Product Supplied, Explained

1. What it measures, and the latest figures

Product supplied is the number the oil market uses for US demand. It is not a count of what drivers, airlines, and furnaces burned. It is the EIA's measure of how much product disappeared from primary supply: the refineries, natural gas plants, blending plants, pipelines, and bulk terminals where the EIA's surveys can see the whole flow. The EIA's glossary definition is the honest one: product supplied approximately represents consumption, because most of what leaves the primary chain is sold onward and consumed not long after. Every Wednesday, the figure arrives inside the Weekly Petroleum Status Report. The latest week in the data archived on this site, the week ending October 23, 2026:

Three readings fall out of that table immediately. Gasoline is roughly two fifths of all US petroleum disappearance, which is why a single product's driving season steers the whole demand story. Distillate, the diesel and heating oil pool, is roughly a fifth, and it answers to freight and winter more than to motorists. And the total, 20,209 thousand barrels a day in the latest week, sat 6.6 percent below its 52 week high of 21,648 and above its 52 week low of 19,226: the band the US demand debate actually lives in is a few percent wide, which is why small weekly moves get argued over so hard.

2. The formula: an accounting identity, not a survey of buyers

Product supplied is computed, per product, from the supply side. In the EIA's formulation: field production, plus transfers to crude oil supply, plus biofuels plant net production, plus refinery and blender net production, plus imports, plus net receipts on a district basis, plus adjustments, minus stock change, minus refinery and blender net inputs, minus exports. Compressed into one line:

Product supplied = production + imports + adjustments - stock change - refinery inputs - exports

Read it as a warehouse ledger. Everything that entered primary supply is on the left. Subtract what is still sitting in primary storage (the stock change), what went back into a refinery as an input, and what left the country (exports). The remainder had nowhere else to go: it disappeared into the domestic distribution chain. Two terms deserve attention. The adjustments line is a balancing item that absorbs survey and statistical error, and for finished gasoline it reconciles the product with its two surveyed components, gasoline blending components and fuel ethanol, which are collected separately and never quite foot. And the stock change term is why the series is noisy: it is computed from the same primary stock surveys as the inventory headlines, so every quirk of tank measurement lands inside the demand number too.

A worked ledger, and how a filing lag fakes a demand swing

The identity is easiest to trust after running it once. Take a hypothetical gasoline week with stated inputs, in thousand barrels a day: refinery and blender net production 9,600, imports 600, exports 900, and primary stocks down 2.1 million barrels over the seven days, which is a stock change of minus 300 a day. Product supplied is 9,600 + 600 - 900 - (-300) = 9,600. Every term is a flow the surveys can see; the result is simply what is left over.

The table shows the failure mode the EIA itself describes in the explanatory notes to the weekly report (checked October 2026): the companies that report inventories are often not the companies that report imports, and one firm's books can lag another's. In the example, a 1.4 million barrel cargo (200 thousand barrels a day over a week) lands in tanks in week 1, so the stockholder's count already includes it, but the importer files it in week 2. Week 1 product supplied prints 200 low, week 2 prints 200 high, and a reader comparing the two weeks sees a 400 thousand barrel a day demand surge that no driver produced. The two week average recovers the true 9,600 exactly. Nothing was misreported; the ledger was simply closed on two different clocks.

3. Why a proxy at all: nobody can meter a continent weekly

The obvious alternative, measuring fuel where it is burned, fails on arithmetic. The United States has well over 100,000 retail gasoline stations, plus fleet depots, farms, rail yards, airports, marinas, and home heating tanks. Surveying final sales weekly would be slow, partial, and late. Surveying the primary chain is the opposite: a limited population of refiners, blenders, importers, pipeline companies, and bulk terminal operators handles effectively the entire national flow before distribution, and the EIA surveys that population. What leaves their tanks for the domestic market is the closest measurable thing to consumption, and it is available five days after the week ends.

The cost of the shortcut is a known blind spot: secondary storage. Fuel sitting in a retailer's underground tanks, a distributor's depot, or an airline's fuel farm has already counted as supplied, whether or not anyone has bought it yet. In the EIA's own illustration of the effect, a refinery that moves an unusually large volume to distributors in one week produces a product supplied spike that week and a sag the next, while final sales never changed pace. The transfer is real; the demand signal is an artifact. This is not a flaw anyone is hiding. It is the stated reason the serious readers of this series, the EIA included, work with averages, the subject of section 5.

How large is that blind spot? The EIA's explanatory notes concede that the most recent study of secondary stocks is old: a National Petroleum Council survey from 1989, which found secondary distillate stocks equal to about 6.9 percent of distillate stocks and 6.7 percent of distillate storage capacity, and noted that secondary capacity was shrinking under environmental rules. The practical reading is that the unmeasured layer is small relative to primary stocks, so it cannot hide a sustained demand trend, but it is large enough to absorb a week or two of distributor restocking ahead of a holiday or a price increase. When distributors expect prices to rise, they pull product out of primary storage early, product supplied jumps, and the next weeks look soft while the secondary tanks are drawn down.

4. Reading the products separately

The total is the headline, but the information is in the split, because the products answer to different economies. Gasoline supplied is road demand: it peaks with summer driving, dips in January, and jumps around holiday weekends, and a single week containing a holiday is not comparable with the week before it. Distillate supplied is freight and heat: trucking, rail, farming, construction, and, in the Northeast, heating oil, so it carries a winter weight and tracks the goods economy more than the consumer. Jet fuel supplied is air travel, with its own holiday and summer peaks. The remainder of the total, roughly 30 percent, is the industrial barrel: residual fuel, propane, and petrochemical feedstocks, where a single plant turnaround can move the line. Ethane is the clearest case: almost all of it goes to ethylene crackers, and the natural gas liquids guide sets its product supplied beside gas plant output and exports for 2015 to 2025.

Two comparisons discipline any reading. The first is against the same week a year earlier, which strips out the season: gasoline supplied in a July week should be judged against last July, not against June. The second is the four week average against its own year-ago value, the form in which the EIA itself headlines the series. A demand story that only exists in one unadjusted week, against last week, is usually a transfer story wearing a demand costume. Our analysis page archives the weekly series so both comparisons can be made against the record rather than against memory.

The annual record is the frame every weekly print should be set against. The EIA's Short-Term Energy Outlook carries US consumption, which is product supplied on the monthly basis, by product in its Table 4a. The October 2026 edition (checked October 2026) gives 2025 as history and 2026 and 2027 as forecasts:

Set the weekly highlights for the week ending October 16, 2026 beside that table and the discipline of averaging becomes concrete. In the report for the week ending October 16, 2026, the EIA put total product supplied over the prior four weeks at 21.1 million barrels a day, 3.2 percent above a year earlier, with gasoline at 8.7 million (up 1.5 percent), distillate at 3.9 million (down 2.4 percent), and jet fuel up 1.4 percent. Yet the same agency's monthly outlook, published the same month, expects 2026 as a whole to come in below 2025, at 20.60 against 20.74, with gasoline down about 2 percent in a year when retail gasoline averaged $4.35 a gallon in September. Both can be true. A four week window in autumn can run hot against a soft year-ago base while the annual average carries the weaker spring and summer the outlook already counts. The weekly number answers what moved out of primary supply lately; the annual table answers how big the market is. Readers who quote one as if it were the other end up arguing past each other.

5. The four week convention, and the weeks that break the single figure

The EIA presents product supplied on a four week average basis in its report highlights, and the convention is worth adopting wholesale. Averaging four weeks cancels most transfer timing: a cargo pushed into week one is missing from week two, and the mean of the two is close to the true pace. It also softens the calendar effects that plague individual weeks, like a holiday landing on a different weekday than last year. What the average cannot fix is a genuine break: a hurricane that closes Gulf refineries and ports cuts production and transfers at once, and product supplied falls for supply reasons while wearing a demand label. The honest reading of a storm week is to say so, and to wait for the average to wash the event out. The release schedule matters here too: a holiday week has fewer working days and can move the publication date, and comparing a holiday week against an ordinary one is a quiet source of phantom demand moves.

6. How close a month of weekly figures lands: 132 months against the monthly survey

Averaging fixes the week-to-week noise, but it leaves a second question that matters more for anyone building a demand argument: once the fuller monthly survey arrives, how far was the weekly picture from it? The EIA publishes both series, so the test can be run directly. We took every weekly product supplied figure from the historical file behind the Weekly Petroleum Status Report's Table 1, spread each one across the seven days it covers (each weekly figure is a daily rate for the week ending Friday), averaged those days by calendar month, and set the result against the same month's product supplied in the Petroleum Supply Monthly. The full span runs from January 2015 to July 2026, the latest monthly figure published as of October 29, 2026; the summary below uses the 132 complete-year months from 2015 to 2025.

Read the table from right to left. Gasoline, the product with the most attention, is also the one the weekly survey tracks best: in seven months out of ten the weekly-based figure landed within 2 percent of the monthly survey, and the typical miss was 1.4 percent, about 134 thousand barrels a day. Total product supplied is close behind, within 2 percent in about two months of every three, with a typical miss of 1.3 percent, which on a total above 20 million barrels a day is still about 261 thousand barrels a day, larger than many of the weekly moves that get headlined. Jet fuel, a smaller pool, misses by more in percentage terms. Distillate is the outlier. Only 48 of the 132 months, a little over a third, landed within 2 percent; the typical miss was 2.9 percent; and in 37 months the gap was larger than 5 percent.

Distillate also carries the one persistent lean in the table. The weekly-based figure came in below the monthly survey in 86 of the 132 months, and the average gap across the period was minus 1.7 percent, about 71 thousand barrels a day. Year by year, the weekly figure averaged below the monthly one in nine of the eleven years; 2017 and 2021 were the exceptions. The EIA's published notes do not, as far as we found, give a cause, so we do not offer one. The practical reading is enough: a weekly distillate demand figure has historically been more likely to understate the eventual monthly number than to overstate it, and a distillate story built on a gap of a percent or two between weekly figures sits inside the normal distance between the weekly and monthly measures.

The EIA has run this check on itself before, with the same method. In This Week in Petroleum for November 23, 2005, it compared monthly data with what it called monthly-from-weekly prorated averages and found the monthly total product supplied figure higher 63 percent of the time since January 2000, by an average of 121 thousand barrels a day, with gasoline revised upward 62 percent of the time; it concluded that the trend had been for weekly data to underestimate demand. A decade later the total no longer leaned that way. From 2015 to 2025 the monthly total was higher than the weekly-based figure in only 57 of 132 months, 43 percent, while for gasoline it was higher in 77 months and for distillate in 86. The old rule that the monthly survey will show more demand than the weekly data held for gasoline and distillate in this period and failed for the total until 2023.

The total line in the chart has a shape worth noticing. From 2015 through 2022 the weekly-based total ran above the monthly survey on average in every year, by between 0.76 and 1.68 percent. From 2023 it ran below: minus 0.48 percent in 2023, minus 1.17 in 2024 and minus 1.91 in 2025. Gasoline averaged below the monthly survey in every year from 2022 to 2025, after mixed signs before that. We can describe the turn but not explain it from the published notes, so treat it as a feature of the record rather than a known cause. What it does establish is that the sign of the weekly-to-monthly gap is not fixed, and a reader cannot simply add a standard correction to the weekly figure to predict the monthly one.

The biggest single misses make the same point at full scale. In January 2022 the weekly-based total averaged 21,666 thousand barrels a day against a monthly survey figure of 19,613, a gap of 2,053, or 10.5 percent. In February 2021 the gap was 1,768, or 10.0 percent. For distillate, July 2017 ran 567 thousand barrels a day, 15.6 percent, above the monthly figure, and February 2016 ran 521, 13.0 percent, below it. Six months out of 132 missed by more than 5 percent on the total. In those months, a demand reading taken from the weekly data alone was off by more than 5 percent.

The most recent months are quieter. From January to July 2026 the weekly-based total averaged 0.02 percent above the monthly survey, with an average miss of 1.23 percent, and five of the seven months landed within 2 percent. July 2026, the latest month published, is a fair example: the weekly data averaged 20,438 thousand barrels a day and the monthly survey 20,612, so the weekly-based figure ran 174 thousand barrels a day, or 0.8 percent, low. Seven months is too short a run to call a change, and the comparison will move as later months arrive.

The table also sizes the weekly headlines quoted in section 4. In the report for the week ending October 16, 2026, the EIA put four-week distillate product supplied 2.4 percent below a year earlier, gasoline 1.5 percent above, and the total 3.2 percent above. Because both sides of each comparison are weekly figures, a steady lean in the weekly series would largely cancel out of a year-on-year change. What does not cancel is the month-to-month scatter around the monthly survey, and the size of that scatter differs by product. For the total, a 3.2 percent rise is more than twice the typical 1.3 percent miss. For gasoline, a 1.5 percent rise is about the size of the typical 1.4 percent miss. For distillate, a 2.4 percent fall is smaller than the typical 2.9 percent miss. On the weekly data alone, the distillate decline is a lead to check rather than a finding, and the monthly survey for the same months is the number that settles it.

A note on method. Weeks do not line up with months: a week ending Friday, October 2 is mostly September. Spreading each weekly figure across the seven days it covers and averaging by calendar day, as above, matches the calendar month the monthly survey counts, instead of assigning a whole week to the month its last day falls in. Only months fully covered by weekly data are included. The monthly figures are the values in the EIA series as downloaded on October 29, 2026, so where the EIA has revised a month since first publishing it, the comparison uses the revised value.

The practical upshot is to compare weekly figures with weekly figures: a year-on-year change built from the same weekly series carries the same measurement on both sides, and it is the form the EIA's own highlights use, four-week averages against their year-ago values. Do not set a weekly figure against a monthly or annual one, such as the Short-Term Energy Outlook consumption table in section 4, and read the difference as a trend; a gap of 1 to 3 percent is within the typical distance between the two measures shown in the table. And when a long history matters, build it from the monthly series once it exists, as the production vs consumption guide does. The crude side of the weekly balance has its own published measure of uncertainty, the crude oil adjustment, which the API vs EIA guide measures week by week.

7. What product supplied is not

• Not retail sales. It counts fuel when it leaves primary supply, which is before the distributor, the retailer, and the driver. Short-run gaps between the two are normal and usually reverse.
• Not exports-inclusive demand. Exports are subtracted by construction. World demand for US products lives in the export lines of the same report, and in the API and EIA tallies alike it is a separate quantity from domestic disappearance.
• Not final. The weekly figure is an estimate built on the weekly surveys. The monthly Petroleum Supply Monthly restates the same quantities on a fuller basis, and annual benchmarks move them again. Treat the fourth decimal of any weekly demand claim as fiction.
• Not a price signal by itself. Demand a few percent above or below last year matters to prices through inventories: strong supplied figures with rising stocks say something different from the same figures with falling stocks. The demand line and the stock lines of the report are one argument, not two.

8. Mistakes readers make with this number

• Headlining the single week. A 500 thousand barrel a day weekly swing in gasoline supplied is usually transfers, weather, or calendar, not a demand break. Check the four week average before writing the story.
• Comparing across seasons. January gasoline against July gasoline measures the calendar. Same week last year is the comparison that isolates demand.
• Calling it consumption without the caveat. The EIA's own wording is "approximately represents consumption." Dropping the "approximately" overstates what any weekly proxy can know.
• Adding crude oil on top of it. In the EIA's balance, product supplied for crude oil itself is assumed to be zero, according to the EIA's monthly explanatory notes. The mistake is adding crude production or refinery crude inputs to total product supplied a second time, which double counts the barrel on its way to becoming products.
• Reading a supply outage as demand weakness. When refineries or ports close, less product leaves primary supply because less exists to leave. The formula cannot tell a closed refinery from a quiet consumer; the rest of the report can.

9. Frequently asked questions

Data and sources: U.S. Energy Information Administration (EIA) - glossary definition of product supplied and table definitions for the Weekly Petroleum Status Report, read October 21, 2026 (product supplied approximately represents consumption because it measures the disappearance of products from primary sources; computed as field production, plus transfers to crude oil supply, plus biofuels plant net production, plus refinery and blender net production, plus imports, plus net receipts and adjustments, minus stock change, minus refinery and blender net inputs, minus exports); EIA frequently asked questions, "How do we calculate product supplied?", read October 21, 2026 (finished motor gasoline identity: net production plus imports plus supply adjustment minus primary storage inventory change minus exports; the supply adjustment reconciles finished gasoline with its blending components and fuel ethanol); EIA Today in Energy, "Understanding petroleum product supplied", on the proxy logic and the primary versus secondary storage distinction. Weekly figures and four week averages on this page are computed at build time from the PetroEyes archive of EIA weekly data (latest week ending October 23, 2026); the table values will move as the archive updates, and the method described does not change. EIA, Weekly Petroleum Status Report explanatory notes (Appendix B), checked October 22, 2026 (product supplied as implied demand from disappearance out of the primary supply chain; the 1989 National Petroleum Council finding that secondary distillate stocks equaled about 6.9 percent of distillate stocks and 6.7 percent of distillate storage capacity; the note that inventory and import respondents differ and their timing differences cause weekly variation in product supplied). EIA, Short-Term Energy Outlook, October 2026, Table 4a, checked October 22, 2026 (US consumption 20.74 million barrels a day in 2025, forecast 20.60 in 2026 and 20.81 in 2027, with the product split shown above; retail gasoline averaged $4.35 a gallon in September 2026). EIA, Weekly Petroleum Status Report highlights for the week ending October 16, 2026, checked October 22, 2026 (four week averages: total product supplied 21.1 million barrels a day, up 3.2 percent year on year; gasoline 8.7 million, up 1.5 percent; distillate 3.9 million, down 2.4 percent; jet fuel up 1.4 percent). The ledger table in section 2 uses hypothetical inputs, stated in the text. All other figures are as published by the sources named, for the periods stated. Section 6, checked October 29, 2026: EIA, WPSR historical Table 1 data (psw01.xls) (weekly product supplied of total products, finished motor gasoline, distillate fuel oil and kerosene-type jet fuel); EIA Petroleum Supply Monthly series MTTUPUS2, MGFUPUS2, MDIUPUS2 and MKJUPUS2 (monthly product supplied, January 2015 to July 2026); EIA, Petroleum Supply Monthly explanatory notes (product supplied for crude oil assumed zero); EIA, This Week in Petroleum, November 23, 2005 (monthly total product supplied higher than the monthly-from-weekly prorated average 63 percent of the time since January 2000, average 121 thousand barrels a day; gasoline revised upward 62 percent of the time). The month-by-month comparison, counts and averages are PetroEyes arithmetic on those files.