14 of 18 projects had difficulty reaching design production. Lenders were protected only by sponsors' completion undertakings.
Conditions: Most acute for juniors and for bond- or project-financed builds.

10. Use the curves as the prior, never as the forecast

Benchmark:
Pick the narrowest honest reference class: copies of an existing plant, conventional concentrators, or novel hydromet plants.
Anchor on its typical outcome.
Adjust with the factor list, in writing.
Never put Series 1A in a base case.
Sources:
McNulty 2014: Series 1A is the best fit to the best 23 of 81 plants. For the Freeport data, "only those which were superior to the Series 1 average were considered".
RAND 1986: planned start-up time predicts actual start-up time poorly (Fig. 8).
Flyvbjerg & Gardner 2023, the framework from Ep. 2.
Conditions: Applies always.

Test cases: FS ramp-up schedules against what happened

|---|---|---|---|

| Côté Gold (IAMGOLD, Ontario; HPGR, 36,000 t/d) | 2018 FS: 10 months to design throughput. 2022 TR update: 20 months, with 90% after 10 months. The same HPGR-precedent paragraph appears in both. | First gold 31 Mar 2024. Commercial production Aug 2024 (30 days at ≥60%). 90% monthly average in March 2025. 36,000 t/d over 30 days on 21 Jun 2025, about 15 months. | Reality landed between the two studies, about 2 months behind the 2022 study's 90% milestone. |

| Red Chris (Imperial, BC; 30,000 t/d) | 2012 TR: full 10.95 Mt/a in Year 1, so no ramp-up. | Mill from 17 Feb 2015. 25,698 t/d average in 2015 and 26,371 t/d in 2016. | About 86% in year 1 and 88% in year 2, against a 100% plan. |

| Las Chispas (SilverCrest, Sonora; 1,250 t/d) | 2021 FS: stockpile ahead of start-up; gradual underground ramp-up to 2025. | Commercial production 1 Nov 2022. Q4 plant 1,135 t/d against a 1,143 plan; December 1,318 t/d. Mine 700 t/d against 750 guidance. Stockpile 261.6 kt against 311.2 kt. | Plant on plan, mine slightly behind, and the buffer did its job. |

| Platreef (Ivanhoe, South Africa; Phase 1 0.77 Mtpa) | IDP25 FS: Plant 1 Q4'25, Shaft 3 hoisting from Feb 2026, mine steady state 4.07 Mtpa in Q4'29. | First concentrate Nov 2025. Q2 2026 milled 37,296 t on development ore. Shaft 3 completed on schedule. Commercial production expected Q4 2026. | Plant on time, mine-constrained. The slow ramp-up is the shaft and the stopes, not the mill. |

| Grota do Cirilo (Sigma Lithium, Brazil; 1.5 Mtpa DMS) | Jan 2023 TR: the Xuxa mine schedule runs at about 0.12–0.14 Mt a month from Year 1, about the full rate. | Not verified. We didn't check Sigma's reported production against the plan. | Off air. Listed so the gap is visible. |

Corrections and inconsistencies found along the way

Hatch Part 2 misreads RAND 1986. "4 months (median) to 8 months (mean) … 3.6% to 5.5%" pairs the medians and means of two separate distributions: time (n=53) and cost (n=51). It is not a causal before-and-after.
McNulty 2014 on Wasmund: McNulty says the failures were "39% of the database". Wasmund et al. 2011 give 7 of 23, which is 30%.
McNulty 2004: says the Series 1 average was 18 months to design. The 1998 Fig. 1 has Series 1 at about 98% by month 12.
Brief vs source: the brief described a "~140-case" update. Hatch Part 5 says "more than 150" projects. The episode uses Hatch's figure.
Côté 2022 TR: one passage (p.~1487 of the text layer) still says a 10-month ramp-up. Sections 1 and 17 say 20 months.
Carmen de Andacollo: the paper's abstract says commercial production was October 2010. The body says "October 1st, 2011" but also "7.5 months" from 15 February. 2010 is consistent.
Authorship: the "Gypton 2003" item in the library is Shillabeer & Gypton 2003. The Cawse paper is by Dobson, Young & O'Sullivan.

Bibliography

Framework and reference-class studies

Merrow, E.W., Phillips, K.E. & Myers, C.W. (1981). Understanding Cost Growth and Performance Shortfalls in Pioneer Process Plants. RAND R-2569-DOE.
Merrow, E.W. (1983). Cost Growth in New Process Facilities. RAND P-6869.
Myers, C.W., Shangraw, R.F., Devey, M.R. & Hayashi, T. (1986). Understanding Process Plant Schedule Slippage and Startup Costs. RAND R-3215-PSSP/RC.
McNulty, T.P. (1998). Developing innovative technology. Mining Engineering 50(10): 50–55.
McNulty, T.P. (2004). Minimization of delays in plant startups. In Management of Process Technology Development, SME, pp. 113–120.
McNulty, T. (2014). Plant ramp-up profiles: an update with emphasis on process development. COM 2014, Conference of Metallurgists, CIM.
Mackey, P.J. & Nesset, J.E. (2003). The impact of commissioning and start-up performance on a mining/metallurgical project. 35th Canadian Mineral Processors, paper 21, pp. 331–347.
Wasmund, B., Voermann, N., Haneman, B., Sarvinis, J. & Sheehan, G. (2011). Implementing new technologies in metallurgical processes: building plants that work. COM 2011, CIM/MetSoc, pp. 37–54. A CIM version is also in the library.
Campbell, F., McConaghy, E. & Vardill, W. (2004). Startup and reliability of nickel laterite plants. International Laterite Nickel Symposium, TMS, pp. 25–42.
Nice, R.W. (2003). Metallurgical processing plant ramp-up periods: how to reap the 'rewards'. Mining Risk Management Conference, AusIMM, Sydney, pp. 257–263.
Tostengard, G.R. (1988). Gold plant start-up performance: the financial impact. Gold Mining 88, SME, ch. 12.
Flyvbjerg, B. & Gardner, D. (2023). How Big Things Get Done. Framework only; see Ep. 2.

Hatch, The Ramp-up Series: Beyond the Curves (blog)

Part 1: MacKinnon, B., McNulty, T. & Mackey, P. (3 Oct 2025). An update and new perspectives on the McNulty curves for capital project success.
Part 2: Bouchard, L. et al. (24 Feb 2026). Rethinking risk to accelerate ramp-up.
Part 3: MacKinnon, B. & Lilley, J. (14 May 2026). Setting the stage for success.
Part 4: MacKinnon, B., Cajueiro, A. & Hill, K. (2 Jun 2026). Enabling strong ramp-up performance with OR.
Part 5: MacKinnon, B., McNulty, T. & Mackey, P. (18 Jun 2026). McNulty curves: new perspectives to enable trust, transformation and technology.
Part 6: MacKinnon, B., McNulty, T. & Mackey, P. (15 Jul 2026). Phronesis at scale.
Part 7: Villanueva, M. & MacKinnon, B. (18 Aug 2026). Accountability without a home.
Part 8: MacKinnon, B. & Hocking, R. (23 Sep 2026). Reality check: reliability matters.
Part 9: MacKinnon, B. & Lombard, J. (27 Oct 2026). Ramp-up is an operating outcome.

Feasibility-study accuracy

Castle, G.R. (1985). Feasibility studies and other pre-project estimates: how reliable are they? Finance for the Minerals Industry, AIME.
Prentice, P.O. (1990). The fall and rise of the feasibility study. AusIMM.
Goode, J.R., Davie, M.J., Smith, L.D. & Lattanzi, C.R. (1991). Back to basics: the feasibility study. CIM.
McCarthy, P. (2003). Managing technical risk for mine feasibility studies. Mining Risk Management Conference, AusIMM, pp. 21ff.
Shillabeer, J. & Gypton, C. (2003). Highlighting project risk following completion of the feasibility study. Mining Risk Management Conference, AusIMM, pp. 101ff.
Mackenzie, W. & Cusworth, N. (2007). The use and abuse of feasibility studies. AusIMM. An SAIMM version is also in the library.
Mackenzie, W.R. & Cusworth, N. (2016). The use and abuse of feasibility studies: has anything changed? Project Evaluation Conference, AusIMM, Adelaide.
Bullock, R.L. (2011). Accuracy of feasibility study evaluations would improve accountability. Mining Engineering, April 2011, pp. 78–85.

Start-up case histories

Kazakoff, J. & Mortimore, A. (1999). The Kumtor gold mill start up and start up challenges. 31st Canadian Mineral Processors, paper 23.
Dobson, T., Young, R. & O'Sullivan, D. (2001). Commissioning under a microscope: the Cawse nickel start-up. AusIMM Proceedings 306(1): 27–33.
Burgess, F., Reemeyer, L., Spagnolo, M., Ashley, M. & Brennan, D. (2003). Ramp up of the Pasminco Century concentrator to 500 000 tpa zinc metal production in concentrate. Eighth Mill Operators' Conference, AusIMM, Townsville, pp. 153ff.
Vanderbeek, J., Gelfi, P. & Enriquez, J. (2017). The Cerro Verde 2 concentrator: design, start up and operation. SME Annual Meeting, Denver, preprint 17-115.
Klohn, S., Stephenson, D. & Granados, H. (2016). Constancia project process plant design and start up. 48th Canadian Mineral Processors, Ottawa.
Rondestvedt, C.S.R., Bustos, M. & Torres, L. (2012). Carmen de Andacollo: start-up of Teck's new copper concentrator. CIM.
O'Hara, C., Swedburg, K., Roy, S. & Katchen, J. (2015). Start-up and early optimization of the New Afton concentrator. CIM.
Yue, J. (2017). An update on reaching nameplate throughput and continuous improvement at Mount Milligan. 49th Canadian Mineral Processors, pp. 142ff.
Chow, J., Fitzpatrick, S., Greenup, D. & Lloyd, S. (2013). Tropicana Gold Mine: mitigating the challenges of a mining start-up. World Gold Conference, AusIMM, Brisbane, pp. 51ff.
Also read but not used on air:
Sleeman 1976 (Sept-Îles)
Dunne et al. 1999 (Cadia)
Andreatidis et al. 2000 (Cannington)
Hart 2012 (Boddington HPGR)
Abols 2015 (Eagle)
Berton et al. 2016 (Canadian Malartic)
Lett et al. 2016 (Cadia East PC1)
Lovitt & Hewitt 2005 (Northparkes Lift 2)
Salvador 2023 (Quellaveco)
Rushton 1978
Fletcher 1981
Agarwal 1983
Smith et al. 1987
Rajan 2000
Broad 2013
Anon. 2016 (SME, stoping ramp-up model)

Technical reports and company disclosures (test cases)

IAMGOLD (Nov 2018). Côté Gold Project NI 43-101 Feasibility Study, §17.3.9.
IAMGOLD / SLR (12 Aug 2022). Côté Gold Project NI 43-101 Technical Report, §§1, 17.3.9, 17.5.
IAMGOLD news releases: first gold pour (31 Mar 2024); commercial production (2 Aug 2024); "Côté Gold achieving nameplate throughput" (23 Jun 2025, SEC Form 6-K exhibit 99.1).
Imperial Metals (2012). Red Chris technical reports (production schedule in §§16 and 29).
Imperial Metals news releases: 2015 production; 2016 production.
SilverCrest Metals (Jan 2021). Las Chispas NI 43-101 Feasibility Study.
SilverCrest Metals news release, 31 Jan 2023 (SEC Form 6-K exhibit 99.1).
Ivanhoe Mines (Mar 2025). Platreef IDP25 NI 43-101 Technical Report, §16.
Ivanhoe Mines, MD&A for the period ended 30 Jun 2026.
Sigma Lithium (Jan 2023). Grota do Cirilo Updated NI 43-101 Technical Report. Schedule reviewed; actuals not verified.

Cited from verified metadata only (full text not read)

McNulty, T., MacKinnon, B. & Mackey, P. (2025). McNulty curves: an update and new perspectives. 12th International Copper Conference (COPR 2025), Springer. doi:10.1007/978-3-032-00102-3_102. Summarised via Hatch Parts 1 and 5.
McNulty, T. & Parameswaran, K. (2024). COM 2024, Springer. Not in hand.
Beauprie, R. & McKen, A. (2025). Côté Gold: from first gold to commercial production and beyond. 57th Canadian Mineral Processors. Cited in Hatch Part 1; not read.

Cited second-hand (via a paper we did read)

Agarwal & Katrak (1983) and the Charles River Associates / World Bank start-up study, via McNulty 2004, Mackey & Nesset 2003, McCarthy 2003 and Tostengard 1988.
Halbe (2003) and Kennedy, Harris & MacRae (2011) NPV comparisons, via McNulty 2004 and 2014.
Partanen et al. (2010), on the Tenke Fungurume start-up, via McNulty 2014.
Ward & McCarthy (1999), on nine Australian underground base-metal mines, via McCarthy 2003.
IPA (2015), about 30 mining projects from 2006 to 2013, months 7–12 output about 20% below target, via Mackenzie & Cusworth 2016. Not used on air.
Fulks (1982), saying inexperienced organisations take 40–50% longer, via Myers et al. 1986.
UBS Warburg (2001), saying ramp-ups at Century's scale typically take up to four years, via Burgess et al. 2003.

Not in hand (and not cited on air)

Rollins & Lam 2017
ALTA HPAL papers (Valle et al. 2017; Ambatovy; Ramu)
Merrow, Industrial Megaprojects
Bertisen & Davis 2008

Glossary

Nameplate / design capacity: the throughput the plant was designed for, usually in t/d or t/a.
Commercial production: a company-defined milestone, often for accounting. In these cases it ranged from 60% to 80% of design.
McNulty Series 1–4: four empirical ramp-up profiles (1998) from fast (1) to slow (4). Series 1A (2014) is a best fit to the best performers.
HPAL / PAL: high-pressure acid leaching of nickel laterite in titanium-lined autoclaves.
CCD: counter-current decantation, the thickener train that washes dissolved metal from leach residue.
Preg-robbing: natural carbon in ore adsorbing dissolved gold during cyanide leaching.
HPGR: high-pressure grinding rolls.
Operational readiness (OR): the people, systems and procedures needed to run a new asset from day one.
EPC / EPCM: engineer–procure–construct (often lump-sum) versus engineering, procurement and construction management (reimbursable, owner bears more risk).