Nelson Complexity Index: Why Refinery Complexity Matters
How the NCI is built from relative build cost, what complexity buys when heavy crude trades at a discount, and why a higher score is not automatically better.
Educational analysis for professional use. This guide is not investment advice or a recommendation to buy or sell any security, and it is not personalised.
A High Score Means an Expensive Refinery, Not a Good One
The Nelson Complexity Index gets read as a quality ranking, and it is not one. It is a cost measure. A refinery scoring 14 did not necessarily earn more than one scoring 7; it cost roughly twice as much to build per barrel of capacity, and whether that was money well spent depends entirely on what crude prices did afterwards.
What the index actually counts is secondary processing capacity: the units that sit behind the crude distillation column and break heavy molecules into light ones. A refinery with plenty of it can buy cheap, dense, sulphurous crude and still produce gasoline and diesel. A refinery without it is stuck buying the expensive light sweet grades that come out of the distillation column already close to the products it needs to sell.
So complexity is optionality, and the optionality is bought with capital. Hold that thought through everything below, because it is the half of the index that gets left out.
How the Index Builds
| Component | Role in the index |
|---|---|
| Crude distillation unit (CDU) | Base factor of 1.0 |
| Each downstream unit | Build cost relative to the CDU × that unit’s capacity as a % of CDU capacity |
| Refinery index | Sum of all unit contributions |
Nelson’s method is cost-based, not a count of units. The relative build costs do the work: a fluid catalytic cracker, which breaks heavy gas oil into gasoline, costs five to seven times as much per barrel of daily capacity as an atmospheric distillation unit, so it carries a heavy factor. A hydrotreater, which strips sulphur, is far cheaper and carries a light one.
The capacity weighting matters as much. A unit sized to handle half the crude the refinery runs contributes half its full factor, so two plants with the same list of units can score differently.
Because the whole thing is proportional to cost, the ratio between two scores is roughly the ratio between two construction bills. Oil & Gas Journal ran that arithmetic itself: a refinery at 12 should cost about 26% more to build than a similar-sized plant at the US average of 9.5, which is simply 12 ÷ 9.5. That is the cleanest way to read any index figure you come across.
What the Numbers Look Like
Oil & Gas Journal put the average US refinery at 9.5 in 1996, and that figure is still the one most often quoted. Treat it as the middle of the distribution rather than as this year’s average: the US system has kept adding conversion capacity since, so the centre of gravity has moved up.
The spread around it is wide, and the EIA has published two plants that show it inside one company’s system. One in Washington state scored 7.0, running a catalytic cracker, alkylation and hydrotreating. One in Los Angeles scored 14.1, adding hydrocracking, reforming and coking. Same operator, double the index, because of the units in the second list.
Practitioner shorthand, not a filed standard, puts simple hydroskimming refineries below 7 and complex coastal plants at 12 to 15 and above. A hydroskimmer runs distillation and treating with almost no conversion behind it, which means its yield pattern is decided by whatever crude it buys and it lives close to the headline 3-2-1 crack.
What Complexity Buys, and When It Pays Nothing
The return on complexity is the price gap between heavy and light crude, and nothing else. If a barrel of heavy sour crude yields the same slate of products after a coker as a barrel of light sweet does without one, the complex refinery keeps whatever discount the heavy barrel was trading at.
That gap moves a long way. Western Canadian Select, the heavy Canadian marker, has traded around $12 a barrel under WTI, and our planning deck runs a deliberately wider $15 because a fat discount is the conservative case for a heavy-oil producer and the generous case for a refiner. It has been under $5 in tight years and far wider than $15 in loose ones.
Put that against the capital. Take a 200,000 b/d refinery and imagine lifting it from 7 to 12 by adding coking and hydrocracking. On the cost relationship above, that is 12 ÷ 7, or roughly 70% more capital per barrel of capacity, spent once and carried forever.
| Wide differential | Narrow differential | |
|---|---|---|
| Heavy discount captured | $10/bbl | $3/bbl |
| Heavy barrels run (half the slate) | 100,000 b/d | 100,000 b/d |
| Barrels a year | 36.5M | 36.5M |
| Gross feedstock saving | $365M | $110M |
The arithmetic is 100,000 × 365 × $10 = $365M, and the same volume at $3 gives $109.5M. Between those two columns nothing about the refinery changed. The coker cost what it cost, it burns the same energy and takes the same maintenance either way, and in the right-hand column it is a very large piece of capital earning very little.
This is why complexity is worth understanding rather than scoring. Ask what differential the plant was built for, then ask where the differential actually is.
Filed Margins Are Not a Complexity Ranking
Refiners do not publish plant-level index scores in their results, so the temptation is to read the margin table as a complexity table. Resist it.
| Company | Throughput (mb/d) | FY2025 margin ($/bbl) | Note |
|---|---|---|---|
| MPC | 2,989 | $16.87 R&M | Gulf Coast weighted, 94% utilisation |
| DINO | 652 | $15.37 adj. gross | Inland Rockies and Mid-continent |
| VLO | 2,988 | $12.29 | Merchant pure-play; the margin is already struck after RIN compliance cost |
| PSX | 1,882 inputs | $10.88 realised | Integrated; refining is one segment among several |
| PBF | 832.9 | $7.72 ($8.77 ex specials) | Coastal; a fire at its Martinez plant |
PBF sits at the bottom of that table and runs coastal refineries, which is exactly the configuration the complexity bands call sophisticated. What separated it from the top was an event year and its regional exposure, not a thin unit list. Meanwhile each of these margins is struck against a different internal benchmark, which is a second reason the ranking does not mean what it looks like. Sorting that out is the job of capture rate.
Geography Overlays Complexity
Where a refinery sits can matter more than what it contains, because the index says nothing about who it can sell to.
Par Pacific processed 187.8 mb/d across four refineries totalling 219,000 b/d of rated capacity: Hawaii 94,000, Tacoma 42,000, Billings 63,000 and Newcastle 20,000. Island and Pacific Northwest markets do not clear at a Gulf Coast price, and a captive market can be worth more than a conversion unit. HF Sinclair runs a 678,000 b/d inland system, 652 mb/d of throughput in FY2025, and carries a lubricants business alongside it that turned $2,519M of external revenue into $261M of segment EBITDA, earnings that do not move with the crack at all.
None of that shows up in an index score. Complexity tells you what crude a plant can digest. It does not tell you what its products fetch, or whether anything else in the group is holding the earnings up when cracks compress.
Complexity scores tell you which crudes a refinery can run. The primer prices that difference across three worked mid-cycle DCFs.
The Excel model is the primer's three refiner DCFs live across 12 sheets: change the crack spread, capture rate or exit multiple and the valuation moves.
Frequently Asked Questions
- What is the Nelson Complexity Index?
- The Nelson Complexity Index (NCI) is a cost measure. The crude distillation unit is set at 1.0, every other processing unit is scored by what it costs to build relative to that unit, and each score is multiplied by the unit's capacity as a percentage of crude distillation capacity. Add the results together and you have the refinery's index. It is a proxy for what the plant cost per barrel of capacity, and for how much heavy or sour crude it can turn into gasoline and diesel. It is not a margin and not a quality score. Oil & Gas Journal put the US average at 9.5 in 1996.
- Why does refinery complexity matter for valuation?
- Complexity is optionality bought with capital. A plant with coking and hydrocracking can buy heavy or sour crude at a discount and still make the same light products, so it earns that discount; a simple refinery has to pay up for light sweet crude. But the capital is spent whether or not the discount is there. Oil & Gas Journal's own arithmetic puts a refinery at 12 on the index at roughly 26% more expensive to build than one at the 9.5 US average. When heavy and light crude trade close together, the complex plant carries that cost for very little return, which is why a higher index is not automatically a better investment.
- What is a typical Nelson Complexity Index for a US refinery?
- Oil & Gas Journal put the US average at 9.5 in 1996, which is still the figure most often quoted. Practitioner bands place simple hydroskimming refineries below 7 and complex coastal plants at 12 to 15 and above. The EIA has published two worked examples that show the spread inside a single company's system: a Washington state refinery at 7.0, running a catalytic cracker, alkylation and hydrotreating, and a Los Angeles refinery at 14.1, which adds hydrocracking, reforming and coking.