A previous white paper introduced Summit Ridge Refining, a fictional refinery presented as a well-run organization. Its reliability, safety record and planning capabilities reflected years of investment in people, technology and business processes. Even so, some opportunities identified through its planning process failed to reach execution.

Leadership examined where value was being lost. Planning knowledge was concentrated among a few people; planners and operations personnel sometimes understood constraints differently; temporary workarounds had become routine; and digital tools could speed up work without correcting weaknesses in the process. Summit Ridge brought planning and operations into closer agreement on constraints, challenged long-standing assumptions, and strengthened how recommendations were reviewed before execution. Those changes helped it act on more credible opportunities and improve profitability.

Summit Ridge had strengthened its planning process, yet its next vulnerability was already taking shape: Several of the planners who made it work were nearing retirement or moving into other roles. They knew which LP relationships deserved a closer look, which assumptions could be trusted, and how to explain an apparent opportunity to operations. While the LP model would continue to run after they left, their experience and judgment would not automatically pass to other planners. Summit Ridge needed a way to examine more of each LP result within the planning cycle and make the reasoning behind sound recommendations available across the planning team.

To see why an LP result requires interpretation, consider the work that begins when the model produces a result. The model might recommend a different crude slate, higher utilization of one process unit, lower severity in another, or a change in product disposition. The result indicates where economic value may lie. The planner must uncover what is creating that value and determine whether the opportunity can survive closer scrutiny.

That analysis might involve comparing a base case with an optimized case, tracing changes in the objective value, identifying active constraints, and testing the result against different prices or assumptions. The planner also must consider whether the physical refinery can execute the recommendation, how it would affect other units or product commitments, and what risks should be communicated to operations and leadership. Each answer could prompt further analysis or cross-functional discussion.

Even a planner with decades of experience may have time to investigate only the most apparent drivers before a commercial decision is made. Other interactions might remain untested because of the planning cycle’s limits, rather than any lack of skill. 

Three elements can help refinery planners examine LP results with consistency and explain the basis for a commercial recommendation:

• Analytical method. A repeatable sequence of case comparisons, constraint reviews and sensitivity checks helps planners investigate changes in economic value and relationships that warrant closer attention.
• Decision context. Prior cases, model notes, backcasting lessons and operating experience help planners test whether an apparent opportunity holds up under current refinery conditions.
• Explanation. Findings should connect the economic and operational drivers to their supporting sources, assumptions and unresolved questions so planners can challenge the interpretation before recommending action.

The immediate purpose is to understand the result more thoroughly. Documenting how planners conduct and explain that analysis also allows others to apply the method as planning teams change. Because markets and refinery conditions evolve, past decisions provide context for investigation rather than an automatic answer. 

Automation and AI can extend the reach of LP output analysis. Automation can perform repeatable case comparisons and structured checks. AI can help connect LP results with relevant refinery records and organize the economic and operational drivers into an explanation for planner review. Together, automation and AI can examine more relationships and available cases across a wider portion of the LP results, surface findings that warrant attention, and give the planner more time to challenge the most important conclusions.

That breadth, however, is useful only if the findings can be trusted. Calculations and explanations should be grounded in LP inputs and outputs and supported by identifiable refinery information. The planner should be able to see the evidence, assumptions and uncertainties behind each finding. When the available information does not support a conclusion, the system should flag the gap rather than supply a plausible-sounding answer. The planner then decides which assumptions to test, which additional LP runs to request, and whether the result supports a recommendation.

The application of automation and AI begins with a solid refinery planning foundation. Reliable process-unit data, sound yield and shift vectors, disciplined backcasting, and model stewardship all influence the quality of the results a planner must interpret. The focus explored here is the analysis of LP output and helping planners examine those results more comprehensively and understand the economic and operational drivers. As the underlying capabilities develop, they could also support other parts of the planning process, with each use validated against refinery data and operating experience. 

One example of this concept is Future of Refinery Optimization (FRO), a vendor-neutral intelligence layer developed by 1898 & Co. to support post-solve analysis of refinery LP results. It draws on existing LP outputs and refinery knowledge to produce findings planners can examine and utilize without replacing the refinery’s LP platform or planner judgment.

At its core, FRO uses a contextualized model that incorporates analytical methods drawn from decades of refinery-planning experience. Applied to a refinery’s LP results and operating context, that knowledge helps planners investigate the relationships behind an apparent economic opportunity.

FRO enables refinery planners to conduct a considerably faster and more comprehensive examination of LP output. By systematically comparing cases and tracing the drivers, it can bring relevant relationships to the planner’s attention that a time-constrained review might miss. Its output identifies the LP results and refinery records supporting each finding, shows the assumptions behind the interpretation, and flags uncertainties for the planner to investigate before making a recommendation.

The intelligence layer is designed to work with output from any LP platform. It uses those results alongside a refinery’s analytical methods, decision history and operating context to help planners investigate opportunities and explain their recommendations. 

A more comprehensive review of LP results would change how planners investigate opportunities and share their findings. Refinery leadership would need to define where the new analysis fits in the planning cycle, how its conclusions will be reviewed and who remains responsible for recommendations and operating decisions. Planners should understand the purpose before being asked to incorporate a new tool into their work.

The commercial manager and planning supervisors can work with planners to make the change useful in their daily work. Experienced planners have an opportunity to investigate relationships they rarely have time to pursue; newer planners gain a more structured way to examine results and the evidence behind them. Both groups can select initial planning questions, identify the records needed to investigate them, and challenge FRO’s findings. They can address concerns about reliability and added work directly, rather than assuming a demonstration would win acceptance.

At a refinery such as Summit Ridge, instruction could be treated as part of the implementation. The implementation team would teach planners how to compare cases, inspect supporting evidence, question an explanation and record corrections. Model owners and operations and commercial personnel would learn how findings would reach them for review. Planning, operations, IT and cybersecurity can align on authoritative LP files and refinery records, access controls, and a process for resolving disputed findings.

An initial application might examine an LP result favoring more of a discounted crude. FRO could help planners compare the base and proposed cases and trace how the crude price advantage interacts with unit utilization, hydrogen demand, product quality and blending constraints. Planners would then test the assumptions behind the apparent gain with operations before deciding whether to recommend a change. Working through a limited set of questions alongside established methods could help the team judge where FRO adds useful analysis and where further investigation is needed.

Summit Ridge could evaluate that work by comparing the time required to reach a reviewed recommendation, the range of cases and drivers examined, and the quality of explanations with its existing process. When recommendations move into execution, comparing actual economic results with the plan could add another perspective. Documented findings and planner feedback would help refine the workflow while preserving the reasoning behind decisions for future planning cycles. 

The story of Summit Ridge illustrates the next challenge for refinery planning: making the work between an LP result and a decision more thorough, traceable and repeatable. An intelligence layer such as FRO could help planners examine more of each result and substantiate the drivers of an apparent opportunity. Planners would still decide which findings need further testing and which recommendations are credible.

As AI capabilities mature, the role of an intelligence layer could extend further. A future version of FRO might work iteratively with an LP model, requesting additional runs and using the results to guide the next test within limits set by the refinery. It might also connect the refinery’s asset data hierarchy with current operating and maintenance information, helping planners recognize when an acute equipment constraint or chronic bottleneck limits an economic opportunity — and where addressing that bottleneck could create value. Documented analytical methods and explanations could help less-experienced planners carry out more of the investigation, with uncertain or consequential findings referred to seasoned colleagues. Each of these possibilities would require reliable data, validation, training and clear accountability.

Today, refiners can begin by revisiting a recent planning decision: Which drivers behind the LP result were investigated, which assumptions went untested before the deadline, and what happened when the recommendation reached operations or the crude purchasers? The answers show where a more systematic review could improve the next decision.

A contextualized model such as FRO could bring extensive refinery-planning knowledge to that review, informed by the site’s operating context and decision history. The goals are to examine more of the result, explain the drivers that matter, and give the refinery a stronger basis for capturing the value its LP identifies.