Rate Accuracy

Once a day an automated job reads published sales tax rates and SaaS taxability from up to eight outside sources and compares them with the rates the AgentTax engine applies. This page shows the latest run, how the comparison works, and each difference the run found, with our reviewed explanation where we have one.

Latest run

Source: /api/v1/rates/verification, cached for up to one hour.

How the cross-check works

The job runs every day at 12:00 UTC. It queries these sources in parallel. A source that cannot be read, or that returns too little data to parse, is marked as failed for that run and is not compared.

What is compared

• Each state rate a source publishes is compared with the rate the engine applies on the run date, including scheduled rate changes that have taken effect. Differences of 0.25 percentage points or less are not reported.
• Local DOR rates are compared ZIP by ZIP with the engine's local rate table.
• When two sources publish different rates for the same state, that is recorded as a disagreement between sources, even if one of them matches the engine, unless both figures are documented definition differences (below).
• States with no state sales tax (AK, DE, MT, NH, OR) must carry a zero rate in the engine. Anything else is recorded as critical.
• SaaS taxability from the SaaS-specific sources is compared with the engine's digital taxability flag. Those sources cover SaaS only, while the engine flag also covers other digital services and goods, so a difference is recorded as a scope note rather than a discrepancy.

Documented definition differences

For four states (VA, CA, UT, LA) a source may legitimately publish a different figure from the engine, such as the state-only component of a rate that the engine combines with a mandatory local rate. A source quoting one of those documented figures is recorded as a scope note. The exception applies only while the engine carries the rate the rule was written for. If the engine rate moves, that is itself recorded as a critical discrepancy.

Severity

• High: any state rate difference against a State DOR page, a local rate difference of more than 0.5 percentage points against a Local DOR authority, or a difference of more than 0.5 percentage points against a primary source.
• Medium: a smaller difference against a primary source or a Local DOR authority, or more than 1 percentage point against a secondary source.
• Low: a smaller difference against a secondary source.
• Review: two sources disagree with each other.
• Critical: a non-zero rate in a no-sales-tax state, or an engine rate that no longer matches a documented definition rule.

Minimum coverage

A run needs at least four sources to return data. With fewer, the run is recorded as unverified, it does not report the rates as checked, and it raises an alert, because a zero discrepancy count would only mean nothing was compared. Where a Department of Revenue page cannot be read and a stored reference value is used instead, that value is reported but not counted as a successful read.

What happens with the result

Every run is stored in a run log and sent to us by email, including clean runs. The job never changes a rate. A discrepancy is reviewed by hand before any rate is updated.

Open discrepancies

Each difference recorded by the latest run. Where we have reviewed the difference, the explanation is shown. Anything not yet reviewed is marked Under review.

Loading the latest run's list.

Explanations on file

What this does not cover

• Capital gains rates. The daily job does not check them. They are reviewed manually, and the /api/v1/rates/capital-gains response carries its own verification block.
• Local rates outside the ZIPs that have a configured local authority. Those are not compared on the daily run.
• Taxability rules beyond the yes or no SaaS answer the sources publish. Category-level rules in the engine, such as partial exemptions, are not compared.
• Agreement with outside sources is evidence, not proof. Third-party tables can be out of date, as the DC entry above shows. This page is not tax advice.

Related

• Tax Methodology: how the engine calculates sales tax, capital gains and 1099 obligations.
• Raw run status as JSON: /api/v1/rates/verification