Digital Exclusive: Margin left on the table: The hidden cost of supply chain capability gaps in hydrocarbon processing—Part 1

S. DUBEY, AVEVA, Singapore

The global hydrocarbon processing industry—refining, LNG and naphtha-based petrochemical operations—has invested heavily in supply chain systems: planning models, scheduling systems, trading and commercial systems. A new wave—digital twins, generative artificial intelligence (AI), knowledge graphs and AI agents—promises to accelerate digital transformation amid today’s volatility.

Yet, for most organizations, the expected margin uplift remains stubbornly elusive. Publicly traded refiners often publish their benchmark indicators (e.g., the 3-2-1 crack spread, the Singapore Complex Margin) only to report realized margins that do not move in lockstep. The common refrain during quarterly earnings calls is a list of explanations for this gap: “prices moved,” “the unit tripped,” “demand was soft.” Without a causal model that traces margin leakage to its organizational roots, the same patterns repeat quarter after quarter, despite continued investment in digital transformation. This is not unique to the hydrocarbon processing industry. Bain & Company’s 2024 survey of over 400 executives found that only about 12% of business transformations achieve their original ambitions.¹

This article provides that causal model. Part 1 begins by examining the structural challenge of hydrocarbon supply chain decisions and traces how these challenges produce four recurring patterns of margin leakage. It then explains why the reflexive industry response—aspirational and technology-driven digital transformation without diagnostic grounding—is not sufficient. Part 2 introduces a supply chain capability framework that provides that grounding, shows how to diagnose capability gaps systematically and outlines how to act on those diagnoses.

The structural challenge of hydrocarbon supply chain decisions. Before one can understand the supply chain capability framework, it is important to appreciate why supply chain decisions in these operations are structurally difficult to get right, and even harder to get right consistently. Three realities define these supply chains:

1. Routine decisions with enormous financial consequences and narrow windows. In many industries, multi-million-dollar decisions are strategic events. In these operations, they are part of the daily routine. A single crude cargo purchase is a $100 MM–$200 MM commitment. A single LNG cargo diversion to capture a European price spike can be worth tens of millions of dollars. For both examples, the opportunity might be presented unexpectedly and with a narrow window. A blending decision on a single batch can cost tens to hundreds of thousands of dollars in quality giveaways. These are not strategic bets reviewed by a board committee, they are routine operating decisions made on a daily or even shift-by-shift cadence.
2. The commitment horizon exceeds the confidence horizon. The hydrocarbon value chain is long and economically coupled. Feedstocks must be selected months before the market conditions under which it will be processed and sold can be known. This is not just transit time; it is the full cascade of commitment lead time, processing time and placement time. The decision to buy one grade of feedstock over another is made today; the economic consequence is revealed 90 days later, after the yield is realized and the product is placed into a market whose prices have moved to a level no one could have reliably forecasted. Each decision narrows what is possible downstream. Feedstock selection constrains yields, yields constrain the product slate, the product slate constrains placement, placement constrains logistics and logistics constrain inventory. By the time the organization is processing and placing products, its degrees of freedom have been substantially determined by commitments made weeks or months earlier under conditions that no longer prevail.
3. Multi-dimensional uncertainty. The hydrocarbon processing industry faces uncertainty from all sides simultaneously. Input prices (feedstock), output prices (products) and the spread between them (margins) all move independently. Yields vary with feedstock quality, catalyst or equipment state, and operating severity. Logistics are subject to weather, port congestion and vessel availability. These uncertainties are not just independent; their correlation structure varies with the nature of the disturbance. A crude price spike driven by a supply threat may coincide with a freight rate surge (same geopolitical trigger) and a crack spread widening (downstream supply fear), or with a crack spread narrowing (if crude rose but product demand is unchanged). The correlation structure is not static; it reconfigures depending on the nature and source of the disruption.

These characteristics produce a set of conditions that complicate decision-making:

• Input unreliability: The inputs on which decisions depend are directly subject to the underlying uncertainty, making them inherently unstable and rapidly stale.
• Decision latency: The time required to assemble inputs, convene the right people and reach a coordinated decision exceeds the window in which the decision has value.
• Decision fragmentation: Given these conditions, each function narrows its aperture. Trading watches the forward curve but loses track of what the plant can actually produce this month. Planning builds a monthly model but cannot track the intra-month moves that change its economics. Operations watch the current shift but has no visibility into the market drivers that informed the plan. Each function makes rational decisions with the slice of information it has, but the slices do not overlap enough to produce coherent value chain outcomes.

This is further complicated by weaknesses common to any industry—person-dependence, process inconsistency and absence of feedback—whose effects are well understood. The next section traces where the resulting value loss shows up.

Where the margin leaks: Four patterns of value loss. The structural challenges described in the preceding section produce recurring, measurable value loss across planning, execution and commercial decisions. Four patterns appear consistently:

1. Phantom margin—Value planned but not realizable. A plan-to-actual deviation of ~10% is not unusual in complex operations. To illustrate, for a 400,000-bpd refinery operating at a $10/bbl gross refining margin, this translates to approximately $130 MM/yr of margin that appears in the plan but is not realized in execution. When the plan misses, the explanation offered to investors is that prices moved and units underperformed; internally, the model gets the blame. It would be simplistic to stop there. Deeper probing always reveals questions the plan did not ask:
   • How much of the product slate has already been hedged, and at what economics?
   • What is the uncertainty bandwidth around the price deck, and does the plan treat all prices as equally reliable when some carry far more uncertainty than others?
   • What volumes in the plan are actually committed vs. still to be procured, and what is the margin exposure if those uncommitted volumes come in at different grades, prices or timing than assumed? Uncommitted volumes represent procurement risk if left unaddressed, but also optionality to select the best available grade and timing—a duality the plan must account for, not ignore.
   • What product placement flexibility exists at execution? Can volumes swing between domestic and export to capture differential moves? A plan that locks placement with no flexibility left for execution leaves margins exposed to adverse movements with no room to recover.

The gap reflects a more fundamental breakdown in how the decision is made: functions planning from different assumptions, uncertainty not sized in the plan, plans not challenged cross-functionally and lessons from past misses not fed back into the next planning cycle. No single input is wildly wrong. It is the compound effect that produces the gap.

2. Missed opportunities—Value available but not captured. Markets do not wait for decision cycles. Crack spread spikes appear and close within days. A competitor outage opens a product placement window that lasts a week. The organization sees these signals—traders watch screens, analysts read reports, dashboards flash alerts. The constraint is not signal visibility but decision latency, assembling a cross-functional decision view and acting within the available window. Acting on a crack spike requires assembling information distributed across functions:

• Does the plant have the feedstock to capture it?
• Is there tank space? Can the incremental volume be placed?
• Is counterparty credit available, or has the preferred counterparty hit its exposure limit?

Each question lives in a different system, owned by a different function. By the time the picture is assembled, the market has moved. The West Texas Intermediate (WTI) 3:2:1 crack spread moved by more than $2/bbl in a single day on 27 occasions in 2024–2025.^a On 16 of those occasions, the move reversed the next day. Even partial capture of such windows represents significant incremental margins—margin that was available, visible and not captured.

3. Defensive operations—The cost of self-protection. These structural challenges not only produce planning failures and missed windows, they also produce a quieter, more persistent consequence: each function learns to protect itself.

• Inventory buffers grow because no one trusts the supply plan enough to run lean, and no one analyzes whether a more reliable supply would cost less than the buffer it replaces.
• Throughput or yields are de-rated because the balance between profitability and reliability is not constantly updated based on real-time signals, but rather on a static number negotiated a year earlier.
• Conservative blending gives away quality because the penalty for off-spec is visible and immediate, while the cost of giveaway is diffuse and quarterly.

These buffers accumulate across the system, converting uncertainty into structural costs. The result can be significant: 2–5 excess inventory days representing $50 MM–$150 MM in locked working capital for a large refinery system, throughput and yields 3%–5% below demonstrated capability and logistics costs of 5%–10% above what a coordinated system would require. These costs recur every period because the defensive posture is structural.

4. Repeated leakage—The cost of not learning. The organization repeats the same leakage patterns because it lacks a systematic mechanism to decompose outcomes into attributable causes. Post-mortems—if they happen—produce explanations, not diagnoses that recalibrate assumptions, models or decision routines. Without that attribution, learning remains informal, and margin loss becomes structural rather than episodic.

Aspirations and technology are not enough. The natural organizational response to these problems is familiar. Call for better agility, so the supply chain can respond faster to market shifts. Better resilience, so it can absorb disruptions without breaking. Better visibility, so everyone can see the same picture. These are sensible aspirations.

The challenge is that they describe the destination without decomposing the capabilities required to reach it.

An organization assessed as having “low agility” knows it has a problem but does not know where the problem originates. Is it a sensing gap—the organization does not detect market changes quickly enough? Is it a planning gap—it detects the signal but cannot translate it into a revised plan within the opportunity window? Is it a governance gap—it knows what to do, but the person with authority lacks the cross-functional picture to act on? Is it an execution gap—it has the plan, but cannot coordinate across functions to implement it within the decision window? Each of these root causes requires a fundamentally different intervention. Calling all of them “low agility” conflates distinct capability gaps that must be diagnosed and addressed at different layers of the organization.

Technology creates a different version of the same problem. Most organizations already have planning models, scheduling systems, process optimization and market intelligence. A new wave—autonomous operations, digital twins, generative AI, knowledge graphs and AI agents—promises to push the frontier further. Each promises transformative value. Each will deliver on that promise only if guided by a clear understanding of which capability gaps it is meant to close, how those gaps should be filled and in what sequence.

Without that guidance, the results are predictable. An AI agent monitoring execution deviations, but with no understanding of their supply-chain-wide impact, will flag every threshold breach indiscriminately—a more sophisticated version of the alarm floods that control rooms have spent decades trying to eliminate. A digital twin built from data feeds that no one has assessed for reliability will propagate bad data. A generative AI summarizing a planning basis that was never cross-functionally validated will produce confident summaries of an incoherent picture. The technology is not the problem. Digital transformation programs face well-documented challenges,¹ from organizational behavior to middle management involvement. However, in refining, petrochemicals and LNG operations, the structural challenges described above are more fundamental determinants of success than the generic transformation challenges: technology deployed without a diagnostic framework that identifies the underlying capability gaps will optimize the pieces while the whole continues to underperform.

Aspirations tell the organization where it wants to be. Technology provides tools to get there. Neither provides the route. What is needed is a structured set of capabilities that, when present and functioning, produce the supply chain excellence these aspirations describe: the ability to plan with confidence, execute with agility, sense and respond between cycles, and learn from every outcome. Part 2 introduces that framework—and shows how to use it.

NOTE

^a Authors’ calculations using WTI Cushing spot, NY Harbor Conventional Gasoline spot and NY Harbor No. 2 Heating Oil spot prices from the U.S. Energy Information Administration (EIA), January 2024–December 2025

LITERATURE CITED

¹ Bain & Co., “88% of business transformations fail to achieve their original ambitions; those that succeed avoid overloading top talent,” April 15, 2024, online: https://www.bain.com/about/media-center/press-releases/2024/88-of-business-transformations-fail-to-achieve-their-original-ambitions-those-that-succeed-avoid-overloading-top-talent/

ABOUT THE AUTHOR

Satyendra Dubey is an industrial software and consulting professional with > 30 yrs of experience spanning oil and gas, chemicals and energy domains. His career has taken him from process engineering at greenfield refineries to global product leadership for supply chain and enterprise software businesses. He has presented at numerous industry forums and has published several articles in industry publications. His current interests lie in connecting industrial AI, optimization and knowledge graphs with insights from behavioral economics and organizational capability theory to improve how industrial enterprises make supply chain decisions. He earned a BTech degree in chemical engineering from IIT Kharagpur.

Related News

From the Archive

Comments

Comments

{{ error }}

Add Comment

{{ comment.name }} Editor • {{ comment.dateCreated | date:'short' }}

{{ comment.text }}