WHITE PAPER · AMA · ASSET MODEL ACCELERATOR
Why the gap between top-quartile and average refineries is really a data gap, and how the drawings you already own close it.
FLATIRONS DIGITAL INNOVATIONS · JULY 2026 · VERSION 1.0
EXECUTIVE SUMMARY
The performance gap between the world's best refineries and the average is not primarily a hardware gap. It is an information gap. Top-quartile operators have a faster, cleaner, more trustworthy path from "we own an asset" to "here is everything anyone needs to know about it."
AMA, the Asset Model Accelerator, turns that path into a repeatable capability: it reads the P&ID library a site already owns and converts it into a queryable model of the plant, with every value traced back to the sheet it came from.
SOLOMON'S WORLD'S BEST REFINERIES VERSUS THE INDUSTRY AVERAGE
20%
more energy efficient
40%
of average maintenance cost
+6 pts
higher utilization
60%
more profitable
Refineries are not graded on a single efficiency percentage. The de facto ranking system for the global refining industry is the biennial Fuels Study from Solomon Associates. The 2020 study covered roughly 280 refineries, approximately 85% of global refining capacity, which is why its benchmarks sit on executive scorecards.
| METRIC | WHAT IT MEASURES | HOW TO READ IT |
|---|---|---|
| Energy Intensity Index | Actual energy consumption divided by standard consumption for the refinery's configuration, normalized to complexity-weighted barrels. | 100 equals industry average; below 100 is better. Used by ENERGY STAR and referenced in Washington state regulation. |
| Opex per UEDC | Cash operating expense per Utilized Equivalent Distillation Capacity barrel. | The primary cost benchmark. Solomon reports an R² of 0.93 between opex and UEDC, enabling comparison across very different configurations. |
Solomon does not publicly disclose quartile cutoffs for individual metrics. This paper therefore leans only on differentials Solomon has published about its World's Best peer group, and on arithmetic any reader can reproduce with their own site's numbers.
Solomon's World's Best Refineries peer group is the strongest public proof point. To qualify, a refinery had to place in the first or second quartile on all four screening metrics, Operational Availability, Energy Intensity, Maintenance Cost Efficiency, and Return on Investment, across three consecutive Fuels Studies: 2016, 2018, and 2020. Only 12 refineries out of roughly 280 cleared that bar. Those 12 run 20% more energy efficient, at 40% of average maintenance cost, six utilization points higher, and 60% more profitable than the industry mean.
The gap also compounds. Long-run Solomon data from a period of stable oil prices shows the top quartile improved opex per UEDC by 22% while the bottom quartile improved by only 7%, widening the absolute gap by 13 index points over the study window. The underlying data covers 1994 to 1998, so it is best read as a long-run pattern rather than a current-cycle claim. The direction, however, is consistent with every study since.
THE STRATEGIC READING
The industry's benchmark authority says the winners are winning by a lot, and the advantage compounds. Doing nothing is not staying flat. It is falling behind.
The rule of thumb that a single percentage point can be worth tens of millions of dollars a year is defensible, provided the arithmetic is shown. The assumptions below are deliberately ordinary: a mid-size complex refinery in a mid-cycle margin environment.
| BASE CASE | ASSUMPTION |
|---|---|
| Nameplate capacity | 250,000 barrels per day |
| Annual throughput at 100% | 250,000 × 365 = 91.25 million barrels |
| Gross refining margin | $15 per barrel, a mid-cycle assumption |
| Baseline gross margin | 91.25M × $15 ≈ $1.37B per year |
| LEVER | THE MATH | ANNUAL IMPACT |
|---|---|---|
| Availability | 250,000 × 365 × 1% × $15 | $13.7M |
| Realized margin | A 1% margin improvement: 91.25M barrels × $0.15 | $13.7M |
| Energy intensity | 1% of a $150M to $250M annual energy spend | $1.5M to $2.5M |
| Maintenance | 1% of an $80M to $120M annual maintenance spend | $0.8M to $1.2M |
The availability lever scales linearly with margin, so the conclusion survives the cycle: at a $10 margin the point is worth roughly $9.1M a year, at $15 it is $13.7M, and at $20 it is $18.3M. The order of magnitude does not move.
REAL-WORLD ANCHOR
Solomon's Dual Refinery Optimization case study covered two refineries with more than 325,000 barrels per stream day of combined capacity. It identified $18M per year of value, of which $6M required no capital investment and came from closing just 10% of the energy gap on the larger site. Roughly a third of the value needed no new equipment, only better decisions about existing assets.
Note what kind of number this is. It is not a cost-trimming story. The largest lever is the availability line: days at full rate that would otherwise be lost to searching, verifying, and waiting. The return arrives as production.
Solomon publishes the outcomes but not the recipe. Read the trade press, consulting research, and operator conference talks together and one theme keeps repeating, and it is not a hardware theme: top-quartile refineries make decisions faster because trusted data reaches the decision-maker faster.
Predictive maintenance instead of calendar-based maintenance. Predictive programs require a queryable asset model: every pump, valve, and instrument mapped to its drawing, datasheet, inspection history, and criticality. Without that context, organizations fall back to fixed schedules and absorb more unplanned downtime.
A single source of truth for assets. Top operators consolidate P&IDs, equipment lists, and vendor documents into one searchable index. An engineer asking what is known about pump 40-P-1043 receives a complete answer rather than a folder of PDFs to inspect.
Digitized engineering-document workflows. Operators are moving from static drawing sets to tag-aware digital viewers because turnaround planning and Management of Change reviews cost less when the P&ID behaves like a database rather than a picture.
A short mean-time-to-information. Define it as the elapsed time from an engineer's question to a trusted answer. Engineer hours spent retrieving and reconciling documents are hours not spent on reliability and optimization. It is also the one lever on this list a site can start measuring this quarter.
THE CORRELATION, HONESTLY STATED
There is no peer-reviewed regression proving that a given increase in data accessibility produces a specific benchmark improvement. The defensible claim is that top-quartile operator profiles share the same shape: they know what they own, they can find what they need, and their engineers spend their day making decisions instead of reading PDFs.
AMA turns a site's existing P&ID library into the asset-model foundation that top-quartile operators often spent years of internal effort and consulting support to build. It does this by inverting the P&ID from an artifact into a database.
Every symbol becomes a queryable asset. Pumps, valves, instruments, and safety devices are identified with computer vision tuned on engineering symbology, then linked to their class, parent equipment, and location on the drawing.
Every tag becomes navigable. Search for 40-P-1043 and receive the pump, its location on the sheet, its parent unit, its child instruments, and every document in the corpus that references it.
Every document becomes searchable in context. Datasheets, isometric drawings, inspection reports, and vendor manuals are indexed against the tags they describe, so the paper trail is attached to the asset instead of scattered around it.
Every P&ID becomes a live model. Additions, redlines, and Management of Change updates flow back into the asset registry instead of drifting out of sync with the static drawing set.
| FINDING AN ANSWER TODAY | WITH AN ASSET MODEL |
|---|---|
| Open the drawing. Hunt for the tag. Cross-reference the equipment list. Search the network share for the datasheet. Ask the one engineer who remembers. | Type the tag. Read the answer. Navigate the relationships. Open the documents, already attached. |
| TENS OF MINUTES ON A GOOD DAY | SECONDS |
Built to be believed. Asset data only matters if engineers trust it. Every value AMA extracts carries the sheet, the region, and the confidence score it came from; low-confidence objects are queued for human review instead of entering the model as fact. On a 56-sheet reference set, AMA reads 13,962 objects into the graph at 98.9% symbol-detection precision, and during a pilot the accuracy is measured against your own drawings, not a benchmark. It runs on your hardware: on premise, in a private cloud, or fully air-gapped, so the drawing set never leaves your network.
Three numbers bracket an honest business case. The floor is what a skeptical CFO should accept; the reasonable case is what belongs in an operating plan; the external reference shows the industry's own benchmark authority landing in the same range.
CONSERVATIVE FLOOR
$6.8M / yr
A 0.5% availability lift on a 250,000 bpd site at a $15 margin.
REASONABLE CASE
$20M+ / yr
1% availability plus a 0.5% margin uplift plus reduced document-search time, moving together.
EXTERNAL REFERENCE
$18M / yr
Identified by Solomon across two refineries, $6M of it requiring no capital.
The recommended headline for an internal business case: a $20M+ annual value opportunity from turning refinery information into operational velocity. The return shows up where it counts, on the availability line, as production at full rate, faster turnarounds, and fewer jobs extended by missing context.
Top refineries win because their people can find things. AMA is how everyone else's people find things.
Everything else, the drawing viewer, the tag search, the equipment hierarchy, the document index, the assistant, is a mechanism for that sentence. The Solomon numbers are the receipts for why the sentence is worth eight figures a year.
PROVING IT ON YOUR OWN SHEETS
Bring one drawing set. AMA runs on your hardware, and in two weeks you review the asset model that comes out: every object linked, every value carrying its evidence, precision measured against your own sheets. You keep the database either way.
Request a guided test drive at fdiama.com or call 888-310-3440.
This paper uses public, citable evidence and separates documented benchmark results from directional industry themes. Where a widely repeated number did not survive source review, it was excluded, and the exclusions are listed below because a claim you can audit is worth more than a claim that impresses.
Solomon Associates, World's Best Refineries. Source for the 20% energy, 40% maintenance-cost, six-point utilization, and 60% profitability differentials; peer-group qualification method and study population.
BOTEC, Solomon Energy Intensity Index glossary. Definition of EII, ENERGY STAR adoption, and the Washington state regulatory reference.
Oil & Gas Journal, Best practices determine refinery pacesetter performance. Source for the opex/UEDC R² of 0.93 and the long-run top-versus-bottom quartile divergence.
Solomon Associates, Dual Refinery Optimization case study. Source for the $18M per year identified value and the $6M no-capital component.
NUMBERS WE CHOSE NOT TO USE
The following circulate widely but did not survive citation review, so they do not appear in this paper: the claim that engineers spend 30 to 50% of their time searching for information; the $50B-per-year industrial downtime figure; the claim that 1% of downtime equals $5M per year for oil and gas operators; a "30% more reliable" differential whose underlying metric is not publicly defined; and named-operator digital transformation ROI figures whose public dollar values lack primary sources.
If a number in this paper matters to your decision, ask us for the source. We will show it.