How to Read the Weekly EIA Petroleum and Natural Gas Inventory Numbers

# How to Read the Weekly EIA Petroleum and Natural Gas Inventory Numbers

Takeaway: The weekly inventory level on its own is nearly meaningless, because oil and gas stocks follow a strong seasonal calendar. The honest lens is the deviation from the same calendar week averaged over the prior five years. A single week's build or draw can move prices sharply on release, but on its own it is a noisy guide to the underlying supply and demand trend.

Kresmion Research. Information only, not investment advice. Figures are for the week ending Friday 2026-07-10, pulled 2026-07-19.

Source: Kresmion production database (`eia_inventories`), cross-checked to the public commodities inventories endpoint; deviations defined in methodology section 21. As of the week ending 2026-07-10, pulled 2026-07-19.

What the weekly EIA releases actually measure

A stock, not a flow

Every week the U.S. Energy Information Administration (EIA) publishes how much crude oil, gasoline, distillate, and natural gas is physically sitting in tanks and underground caverns. This is a stock: a snapshot of a level at one moment, not a rate of movement. When the level rises week to week it is called a build; when it falls it is a draw. That weekly change is a flow you derive by subtracting two consecutive snapshots.

The four series Kresmion tracks (and their units)

Kresmion carries four EIA weekly series (the full list and definitions are in methodology section 21):

• Commercial crude oil stocks excluding the Strategic Petroleum Reserve (series WCESTUS1)
• Total motor gasoline stocks (WGTSTUS1)
• Distillate stocks, meaning diesel and heating oil (WDISTUS1)
• Working gas in underground storage, Lower 48 states (NW2_EPG0_SWO_R48_BCF)

The first three are reported in thousand barrels (the EIA unit label is MBBL, so 409,665 MBBL equals 409.665 million barrels). Natural gas is reported in billion cubic feet (Bcf). Because the units differ, a barrel figure and a Bcf figure are never compared to each other, and mixing MBBL up with millions of barrels is the easiest way to misstate one of these numbers by a factor of a thousand. The related futures contracts a reader will see quoted are crude (CL=F), natural gas (NG=F), RBOB gasoline (RB=F), and heating oil (HO=F).

Two reports, two release days

Two separate EIA reports feed these numbers. The Weekly Petroleum Status Report covers crude and refined products and normally publishes Wednesdays after 10:30 a.m. Eastern (it shifts later in holiday weeks). The Weekly Natural Gas Storage Report covers underground gas and publishes Thursdays at 10:30 a.m. Eastern. Each covers the week ending the prior Friday, so the figure is weekly by design and stale between releases. For the current cycle, EIA released the week-ending-2026-07-10 petroleum data on 2026-07-15 and the gas data on 2026-07-16, with the next releases dated 2026-07-22 and 2026-07-23 (EIA petroleum page; EIA gas data series).

Why the raw level is nearly meaningless

Inventories follow a strong, predictable annual cycle, so the bare level tells you almost nothing on its own. Natural gas is injected (built up) roughly April to October when heating demand is low, then withdrawn November to March during winter, so storage normally peaks in late autumn and bottoms in early spring. Gasoline is drawn down through the summer driving season and rebuilt in winter. Distillate is built through the non-heating season, roughly spring through autumn, then drawn down in winter, which is why even a mid-summer week can post a build (as this July week's distillate figure does). Crude swings with refinery maintenance, the "turnarounds" that cluster in spring and fall (CME Group education). The identical level in January and in July can mean opposite things, so a number with no seasonal context is not a signal. Judging inventories by the raw level is like counting a store's shelves the morning after a holiday rush versus a quiet weekday: a half-empty shelf is normal after a rush and striking on a slow day, so you compare it to what is normal for that day.

The five-year same-week average is the honest lens

The number that carries meaning is the deviation from the five-year average for the same calendar week. Kresmion finds the matching week in each of the prior five years and expresses the current level as a percentage above or below that benchmark (the exact definition lives in methodology section 21). Subtracting the seasonal "normal for this week" strips out the annual cycle and leaves the part that is actually informative: whether inventories are higher or lower than usual for this time of year.

Here is what that lens shows for the week ending 2026-07-10, from the Kresmion commodities inventories dashboard (members):

• Crude ex-SPR sat at 409.665 million barrels, about 6.3% below its five-year same-week average of roughly 437,220 thousand barrels.
• Gasoline sat at 210,529 thousand barrels, about 8.4% below its five-year average of roughly 229,819 thousand barrels.
• Distillate sat at 108,175 thousand barrels, about 10.9% below its five-year average of roughly 121,348 thousand barrels.
• Natural gas working storage sat at 3,024 Bcf, about 5.7% above its five-year average of roughly 2,862 Bcf.

These deviations line up with EIA's own published highlights for the same week, which describe crude as about 6% below, gasoline about 8% below, and distillate about 11% below the five-year average (EIA highlights PDF). A level below the seasonal average is what analysts mean by a "tight" balance and a level above it is "ample" or "loose"; this is a comparison to history, not a forecast and not a view on prices.

Week-on-week: market-moving on release, but a noisy guide to the trend

A single week's build or draw is a jumpy figure, and yet it is also one of the most closely watched scheduled releases in energy. On release, the gap between the reported change and the market's consensus expectation can move crude and natural gas futures sharply within minutes, so the print is a real, tradeable event rather than something the market shrugs off (CME Group). What makes it "noisy" is a separate point: as a read on the underlying supply and demand balance, any one week tells you little. Over the latest week, crude drew 1,692 thousand barrels, gasoline drew 1,533 thousand barrels, distillate built 4,556 thousand barrels, and natural gas injected 41 Bcf (from 2026-07-03 to 2026-07-10, Kresmion database). The weekly petroleum numbers are survey-based estimates that get revised against more accurate monthly data, and the crude balance carries an "adjustment" term that can be large. Cargo timing, refinery outages, holidays, and weather can move one week and reverse the next. The durable signal lives in the trend and in the position relative to the seasonal band, not in any one week's wiggle. As exchange education notes, the reported change matters most relative to expectations, and a draw smaller than the seasonal norm can actually leave inventories looser versus the five-year seasonal average than they were before (CME Group).

Two things these numbers do not include

The crude series is commercial crude, and it excludes the Strategic Petroleum Reserve, the government stockpile filled and drained by policy rather than by market supply and demand; do not read it as total U.S. crude. For natural gas, the reported figure is working gas, the portion that can actually be withdrawn. It excludes the "base" or "cushion" gas that must stay underground to keep reservoir pressure (EIA gas methodology), so "storage" here means usable inventory, not the total gas in the ground.

Honest limitations

• Four headline series only. Kresmion tracks national crude ex-SPR, gasoline, distillate, and Lower 48 working gas. It does not break out the Cushing, Oklahoma hub, the SPR itself, propane, product-export splits, or the regional PADD and gas-region detail EIA also publishes, so a national number can mask regional tightness.
• Weekly data is revised. Petroleum weekly figures are estimates later reconciled to the monthly Petroleum Supply Monthly, and weekly gas storage differs from final monthly data by an average absolute 11 Bcf, or 0.4% (EIA, January 2010 to April 2026) (EIA weekly versus monthly). Treat any single week as provisional.
• The five-year window can contain an unusual instance of this week. Because this is a July reading, the benchmark averages the same July week across 2021 to 2025, so only a July anomaly can distort it: a summer with a major Gulf Coast hurricane that knocked out refining, an unusually weak or strong driving season, or a year of unusual economic conditions in that particular July. A "normal" built partly from one abnormal July is not a perfectly clean normal. A winter cold snap, a polar-vortex week, or a spring demand collapse would land in a different calendar week's benchmark, not in this July one.
• The deviation is a derived, approximate figure. It depends on how "same week" is matched across prior years. Kresmion's endpoint and an independent reconstruction from the same database agree in sign and closely in magnitude but can differ by up to about 0.9 percentage point (for natural gas, roughly 5.7% versus 6.6% above average), so deviations are cited here as approximations. The raw levels, by contrast, are exact and were cross-checked to the barrel and the Bcf.
• Storage capacity and the futures curve are not modeled. A level near a tank or cavern's physical limit behaves differently from the same level with room to spare; Kresmion does not model capacity, "days of supply," or the futures curve, so nothing here speaks to carry, roll, or whether the curve sits in contango or backwardation.
• Descriptive, not predictive. The deviation says where inventories sit versus their own history, not where prices go; production, exports, geopolitics, and demand all move independently.

Methodology and sources

Kresmion ingests the four EIA weekly series named above into its `eia_inventories` table and recomputes, per series, the week-over-week change, the five-year same-week average, the percentage deviation from it, and a 52-week history tail (field definitions in methodology section 21). The five-year benchmark requires a matching-week observation in at least three of the prior five years, otherwise the comparison is left blank rather than shown as zero. Levels here are single-sourced to the Kresmion production database and were cross-checked against the public commodities inventories endpoint to the barrel and the Bcf. History depth is ample: crude and distillate go back to 1982 (2,285 weekly rows), gasoline to 1990 (1,906 rows), and natural gas storage to 2010 (863 rows); all four series carried a latest period of 2026-07-10, fetched 2026-07-19 at about 15:18 UTC. The primary sources are EIA's Weekly Petroleum Status Report and Weekly Natural Gas Storage Report at www.eia.gov; members can view the live tracker on the Kresmion commodities inventories dashboard.

Frequently asked questions

Why compare to a five-year average instead of just last week or last year?

Because inventories are seasonal, and both shortcuts can mislead. Last week is the same season, so comparing to it only shows the very short-term flow, which is noisy and often revised. Last year is the same calendar week, but a single prior year can be a one-off. For this July reading, last year means the July 2025 week, which could have been distorted by a one-off event such as a Gulf Coast hurricane that shut refineries or an unusually weak summer driving season. Averaging the same calendar week across five prior years smooths out one-off years while still holding the season fixed, so what remains is a cleaner read on whether stocks are unusually high or low for this time of year, much like comparing today's temperature to the normal high for the date rather than to yesterday.

Does a weekly "draw" mean supply is tightening?

Not by itself. A draw simply means the level fell that week, which is often exactly what the calendar expects (crude and gasoline typically draw in summer). What matters is whether the draw is larger or smaller than the seasonal norm and how the level now sits against its five-year same-week average; a draw smaller than usual can leave inventories looser versus that seasonal norm than they were before. As exchange education puts it, the reported change matters most relative to expectations, not in isolation (CME Group). Kresmion presents these as descriptive readings, not as trading signals.

Sources

1. EIA, Weekly Petroleum Status Report (release schedule, coverage, SPR, refinery utilization): https://www.eia.gov/petroleum/supply/weekly/ 2. EIA, WPSR highlights PDF (five-year-average language: crude about 6% below, gasoline about 8% below, distillate about 11% below; refinery utilization 96.2%; crude inputs 17.1 million barrels per day; week ending 2026-07-10): https://www.eia.gov/petroleum/supply/weekly/pdf/highlights.pdf 3. EIA, Weekly Natural Gas Storage Report methodology (working versus base gas, EIA-912 survey, Thursday 10:30 a.m. ET, six regions): https://ir.eia.gov/ngs/methodology.html 4. EIA, Differences Between Monthly and Weekly Working Gas in Storage (average absolute error 11 Bcf, or 0.4%, January 2010 to April 2026): https://ir.eia.gov/ngs/mthdiff.html 5. EIA, Weekly Working Gas in Underground Storage data series (Lower 48 total 3,024 Bcf, week ending 2026-07-10, regional breakdown): https://www.eia.gov/dnav/ng/ng_stor_wkly_s1_w.htm 6. CME Group OpenMarkets, What API and EIA Data Reveal About Crude Oil Markets (builds and draws, market reaction to the releases, refinery turnaround seasons, API versus EIA survey differences, correlation roughly 0.6 to 0.8): https://www.cmegroup.com/openmarkets/energy/2025/What-API-and-EIA-Data-Reveal-About-Crude-Oil-Markets.html 7. Kresmion Research, methodology section 21 (series list and same-week average definition): https://kresmion.com/about/methodology 8. EIA primary-source portal: https://www.eia.gov 9. Kresmion commodities inventories dashboard (members): live tracker for the four series.