How Reliable Are Jurisdiction Fields in Truck Fuel-Purchase Records?

Published September 3, 2026.

Research question and practical stakes

How can a carrier test the reliability of jurisdiction fields in fuel-purchase records before using them in an administrative reconciliation? A merchant name can include one place while the transaction occurred in another. Card feeds may supply city and state fields, location identifiers, reversals, corrections, or only free text. Treating every populated state code as verified creates false precision.

This brief proposes a data-quality study, not a fuel-tax calculation. It does not classify taxable activity, prepare a return, decide credit eligibility, determine route mileage, or interpret IFTA or other tax rules. A qualified tax professional and the carrier's authorized finance owner retain those decisions.

Evidence used

IFTA's published manuals provide program context, the IRS excise-tax page establishes a federal tax-information boundary, EIA publishes fuel-market information, and BTS supplies freight-measure context. These sources do not certify a carrier's transaction feed or prescribe the exact quality tests below. The tests are a local research design.

Choose a closed transaction period and freeze the native card export before enrichment. Include purchases, reversals, credits, declined transactions if present, and later corrections. Store the card or account token permitted by policy, transaction identifier, authorization identifier, event time and timezone, posting time, merchant identifier, merchant name, street, city, jurisdiction, postal code, product, quantity, amount, and source-file reference.

Establish the transaction population

The primary unit is a native transaction event. Link a reversal to its original event rather than deleting both before quality review. If the reporting purpose later uses net purchases, publish that as a derived view. Keeping the event population makes duplicate feeds, partial reversals, and delayed corrections observable.

Define how repeated authorizations, split purchases, and file reimports are recognized. A same amount at the same merchant is not enough to declare a duplicate. Use stable provider identifiers when available and place uncertain matches in a review class. Report event counts before and after documented deduplication.

Treat jurisdiction as a claim with provenance

Create separate fields for jurisdiction supplied by the card processor, jurisdiction parsed from an address, jurisdiction returned by a carrier-approved merchant table, and jurisdiction determined by an authorized reviewer. Never collapse them into one field without retaining origin. The final administrative field should show its method and evidence grade.

A two-letter code may be absent, malformed, or inconsistent with the city and postal code. Free-text merchant labels can contain a headquarters location or brand reference unrelated to the point of sale. Location identifiers can be reused after ownership or processor changes. These are reasons to test the field, not reasons to guess.

An example that resists automatic cleanup

A transaction lists "River Junction Travel Center" with a processor jurisdiction of State A. The street and postal code resolve under the approved reference table to State B, and the purchase timestamp falls near a route crossing. Route proximity does not settle the conflict. Preserve the processor value and reference-table result, record the table version, and route the discrepancy.

The carrier's tax owner may have an approved hierarchy or request source documentation. Administrative research should not choose the value that fits reported mileage, produces a desired tax result, or matches the nearest GPS point. Those choices turn a quality screen into an unsupported tax determination.

Build transparent quality classes

Classify events as source-complete, source-missing, format-invalid, cross-field conflict, reference-table unmatched, reversal-unlinked, duplicate-candidate, or resolved under an approved rule. An event may carry more than one flag. Publish each class count and the combinations that materially affect the review.

Calculate completeness for native merchant identifier, event time, quantity, product, and processor jurisdiction. Calculate agreement only among records eligible for a stated comparison. For example, city-to-jurisdiction agreement excludes missing cities from its denominator but reports those missing cities beside the result. This keeps a high agreement percentage from concealing weak coverage.

Sample by risk of error, not by convenience

Review all cross-field conflicts and unmatched high-consequence records within the approved scope, then draw samples from common merchants, uncommon merchants, reversals, nighttime postings, border-area locations, and clean records. Border-area sampling can expose reference errors, but physical proximity alone must not be used to correct the jurisdiction.

A second reviewer should reproduce the classification from the frozen export and the named reference-table version. Record disagreements about parsing, merchant matching, and reversal linkage. If the rule changes, rerun the full frozen cohort rather than repairing only the sampled records.

What the measures can support

Data-quality measures can identify merchants requiring a maintained crosswalk, processor fields that arrive inconsistently, unlinked corrections, or queue items that need owner review. They can support a cleaner handoff to accounting or tax specialists. They cannot determine a filing position or prove where taxable fuel was consumed.

TruckingVA can import carrier-supplied exports, preserve transaction lineage, apply approved formatting rules, maintain an authorized reference table, calculate quality classes, and prepare exception packets. It should not infer a jurisdiction from GPS, amend a tax record, choose a filing treatment, or communicate a tax conclusion.

Limitations and reproducibility

Processor exports vary and may omit native location attributes. Merchant reference tables become stale. Reversals can post outside the selected period, and timezones can make a transaction appear on a different day. A cohort should include a stated look-forward window for corrections or disclose that it does not.

Geocoding quality depends on its input and reference version. Even a strong geographic match does not prove that a transaction field should receive a particular tax treatment. Preserve the query, transformation logic, source hashes, exclusions, and review decisions so another analyst can reproduce the administrative result.

Evidence-led conclusion

Jurisdiction data is dependable only to the extent that its source and conflicts remain visible. The evidence supports this restrained conclusion: test merchant jurisdiction as a sourced field, disclose unresolved locations and reversals, and keep tax interpretation outside the administrative measure. That design helps the carrier locate weak records before specialist review without allowing a clean dashboard to substitute for tax authority.

Sources

1. https://www.iftach.org/manuals.php, retrieved 2026-09-03
2. https://www.irs.gov/businesses/small-businesses-self-employed/excise-tax, retrieved 2026-09-03
3. https://www.eia.gov/petroleum/gasdiesel/, retrieved 2026-09-03
4. https://www.bts.gov/content/average-freight-revenue-ton-mile, retrieved 2026-09-03