On this page Jump to section
Abstract. Sales tax code errors are usually found by filing or by notice. This paper proposes a sampling method that estimates misclassification in a QuickBooks Online ledger, with strata based on value and change exposure, and a review-cost estimate that keeps the work proportionate.
The method estimates local data quality. It does not determine taxability, nexus, or filing obligations.
The difference between a code error and a tax error
A code error is a mismatch between the transaction and the rule the business intended to apply. A tax error is a mismatch between the return and the law.
The two are related but not identical. A business can apply a consistent, documented internal rule and still need its adviser to confirm the rule is correct.
Sampling addresses the first problem. It gives the adviser a cleaner population to work with.
Define the population
Use the sales transactions in a defined period that carry a tax code, plus those that should carry one. Include invoices, sales receipts, and imported orders.
Exclude transactions that are out of scope by policy, such as wholesale sales with a valid exemption certificate on file, but record the exclusion count.
Note the system boundaries. An order that starts in a storefront and posts to QuickBooks may carry a different code at each step.
Stratify by value and change exposure
Three strata keep the sample practical:
- High value: the largest transactions by amount
- Changed: customers, products, or codes modified during the period
- Routine: the remaining population
Review all high-value items where the count is manageable. Sample the changed and routine strata at rates that fit the review capacity.
Changed items are the most informative. A new product, a new ship-to state, or a modified customer exemption is where the code is most likely to be wrong.
Set the sample size from the decision
The sample does not need to support a precise national estimate. It needs to support a local decision: is the error rate low enough to rely on the current process, or does the mapping need repair?
A sample of fifty to one hundred transactions is often enough to distinguish a rare exception from a systemic problem. State the uncertainty honestly rather than reporting a false precision.
Record the inclusion rule and the sampling method so the next period can repeat it.
Review against written evidence
For each sampled transaction, compare the code with the customer record, the product or service setup, the ship-to address, and any exemption certificate.
Record the result as consistent, inconsistent, or undetermined. Undetermined is common when the certificate is missing or the address is ambiguous.
Do not let the reviewer rely on memory of the rule. Attach the internal taxability matrix or the adviser's written guidance to the review.
Estimate the cost and the benefit
Time the review of a sample item. Multiply by the estimated population size to show the cost of a full review.
Compare that with the value at risk in the high-value and changed strata. This often shows that a focused review of a few hundred transactions covers most of the exposure.
The comparison also justifies the sample size. A larger sample that finds nothing new is not free.
Turn findings into fixes
Group inconsistencies by cause: product setup, customer exemption, ship-to mapping, manual override, or retired code.
Fix the default rather than the individual transaction when the cause is setup. A product with the wrong default code will keep producing exceptions after the sample is closed.
For items that need adviser judgment, prepare a packet with the transaction, the code, the evidence, and the question. Do not decide taxability locally.
Report and repeat
Report the population, sample, inconsistencies, undetermined items, estimated range, and the fixes proposed.
Repeat the sample after the fixes with the same strata. A falling inconsistency rate is the evidence that the mapping repair worked.
Limitations
The method measures code consistency, not legal correctness. A consistent code can still be the wrong treatment.
It depends on the accuracy of the customer and product records used as evidence. If those are stale, the review inherits the error.
And it cannot replace the adviser's review of nexus, registration, and filing positions.
Sources
- Internal Revenue Service, How long should I keep records: https://www.irs.gov/businesses/small-businesses-self-employed/how-long-should-i-keep-records
- Committee of Sponsoring Organizations of the Treadway Commission, Internal Control over Financial Reporting guidance: https://www.coso.org/guidance-on-ic
- PCAOB Auditing Standard 1105, Audit Evidence: https://pcaobus.org/oversight/standards/auditing-standards/details/AS1105
- U.S. Government Accountability Office, Standards for Internal Control in the Federal Government: https://www.gao.gov/green-book