Outsource Payroll CompanyPowered by StealthAgents
Table of contents
  1. Research finding
  2. Methodology
  3. Key takeaways
  4. Question and scope
  5. Evidence classes
  6. Decision log
  7. Reconstruction task
  8. Scoring and disagreement
  9. Privacy and access
  10. Replay across cycles
  11. Limits and decision link
  12. Index quality
  13. Provider evidence
  14. Exception carryover review
  15. Reporting the reconstruction result
  16. Artifact naming consistency
  17. Retest after the index change
  18. Sources
  19. FAQs
  20. Related research

Research brief · Published

Payroll close evidence sufficiency

A study design for testing whether a payroll close package lets an independent reviewer rebuild the run, its decisions, and its open exceptions.

Research finding

Test the close package by asking a reviewer who did not run payroll to rebuild it.

4Headline measureSee numbered sources below.

Methodology

This brief triangulates the headline measure against official Philippine government, regulatory, development, and labor sources. It translates evidence into an operating control and separates context from recommendations.

Key stats and interpretation
MeasureInterpretation
4 reconstruction tasks scored for each payroll close package.Context signal for planning, not a promise about an individual worker or provider.
4 source recordsPrimary source links are listed and numbered below for review.

Key takeaways

  • Test the close package by asking a reviewer who did not run payroll to rebuild it.
  • Score population, amounts, release, and exception decisions as separate tasks.
  • A polished report without an owner or period reference is weak evidence.
  • Keep open exceptions visible instead of treating release as closure.

Question and scope

This study asks whether a payroll close package contains enough evidence for an independent reviewer to rebuild the run. The population is the close packages for one employer's pay groups across three consecutive cycles. The unit is one package with its index, artifacts, decisions, and open exceptions. The study measures reconstruction rather than document count, because a large folder can still omit the one record that explains a manual adjustment. It separates population, amounts, release, and exception decisions as four tasks. The result describes the sufficiency of the package for this process, not the correctness of the payroll calculations themselves.

Evidence classes

Define the expected evidence classes before reviewing a package. Include the payroll population, source records, gross-to-net reports, tax and deduction reports, the release confirmation, the decision log, and the open-exception list. Record whether each class is present, missing, or not applicable for the cycle. A class that is present but lacks a period or owner is partial evidence. Keep the classes stable across cycles so the comparison is meaningful. If a class is added after the first review, apply it to all cycles or mark the earlier result as mixed rather than silently improving the score.

Decision log

The decision log is the evidence class most often missing from an otherwise complete package. Record each manual adjustment, off-cycle payment, tax correction, and exception that required an owner decision. For each item, capture the question, the evidence reviewed, the decision, the approver, and the date. A log entry that says only approved without naming what was approved is insufficient. Compare the log with the payroll output to find adjustments that appear in the numbers but not in the record. The study should report both missing entries and entries that cannot be linked to a specific pay period.

Reconstruction task

Ask a reviewer who did not prepare the close to rebuild four things from the package: the employee population, the gross-to-net result, the release confirmation, and the disposition of each open exception. Give the reviewer the index and the approved artifacts but not the original commentary. Record where the reviewer stops, what they cannot find, and which document they used to continue. A successful reconstruction does not prove the numbers are correct, but a failed reconstruction proves the package is insufficient for its purpose. Keep the reviewer's questions as evidence of the specific gap.

Scoring and disagreement

Score each reconstruction task as complete, partial, or blocked, and record the evidence that supports the score. Have a second reviewer score the same package independently. Compare the scores and inspect the artifacts where they differ. Disagreement about whether a document is sufficient is useful evidence about the standard, not a nuisance to average away. If the reviewers cannot agree on what counts as release confirmation, the definition needs repair before the next cycle. Report the disagreement rate and the reasons so the team can distinguish a weak package from an unclear rule.

Privacy and access

Use de-identified extracts where possible and keep the identity crosswalk in the approved payroll system. Limit the reviewer's access to the artifacts needed for the reconstruction, and remove access when the review closes. Record redactions and unavailable documents rather than reconstructing them from memory. A package that is easy to review but exposes unnecessary personal data is not a good control. Privacy handling and evidence sufficiency are separate findings, and both should be reported. The reviewer should not need to open full bank details to confirm that a release occurred.

Replay across cycles

Repeat the reconstruction on three cycles and compare the pattern. A package that works for a routine cycle may fail during year-end, a provider change, or a period with several manual adjustments. Record the cycle type and the events that made it different. If the package fails only under a specific condition, report that condition rather than declaring the process generally weak. Use the next cycle as a retest after one repair, and keep the earlier result visible so the team can see whether the change actually improved reconstruction.

Limits and decision link

The result depends on the package index, the reviewer's familiarity with payroll, and the quality of the available artifacts. It does not prove that the payroll was calculated correctly or that every legal obligation was met. The method is a test of explainability, not a substitute for a control review or an audit. Use the finding to repair one evidence gap, such as requiring every manual adjustment to name an approver and a pay period. Then name the decision that would change if the result is wrong, such as whether a close can be accepted without an independent reconstruction.

Index quality

Grade the index before grading the underlying artifacts. A useful index names the period, the owner, each expected artifact, its location, and its status. It should distinguish missing from not applicable and show who accepted a gap. An index that lists folders rather than documents makes the reviewer search and increases the chance that two people review different evidence. Compare the index with the actual folder to find files that exist but are not listed and entries that point to nothing. A strong index does not guarantee sufficiency, but a weak index makes sufficiency difficult to test consistently.

Provider evidence

When a payroll provider prepares or releases the file, record the provider contact, the file reference, the confirmation, and the time received. A provider portal screenshot may show a submission without showing acceptance, so grade that evidence separately. Include the reconciliation between the employer's approved totals and the provider's processed totals. If the provider made a correction, record the reason and the cycle affected. The study treats provider evidence as its own class because an internal close can look complete while the actual release path is documented only in an email thread.

Exception carryover review

Review the open-exception list at the start of the next close, not only at the end of the prior one. Confirm that each item still has an owner, a due date, and a next action. Record items that were closed without evidence and items that rolled forward more than once. A repeated carryover may mean the owner lacks authority, the evidence is unavailable, or the item was never a real close requirement. Do not let the list grow into a permanent archive. The study should show whether the next cycle inherited a managed list or a forgotten folder.

Reporting the reconstruction result

Report the score for each reconstruction task, the evidence used, and the point where the reviewer stopped. Show disagreement between reviewers and the reasons for it. Include the number of open exceptions carried into the next cycle and whether each had an owner. State the limits of the test and the one repair chosen for the next run. A useful report does not claim that a high score proves payroll accuracy. It shows whether the package can explain the run, which is the precondition for a reviewer to challenge the numbers with evidence.

Artifact naming consistency

Review file names across the package for period, artifact type, and version. A folder that mixes final, latest, and dated copies forces the reviewer to guess which file governed the run. Record naming exceptions and the reason they exist. If a provider supplies files with fixed names, map those names to the internal index rather than renaming the source. Consistent naming is not cosmetic; it determines whether two reviewers open the same evidence. Count the files that could not be matched to an index line and treat that count as a separate finding.

Retest after the index change

After revising the index, ask a new reviewer to reconstruct a different cycle using the same four tasks. Compare the score, the time taken, and the questions raised with the prior result. Keep the artifact definitions stable so the improvement is attributable to the index rather than to a simpler payroll period. If the reviewer still stops at the same point, the index did not address the underlying gap. Record the revision date and the reviewer. Decide whether to keep the change, extend it to another pay group, or repair the artifact itself.

Sources

  1. U.S. GAO: Assessing Data Reliability
  2. NIST Cybersecurity Framework 2.0
  3. IRS Publication 15 (Circular E)
  4. U.S. Department of Labor Wage and Hour Division

FAQs

Does a complete folder mean the close is well documented?

No. The test is whether an independent reviewer can rebuild the run and explain its decisions from the package.

Why use a reviewer who did not run payroll?

The original preparer may fill gaps from memory. An independent reviewer exposes what the record itself can support.

What should be done with open exceptions?

Keep them in a visible list with an owner and due date, then review the list at the next close.

For adjacent operating context, see Payroll Preparation and the payroll operations guide library.

Related research