| Series | First | Latest | Min | Max |
|---|---|---|---|---|
| Refinery utilization | 84.00 | 96.50 | 56.00 | 100.50 |
| Five-year average | 86.00 | 93.07 | 81.33 | 96.97 |
seasonalAverage(t) = mean of same EIA week (±1 week) in the five complete years before tWeekly U.S. percent utilization of refinery operable capacity (EIA series WPULEUS3), plotted against the five-year seasonal average only for the same calendar week. The reference uses the same EIA week number (plus or minus one week) in the five full calendar years before each observation. The current year never enters its own baseline. Each reference year contributes the mean of its tolerated observations. The average line starts once a five-year window exists.
As of 31 July 2026, the latest readings are Refinery utilization at 96.50 Percent and Five-year average at 93.07 Percent. The Refinery utilization line is down 0.4 pp over the past year and above its long-run median of 90.30 Percent.
Utilization is the refinery fleet's workload as a percentage of operable capacity. The industry's gauge of how much product supply the system can push out. A reading at the top of the band means refineries are at or near maximum run rates, leaving little spare product capacity: strong crack spreads but fragile supply. A reading below the band means capacity is idle or offline, which tightens product supply even when crude is abundant.
Energy series move with the calendar year. A raw level cannot say if the market is tight or slack. This page ranks each week against the same week in the five full years before it. A July reading is judged against prior Julys, not against the full year. The baseline uses only earlier years. The current year never enters its own baseline. The rule allows one week of tolerance when holidays shift the report week.
For each date, the reference is the same EIA week number (plus or minus one week) in the five full calendar years before it. Each reference year adds one value (the mean of its tolerated observations). The five-year average uses those yearly values. This page plots the average only. The average line starts once a five-year window exists and rolls forward.
The five-year average moves as history rolls forward. Source revisions can move history. Refinery utilization is one input to the market balance. Prices also move on the weekly surprise versus expectations, not on the level alone. Treat the chart as context for supply and demand conditions, not investment advice.
Refining capacity is the binding constraint on product supply, and utilization is its dial: since the 2021-22 capacity retirements, the fleet has run near the top of its historical band, and every percentage point above the seasonal norm is product supply with no spare capacity behind it. Crack-spread traders watch this page for exactly that headroom.
This page uses a same-week rule over five full prior years. It allows one week of tolerance. The current year is excluded. It plots the seasonal average only. Compare the reading to the average line, not to a min or max range.
As of 31 July 2026, the latest readings are Refinery utilization at 96.50 Percent and Five-year average at 93.07 Percent. The Refinery utilization line is down 0.4 pp over the past year and above its long-run median of 90.30 Percent.
This benchmark is built on weekly data. The page is refreshed when the source publishes new observations; the freshness block below the chart shows the exact data-through date and when PIER20 last fetched the file.
The chart is built from Refinery utilization and Five-year average, sourced from the U.S. Energy Information Administration (EIA).
The percentile cell ranks the latest reading against the full sample. The 100th percentile is the highest on record. The 0th is the lowest. This page charts the five-year seasonal average only. Read extremes from the full history and the percentile cell.
Utilization normalizes by capacity: the same crude input volume is 80% utilization on a large fleet and 95% on a small one. The inputs page answers 'how much crude is being processed'; this page answers 'how much of the fleet is working'. The spare-capacity question that matters for product prices.
A baseline needs five full prior years of observations. The current year never enters its own baseline. The first five years of a series have no complete reference window. Earlier readings plot without a seasonal reference.
One tested idea, through the evidence stack. No migration of any kind.