Preserve the release vintage behind an inventory claim

Three dates answer three questions

The observation date identifies the stock reference period. The release date identifies when a report was published. The retrieval date identifies when the researcher obtained the artifact. These can differ by days or years. Keeping only the observation date erases the information needed to distinguish a historical description from a real-time release study.

A weekly row dated Friday should not be treated as public information available at that Friday’s market close without release evidence. Likewise, a file downloaded in 2026 containing 2020 observations can be perfectly useful historical data while lacking proof of the values and timestamps originally published in 2020.

Work through a revision example

Imagine a synthetic inventory observation initially published as 450 million units, later corrected to 452. A study of original release surprises needs the 450 vintage and the original publication context. A study of currently retained historical stock levels may appropriately use 452, provided it labels that purpose and retrieval vintage.

Replacing 450 with 452 in an old trading narrative can introduce information that was not then available. Conversely, using the old value in a current balance reconstruction may ignore a justified correction. Neither vintage is universally preferable; the research question determines which evidence is required.

Distinguish two types of dated source

EIA’s downloadable historical series can support a current-vintage descriptive comparison when retained with retrieval metadata. A dated archived report can support a statement about that edition’s contents. These are complementary artifacts, and their values should not be assumed identical without checking.

The USDA corn studies illustrate a separate forecast-vintage question: May 10 and August 12, 2024 WASDE reports discuss the same 2024/25 marketing year from different information dates. Comparing their forecasts does not reveal the final realized crop balance. Preserve report number, publication date, table and column so the comparison remains anchored to the intended editions.

Build a source record that survives a changing URL

Record the original URL, retrieved artifact, digest, extraction method and units. A latest-report link can serve different content next week. Its presence in a bibliography does not preserve the edition used for a chart. Stable archives and retained lawful source files make a later review more reliable.

Also record corrections or known methodological changes that affect comparability. If a series definition changes, a continuous date column does not guarantee a continuous economic concept. Mark the boundary and explain whether the analysis adjusts, separates or excludes the affected period. Do not silently smooth away a source break.

Make the limitation visible in the result

The inventory lab’s monthly comparison should identify the retained data vintage and its own aggregation method. It is a description of those observations, not a point-in-time backtest or a release-surprise trading study. A current-vintage seasonal baseline can be informative without claiming that all its inputs were historically available unchanged.

If the intended question requires original releases and those artifacts are missing, state that gap exactly. Preserve the descriptive work rather than relabeling it as causal or tradable evidence. Good commodity research is strengthened by a precise account of what the retained vintage can establish and what would require a different dataset.

Keep in the decision record

• Observation, release and retrieval dates serve different purposes.
• Choose the vintage that matches the research question.
• Current-vintage history is not automatically point-in-time evidence.

Source notes

1. EIA — Weekly Underground Natural Gas Storage methodology, updated 10 September 2018
2. EIA — Weekly Petroleum Status Report explanatory notes
3. USDA WASDE 648 — 10 May 2024, U.S. corn table, printed page 12
4. USDA WASDE 651 — 12 August 2024, U.S. corn table, printed page 12

Source dates and scope matter. Historical inventory observations are retained snapshots; forecasts and synthetic curve assumptions are labelled separately. No investment outcome or product coverage is implied.

Continue the analysis