Twice a year, U.S. refinery runs fall by hundreds of thousands of barrels a day for weeks at a time, on purpose. The cause is the turnaround: a planned shutdown in which process units are taken apart, inspected and rebuilt. Turnarounds are the main reason the EIA's refinery utilization rate has a calendar, with troughs in late winter and fall and a plateau across the summer, and they are the first explanation to check before reading a drop in refinery runs as a supply problem. This guide explains what happens inside a turnaround, why the industry schedules them when it does, how large the seasonal swing is in the EIA data, and how a planned maintenance dip differs from the unplanned outages that actually move fuel prices. The utilization rate itself, its formula and its history, are covered in the companion guide to refinery utilization.
What a turnaround actually is
A refinery is a chain of process units: crude distillation first, then conversion units such as the fluid catalytic cracker and the coker, then treating units that remove sulfur and finish products to specification. These units run continuously, at high temperature and pressure, with corrosive materials moving through them. Metal thins, catalysts lose activity, heat exchangers foul, and safety valves and instruments drift out of calibration. Some of that wear can be managed while a unit runs. Much of it cannot: vessels have to be emptied, cooled, opened and entered before anyone can inspect a wall, replace a catalyst bed or retube an exchanger. Regulations and insurers also require periodic mechanical-integrity inspections that are only possible on a shut unit.
So each unit is taken down on a cycle, typically every few years, in an event the industry calls a turnaround, a shutdown or planned maintenance. A turnaround is closer to rebuilding a section of a factory than to servicing a car. Planning starts years ahead: engineers scope the work, order long-lead equipment, book contract crews that can number in the hundreds or thousands at a large site, and schedule the window against the plant's other units and the market calendar. The work itself runs around the clock. While a unit is down, the rest of the plant either keeps running at reduced rates, if the configuration allows it, or shuts with it when the downed unit is the crude unit that feeds everything else. When the work finishes, units restart in sequence, crude first, and a plant can take days to return to full rates even after the mechanical work is complete.
Two features of turnarounds matter for anyone reading market data. First, they are lumpy: a single large plant can be several percent of its district's capacity, so one site's maintenance window moves the regional numbers, which is the arithmetic worked through in the utilization guide. Second, they are known in advance inside the industry. Specialist services track planned maintenance plant by plant, traders price it in, and refiners prepare for it, which is why a normal turnaround season passes with so little drama compared with its size on paper.
Inside the fence: what the crews actually do
A visitor during a turnaround sees a plant wrapped in scaffolding, with crane booms over units that are normally an untouchable thicket of hot pipe. The work falls into a few families. Vessels and towers are opened, cleaned of coke and scale, and inspected from inside; inspectors measure wall thickness ultrasonically to find corrosion before it finds them, and any section below its minimum thickness is cut out and replaced. Heat exchangers are pulled apart bundle by bundle, cleaned or retubed. Catalyst beds, the chemicals that make conversion and treating units work, are dumped and replaced once their activity has faded, which is one of the hard clocks that sets a unit's cycle length: a hydrotreater runs until its catalyst dies, and no amount of mechanical soundness elsewhere extends that date by much. Rotating equipment, pumps, compressors and turbines, is opened and rebuilt. Safety systems, relief valves, alarms and shutdown instrumentation, are tested and recalibrated, because a turnaround is the only time the plant can prove its last lines of defense work.
Almost every turnaround also finds discovery work: corrosion worse than the last inspection predicted, a cracked fitting, a valve that fails its test. Discovery is the main reason turnaround schedules slip, and experienced planners budget for it in both time and money rather than treating the plan as fixed. That is why restart dates quoted at the start of a maintenance season should be read as targets, and why the data check in the final section of this guide watches for the restart in the input numbers instead of trusting the announced date.
Why spring and fall: the demand calendar and the fuel switch
Refiners would rather maintain equipment when the market least needs the product, and the U.S. fuel calendar offers two such windows. Winter demand peaks with heating oil and propane in the Northeast and Midwest. Summer demand peaks with gasoline during the driving season, when refiners run hardest and when a missing plant is felt fastest. Between those peaks sit late winter into spring, and early fall: demand is at its lowest, product inventories can be built ahead, and a plant can be down for six weeks without the market running short. That is where the turnarounds go.
Regulation sharpens the spring window. Summer-grade gasoline, with lower vapor pressure to limit evaporation and smog formation, is required at retail from June 1 to September 15 under the EPA's volatility program, and refiners and pipelines switch production and shipments weeks earlier to move winter-grade fuel out of the system and summer-grade in. The summer and winter gasoline guide explains the specification change and its cost. A refiner wants its spring maintenance finished before that switch, because the changeover itself temporarily tightens supply: tanks and pipelines cannot hold both grades at once, and output during the transition is constrained. Fall has the mirror image: after September 15 the specification relaxes back to cheaper winter-grade blends, demand falls with the end of driving season, and plants take their second maintenance window before winter heating demand arrives.
Table 1. The turnaround calendar against the fuel year
| Window | Typical timing | Why then |
|---|---|---|
| Spring turnaround season | February to April | Lowest gasoline demand of the year; work finishes before the summer-grade switch and driving season |
| Summer run peak | May to August | Driving season demand; plants run at their highest rates and defer all but emergency work |
| Fall turnaround season | September to November | Driving season over, winter-grade switch made, heating demand not yet at peak |
| Winter run period | December to January | Heating demand in northern districts; weather risk makes planned work unattractive |
Timing is the industry pattern described in the text; individual plants schedule their own windows, and Gulf Coast plants also plan around the June to November hurricane season.
The industry also tries, loosely, not to trip over itself. No authority assigns maintenance dates, but planned work is reported through the trade press and specialist data services, competitors can see roughly when rival plants in their district will be down, and a refiner choosing between two possible windows has every reason to pick the one when its neighbors are running, because product prices in the district will be firmer while a competitor is down and because the contract crews, cranes and specialized shops that turnarounds depend on are a shared regional resource that cannot serve everyone in the same month. The result is a rough staggering within each district rather than a single cliff, most of the time. It breaks down when many plants deferred work in the same earlier period, after a storm season or a stretch of unusually strong margins, and then crowd into the same window to catch up. Shutdowns and restarts also carry an environmental footprint that regulators track: taking units down and bringing them back involves flaring and venting that steady operation does not, and large events are reported to state air agencies, which is one more public record by which a genuine turnaround can be told apart from an unplanned trip.
The fingerprint in the EIA data
The seasonal pattern is not folklore; it is in the monthly averages. The table below takes U.S. refinery utilization by month for the four full years 2022 to 2025 and averages each month. The shape repeats every year: a February low, a climb through spring, a June to August plateau, a September and October fall, and a recovery into December.
Table 2. U.S. refinery utilization by month, mean of 2022 to 2025 (percent)
| Month | Jan | Feb | Mar | Apr | May | Jun | Jul | Aug | Sep | Oct | Nov | Dec |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Mean 2022-25 | 87.2 | 85.9 | 89.3 | 90.1 | 93.2 | 94.6 | 94.0 | 94.4 | 91.9 | 88.7 | 92.3 | 92.5 |
| 2022 | 89.0 | 88.9 | 91.5 | 89.9 | 93.3 | 95.4 | 94.1 | 94.1 | 92.5 | 90.3 | 94.1 | 88.0 |
| 2023 | 86.5 | 86.7 | 87.7 | 90.1 | 91.7 | 93.1 | 93.8 | 94.1 | 91.5 | 86.7 | 89.9 | 93.1 |
| 2024 | 86.1 | 82.7 | 88.7 | 89.5 | 93.5 | 94.2 | 92.4 | 94.0 | 90.4 | 89.7 | 91.6 | 93.0 |
| 2025 | 87.1 | 85.2 | 89.5 | 90.8 | 94.3 | 95.6 | 95.8 | 95.4 | 93.2 | 88.1 | 93.4 | 95.7 |
Source: EIA, Percent Utilization of Refinery Operable Capacity, monthly (series MOPUEUS2), checked October 23, 2026. The mean row is this page's arithmetic on the four published years.
The mean February reading is 85.9 percent and the mean for the peak month is 94.6, a swing of about 8.7 points. At the current U.S. operable capacity of about 18.0 million barrels a day, each point of utilization is roughly 180 thousand barrels a day of refinery input, so the average seasonal swing between the February trough and the summer plateau represents on the order of 1.5 million barrels a day of capacity cycling out for maintenance and back. The fall dip is smaller on average, with October at 88.7 percent, but it varies far more by year: October printed 90.3 in 2022 and 86.7 in 2023, because how much work the industry schedules, and how much of it slips, differs each fall.
A live example: September 2026 in the Midwest
The weekly data shows the same event at higher resolution. Between the weeks ending September 11 and September 25, 2026, U.S. gross refinery inputs fell from 17,449 to 16,670 thousand barrels a day, a drop of 779 thousand, and national utilization fell from 96.8 to 92.5 percent. The regional pages show where it happened: the Midwest (PADD 2) lost 619 thousand barrels a day of gross inputs, about four-fifths of the national fall, and its utilization dropped from 100.0 to 85.6 percent, while the Gulf Coast, with more than half of national capacity, barely moved. The weekly report does not name causes, and an unplanned event can produce the same shape, but a two-week, one-district step down of that size, arriving in the middle of the fall maintenance window, is the textbook fingerprint of turnaround work concentrated in one region. The guide to the weekly report shows where each of those series sits in the release, and the PADD guide explains the districts.
What a turnaround does to supply, stocks and prices
A refinery in turnaround makes less gasoline, diesel and jet fuel, so the first-order effect is a draw on product inventories in its region. Refiners plan for exactly that: they run hard ahead of a turnaround to build stocks, buy product from other regions or from imports to cover their supply contracts, and schedule their sales around the window. The system absorbs a normal turnaround the way it absorbs a holiday weekend, because everyone involved knew the date. Three situations break that calm, and they are the ones worth watching.
- Overlap. When several large plants in the same district are down together, regional stocks can fall faster than imports and pipeline shipments from other districts can replace them. District-level data, not the national average, shows this.
- Overrun. Turnaround work regularly finds more damage than the inspection plan assumed, and a unit that was due back in April returns in May, into the start of driving season, when the market has no slack left.
- Stacking with an unplanned event. A hurricane, freeze or fire that lands during maintenance season removes capacity the schedule had not spared, on top of the capacity already down by plan.
The price signal, when there is one, appears first in wholesale markets: regional gasoline and diesel prices strengthen against other regions, and the refining margin, the crack spread, widens for the plants still running. Retail prices follow only when the tightness lasts, because the pump price also carries crude cost, taxes and distribution, which the gasoline price breakdown itemizes. The honest summary is that turnarounds set the background level of supply for a season; they rarely set the headline price. Outages do that.
The crude side of the same event
A turnaround also removes a crude oil buyer. A district running 600 thousand barrels a day less crude, as the Midwest did across two weeks in September 2026, is a district whose refineries are not calling for deliveries, and the effect shows up in regional crude markets before it shows up at the pump. Barrels that normally move to Midwest plants back up toward the producing regions and the storage hubs, and regional crude grades can soften against the benchmarks while the maintenance lasts, the mirror image of the product side, where the same event tightens supply. Traders watch the two together: product cracks strengthening while a regional crude differential weakens is the classic market signature of refinery maintenance rather than of a crude supply problem, and it unwinds as the plants restart and resume pulling crude. For a reader of this site's price pages, the practical lesson is that a soft regional crude price during turnaround season is usually a demand story about refineries, not a supply story about oil fields.
Turnaround, outage, run cut, closure: four different events
Table 3. Four ways refinery output falls, and how to tell them apart
| Event | Planned? | Typical duration | What brings output back |
|---|---|---|---|
| Turnaround | Yes, years ahead | Two to eight weeks per unit | Completion of the work and a sequenced restart |
| Unplanned outage | No | Days to months, unknown at first | Repair of the failed equipment; fires and storm damage take longest |
| Economic run cut | Yes, weeks ahead | As long as margins stay weak | A recovery in crack spreads; no mechanical work involved |
| Permanent closure | Yes, announced ahead | Permanent | Nothing; the capacity leaves the EIA denominator, covered in the capacity guide |
Categories are this page's summary of industry usage. The EIA weekly data records the resulting rates, not the category, so classification comes from company and trade reporting.
The distinction matters because the four events have opposite implications. A turnaround dip is supply coming back on a known date, with the plant returning in better condition than it left. A run cut says refiners see weak margins, which is demand information, not supply damage. A closure, such as the conversions and shutdowns listed in the 2026 refinery capacity guide, permanently removes the barrels and, by shrinking the denominator, can even raise the reported utilization rate of the plants that remain. An outage is the only one of the four that is both unscheduled and open-ended, which is why it is the only one that reliably moves prices on the day it happens.
How to read a turnaround season: a five-step check
- 1. Compare with the season, not the summer.Hold the weekly utilization reading against the same month's average in Table 2. A 90 percent reading in October is ordinary; the same reading in July is not.
- 2. Find the district. Use the PADD-level inputs before the national figure. A national dip carried by one district is a plant story; a dip in all five is a market story.
- 3. Watch product stocks, not just runs. Gasoline and distillate inventories in the weekly report show whether the maintenance is being covered from storage or is genuinely tightening supply.
- 4. Check the restart. Turnarounds end in steps. Rising district inputs over two or three consecutive weeks usually mark the restart sequence, and a flat line past the expected window is the first sign of an overrun.
- 5. Look for the second cause. If prices are moving during a routine maintenance dip, something else, an outage, a pipeline problem, a crude move, is doing the work. The turnaround alone is rarely the story.
The one-sentence version
Refinery runs fall every spring and fall because the industry takes its plants apart then, on purpose, in the weeks when fuel demand is lowest; read a dip against the calendar and the district before reading it as a shortage.
Turnaround questions
What is a refinery turnaround?
A turnaround is a planned shutdown of part or all of a refinery for inspection, repair and equipment replacement that cannot be done while units are running. Crude distillation units, crackers, hydrotreaters and other process units are opened, inspected, cleaned and rebuilt on a cycle of several years per unit. A major turnaround takes weeks, employs large temporary crews, and is scheduled years in advance.
When is refinery turnaround season?
Twice a year: late winter into spring, and fall. Refiners place maintenance in the shoulder seasons between the winter heating peak and the summer driving peak, and finish spring work before the switch to summer-grade gasoline. In the EIA monthly data for 2022 to 2025, average U.S. utilization is lowest in February (85.9 percent) and dips again in October (88.7 percent), against a summer plateau around 94 to 95 percent.
How long does a refinery turnaround last?
It depends on the scope. Work on a single unit can take two to four weeks; a full-site turnaround involving several units commonly runs four to eight weeks. Restarts are phased, unit by unit, and a plant often runs at reduced rates before and after the main window as units are taken down and brought back in sequence.
Do turnarounds raise gasoline prices?
Planned turnarounds, on their own, usually do not. They are scheduled in the lowest-demand weeks of the year, refiners build product inventories ahead of them, and the market knows the calendar in advance. Prices react more to unplanned outages, to several large plants being down at once, or to a turnaround that overruns into the driving season, because those remove supply the market had not already replaced.
What is the difference between a turnaround and an outage?
A turnaround is planned, permitted and announced to the market through industry reporting, with crews and parts booked long ahead. An outage is unplanned: a fire, a power failure, a freeze or a hurricane forces units down without warning. The two look identical in the weekly utilization data, which records rates and not reasons, so the distinction has to come from reporting about the specific plant.
How can I see a turnaround in the data?
Watch the EIA weekly refinery utilization by region (PADD). A turnaround at a large plant shows up as a district-level step down in gross inputs lasting several weeks, followed by a step back up, at the same point in the calendar each year. In September 2026, for example, Midwest gross inputs fell 619 thousand barrels a day in two weeks, about four-fifths of the national fall, the classic shape of fall maintenance concentrated in one district.
Data and sources: U.S. Energy Information Administration, Percent Utilization of Refinery Operable Capacity, monthly, checked October 23, 2026 (monthly utilization 2022 to 2025 behind Table 2), and U.S. Total Weekly Inputs & Utilization with the regional PADD pages, checked October 23, 2026 (weeks ending September 11 and September 25, 2026, and the capacity figure used for the per-point conversion). U.S. Environmental Protection Agency, gasoline standards and fuel volatility program (summer gasoline season of June 1 to September 15 at retail). Turnaround practice, durations and scheduling are described from standard industry usage; the EIA series record rates and volumes, not the reasons for them, and this page does not attribute any specific week's movement to a named plant.