Primary audience
Commodity analysts, forecasting teams, banks, consultants, data scientists, and model-risk reviewers who need to reproduce what a dataset showed on a past date.
Energy data research
Experiment 2 · point-in-time production data
Oil production vintages: a revision-history experiment
Can repeated snapshots reveal how federal and state oil-production estimates change after first publication—and create an auditable ‘as known on’ history for models and backtests?
Status
collection probe
Initial scope
EIA API documentation and crude production browser, Texas RRC downloads, and New Mexico OCD production reports
Collection cadence
Daily availability checks plus immutable snapshots at each official release
Last reviewed
2026-08-11
This is not a live production-vintage product or a guarantee of historical revision coverage. It is an active snapshot experiment.
Decision context
What this experiment must decide
Decision question
Do official oil-production series revise often and materially enough—and disappear thoroughly enough—to justify a retained point-in-time API?
Why collect now
A vintage archive can only be built prospectively. If an official source overwrites a prior estimate, waiting until customers ask for the old value may make reconstruction impossible.
P · Problem
Problem
Most production datasets answer ‘what is the latest value?’ A backtest, audit, or forecast review often asks a different question: ‘what value was available when the decision was made?’ If a regulator revises prior months in place, a normal latest-only database cannot answer it.
Federal and state systems use different publication schedules, identifiers, units, reporting lags, and correction processes. A state total may also represent a different reporting basis from a federal estimate. Combining them without source, vintage, and methodology metadata creates false precision.
The central uncertainty is not whether production data exists. It is whether successive releases contain material, decision-relevant revisions that users cannot recover elsewhere. That must be observed release by release.
A · Approach
Approach
The collection design is intentionally narrower than a production data platform. It preserves evidence, makes uncertainty explicit, and produces a measurable build/no-build decision.
- 1
Pin the query
Define exact source routes, filters, units, frequency, sort order, and requested date range so successive responses are comparable.
- 2
Snapshot the same history
At every release, retrieve the entire chosen observation window—not only the newest month—and store the raw response immutably.
- 3
Separate vintage from observation
Record the retrieval/release vintage independently from the month or period described by each production value.
- 4
Diff at field level
Calculate first reported value, latest value, absolute revision, percentage revision, number of releases changed, and days to stabilization.
- 5
Preserve methodology boundaries
Do not reconcile EIA estimates and state-reported production into one number until definitions, lags, and units are explicitly mapped.
- 6
Test model impact
Run representative forecasts or backtests using first-release versus latest-vintage data and quantify whether conclusions change.
R · Research
Research
These are current findings and their decision implications. They are not promises about future OilPriceAPI coverage.
| Issue | Observed finding | Decision implication |
|---|---|---|
| EIA provides parameterized API access | EIA documents its version 2 API and a petroleum crude-oil production browser. | The federal source can support repeatable pinned queries, but a key and exact route configuration are operational requirements. |
| Texas publishes several production products | The RRC lists monthly ledgers, statewide files, and a Production Data Query CSV dump with stated publication schedules. | The experiment must choose one stable extract first; mixing products would confound the revision comparison. |
| Texas values can roll through systems | RRC documentation describes monthly processes and a 26-month production system for some datasets. | Prospective snapshots may preserve states that are otherwise difficult to reconstruct after roll-off or correction. |
| New Mexico discloses reporting lag | OCD states that C-115 reported production updates nightly with a 45-day lag, while other reports have separate schedules. | Freshness and finality must be represented separately; a late value is not necessarily a collector failure. |
| State schemas are not interchangeable | Texas offers lease- and district-oriented products, while New Mexico exposes C-115 and statewide summaries. | The initial moat is point-in-time provenance per source, not a premature national production total. |
M · Measurement
Measurement
Primary decision metric
Decision-relevant revision yield: materially changed historical observations captured per source release, weighted by buyer-reported model impact.
Revision incidence
Historical observations changed in a new release divided by observations compared.
Revision magnitude
Absolute and percentage change from first-reported to later values.
Revision depth
How far back a release changes prior observation periods.
Time to stabilization
Days and releases between first publication and last observed change.
Snapshot reliability
Successful, comparable captures divided by scheduled release captures.
Schema drift
Source format or identifier changes requiring collector intervention.
Reconstruction gap
Vintages unavailable from the source after publication but retained by the probe.
Model impact
Difference in selected forecasts or backtests when first-release data replaces revised data.
Continue when
- ✓At least two priority sources revise historical observations in ways that cannot be reconstructed easily from their current interface.
- ✓Snapshots complete at least 99% of scheduled release captures or surface explicit failure status.
- ✓At least three target users demonstrate a model, audit, or research workflow where vintage choice changes reproducibility or conclusions.
Stop or redesign when
- ×Observed revisions are negligible for target workflows or authoritative sources already expose complete vintage histories.
- ×Source definitions change too frequently to maintain comparable series at viable cost.
Initial collection
Started
2026-08-11T21:35:40Z
Current status
All three official source-family pages returned HTTP 200; no release-to-release comparison has occurred yet.
Baseline observations
- •EIA documents a versioned API and exposes a crude-oil production browser; an API key is still required for the pinned data query.
- •Texas RRC lists monthly production extracts and states when several downloads are updated.
- •New Mexico OCD states a 45-day lag for C-115 reported production and nightly updates for that reporting surface.
- •No revision-frequency conclusion is possible from a single baseline; successive release captures are the experiment.
Next capture
Pin one EIA series and one extract each from Texas and New Mexico, then preserve the complete selected history at every release.
Proposed record schema
The schema is a test contract, not a public endpoint specification. Unknown values remain unknown, and every derived field must retain source evidence.
- source_series
- Exact agency route, series, filters, and unit.
- observation_period
- Month or reporting period described by the value.
- vintage_at
- Release or retrieval time for this version.
- value
- Reported value without cross-source reconciliation.
- revision_number
- Observed sequence of distinct values for the same source period.
- first_reported_value
- Earliest captured value for that source period.
- revision_delta
- Absolute and percentage change from the prior vintage.
- source_hash
- Hash of the raw response that supports the version.
Sources and scope
OilPriceAPI Research prepared this page to test a product decision. The method combines direct-source review, immutable collection probes, and explicit decision gates. External links establish what each source publishes; they do not grant OilPriceAPI redistribution rights or prove complete coverage.
EIA API technical documentation
U.S. Energy Information Administration
Supports: API v2 query behavior and reproducible parameterization.
Limitation: Documentation is not itself a production vintage, and API access requires configuration and a key.
Texas production datasets available for download
Railroad Commission of Texas
Supports: Official extract inventory, formats, and stated monthly schedules.
Limitation: The page describes multiple products with different structures; the experiment must not treat them as one series.
New Mexico OCD statistics
New Mexico Energy, Minerals and Natural Resources Department
Supports: Official production report inventory, update frequency, and disclosed C-115 lag.
Limitation: Report pages and downloads may have different update timestamps and aggregation levels.
Help validate the next decision
If this dataset is part of your workflow, tell us which sources matter, what failure looks like, and what evidence would make the data useful enough to buy. A conversation is evidence; a page view is not.