haql-qc
Data qualityFlags every record in a production export that is missing, negative, contradictory or physically impossible — with the rule ID and the source row.
Deterministic software for oil & gas data
Software that catches what the summary hides.
A monthly total still adds up when four wells carry negative volumes. Haql tools read the export you already have and point at the exact row where the numbers contradict each other.
- Runs offline
- Read-only by construction
- Row-level traceability
- Deterministic findings
$ haql-qc -i production.csv -r audit.json rows=15634 wells=7 findings=80 high QC010_NEGATIVE_RATE well NO 15/9-F-12 H date 2008-04-23 row 1984 value −14.19 Sm3 (BORE_WAT_VOL) # read-only: the export is never modified # input SHA-256 recorded in every report # identical bytes in → identical findings out
An example finding from the published validation run. Every finding carries a stable rule ID and a 1-indexed source row; the report records the SHA-256 of the exact input audited.
Trans-Alaska Pipeline · Craig McCaa / BLM Alaska · Public domain
The suite
Ten tools. Each one answers one question you already ask.
Eight commercial products — six self-contained binaries, one independent simulation audit and one purchaser-statement audit — plus two MIT libraries. All share one stance: read-only, offline, deterministic, traced to a row in your own file.
haql-closeout
Month-endIs this month complete and reconciled enough to lock? Cross-checks well tests, downtime, field balances and date continuity.
haql-45q
45Q MRVTurns a year of capture, injection, monitoring and lab records into a defensible 45Q claim: mass balance and a certification pack.
haql-emission
Emissions MRVReconciles a GHG inventory from activity, flare and vent exports — mass balance, coverage, and an EU ETS allowance position.
haql-rrc
RRC filingChecks a Texas RRC Form PR export before it is filed — operator numbers, disposition codes, district range, flare remarks, filing status.
haql-dataroom
A&D dataroomCross-checks an acquisition dataroom offline: production, revenue and severance tables, reserves and LAS headers — every contradiction traced to a source row.
haql-revenue
Revenue auditReconciles produced volumes against purchaser statements: missing statements, price below contract, volume short, duplicate lines and unscheduled deductions, ranked by dollars.
haql-engine
Simulation auditIndependent audit of an Eclipse/OPM run: mass balance, convergence, reproduction and input integrity — with provenance receipts, quarterly revision diffs and ensemble audits. The engine is the verification instrument; the audit is the deliverable, and it reproduces byte-identically on your machine.
haql-units
MITField-to-SI unit conversions with explicit time bases and round-trip checking, so a unit mismatch is caught rather than inherited.
haql-coords
MITWellbore geometry and coordinate math engineers use: minimum curvature, DLS, UTM/WGS84 — with round-trip verification.
Builds are sent as private links; Linux x86_64 is available now, Windows and macOS on request. See the downloads page for the request process, hashes and the 30-day evaluation.
From the field
The numbers start at the wellhead. They end up on a statement.
Between the two, a hundred small handoffs happen. Haql reads the export at the end of that chain and points back to the row where it stopped adding up.
Rates, downtime, well tests and the daily records behind them.
Volumes, destinations and the paper trail between field and purchaser.
Purchaser lines, deductions, prices and the dollars that reach the ledger.
Field photography is illustrative and does not depict customer assets. Credits: Craig McCaa / BLM Alaska (public domain), Roy Luck (CC BY 2.0), U.S. Department of Energy (public domain), GuavaTrain (CC0), Maxx@night (CC BY 2.0). Full image credits.
How it works
One step added. Nothing replaced.
No change to how you capture or account for production. Export the file you already have, run the check, and decide what to fix in your source of truth.
- ExportYour production, monitoring or MRV data — no change to your capture or accounting system.
- RunRun the CLI yourself offline, or email the file for a free sample check. Both use the same rules.
- ReviewEvery finding names the rule, the well, the date and the 1-indexed source row.
- FixYou correct your source of truth. The tool never writes to your file.
Read-only in both directions · no network call · no telemetry · no upload · no model.
The field doesn't stop
The field runs around the clock. The reconciliation shouldn't wait for month-end.
Production, downtime, run tickets, purchaser statements — the data is already there. Haql turns it into a repeatable check, not a once-a-quarter scramble.
Pricing
Priced against the hours you already spend.
Licenses are annual and verified offline. Evaluation runs for 30 days with no license key, and the two MIT libraries are free forever.
Sample check
Free
One month of data, on the house. No key, no seat limit, no card. The point is to hit one finding you did not know about.
Send me your CSVTeam
$349/month
Annual commitment, up to 10 users, unlimited fields. Roughly one engineer-day of loaded cost.
On-premise
$1,750/month
Annual, per site, unlimited users on that site. On-premise is not a feature — under most data-residency policies it is the premise.
Field-specific work is available separately at $175/hour — the per-basin knowledge of how your operator's export actually behaves. haql-units and haql-coords are MIT and free forever.
Trust & FAQ
Why hand a month of data to a stranger.
The six commercial data tools are single self-contained binaries that run entirely on your machine, offline, with no installer and no Python. haql-revenue is a licensed Python 3.8+ package, standard library only, with a local review UI — still no network call. The simulation audit is a service: send a deck and run, and we return a signed report. The MIT libraries install from PyPI. None of these tools opens a network connection. The rules are numbered permanently — adding a rule never renumbers an existing one.
- Never corrupt source data. Read-only by construction. Correcting requires knowing which of two contradictory fields is wrong, and the data cannot supply that.
- Never guess. Units are declared by you and recorded in the report as your declaration.
- Never be untraceable. Every finding ties to a stable rule ID, a row you can open, and a hash of the input.
haql-qc is validated against one public field — Equinor's Volve dataset, 15,634 daily and 526 monthly rows, producing 80 findings. One field is a starting point, not a general guarantee, and we will not claim otherwise. Download the sample audit report.
Does it use AI?
No. Every rule is deterministic arithmetic on your data — comparisons, monotonicity checks, medians. No model, no network call. Running the same file twice gives the same answer.
Does my data leave my computer?
Run locally, nothing leaves your machine — the code makes no outbound calls. If you use the free sample check, you choose to email us the file, and you can ask for it back anonymised or deleted.
What do I need to install?
Nothing, if you use the free sample check. To run the tools yourself, request a private build link from the downloads page — no Python, nothing else to configure. haql-revenue is the exception: it runs from Python 3.8+ with no third-party packages. Linux x86_64 is available now; Windows and macOS builds are made on request.
How is it licensed?
The commercial tools are proprietary: running them requires a signed Ed25519 token carrying company, tier, seats and expiry, verified entirely offline. No activation server. haql-units and haql-coords are MIT and need no license.
Send us a month of data.
Tell us which field, anonymise it if you need to, and mention anything unusual about how your operator reports — a sign convention, a units quirk, a column that means something other than its name.
This site is static, hosted on GitHub Pages. No cookies, no trackers, no analytics, no backend. Findings are advisory; the judgment is yours.