Outsource Payroll CompanyPowered by StealthAgents
Table of contents
  1. Research finding
  2. Methodology
  3. Key takeaways
  4. Question and scope
  5. Request intake
  6. Identity evidence
  7. Authorization chronology
  8. Bank validation records
  9. Payroll-cycle linkage
  10. Replay and exception review
  11. Limits and decision link
  12. Request channels and timing
  13. Approval and substitution evidence
  14. Data quality checks
  15. Reporting the coverage result
  16. Sampling the request population
  17. Retest after the repair
  18. Sources
  19. FAQs
  20. Related research

Research brief · Published

Direct deposit change verification coverage

A bounded study design for measuring how many direct deposit changes have evidence for the request, identity check, effective cycle, and bank-file match.

Research finding

Measure the change request, the identity check, the approval, and the file match as separate gates.

5Headline 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
5 fixed evidence gates applied to each direct deposit change.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

  • Measure the change request, the identity check, the approval, and the file match as separate gates.
  • A request received is not the same as a change authorized.
  • Deferred changes and rejected changes need their own disposition categories.
  • Report coverage by pay group and request channel so one easy segment cannot hide another.

Question and scope

This study asks what share of direct deposit changes has evidence that the request, identity check, approval, and effective pay cycle were verified before the bank file was released. The population is every direct deposit change requested for one employer's pay groups during three consecutive cycles. The unit is one change request with a status of applied, deferred, rejected, or withdrawn. The question separates intake from authorization and release, because a form on file does not show that anyone confirmed the employee's identity or that the payroll file matched the approved account. The result describes evidence coverage for this process, not whether any individual bank account is valid.

Request intake

Start with the request as its own record. Capture the date received, the channel, the employee identifier, the old account ending, the new account details, the requested effective date, and the person who accepted the request. A message that merely says to update direct deposit is incomplete evidence because it does not state what should change or when. Grade the request as complete only when the required fields are present and the employee can be linked to the payroll population. Record missing fields as gaps rather than filling them from memory or from a later conversation that was not documented.

Identity evidence

The identity check is a separate gate. Record the channel used, the date, the person contacted, and the result. A response to the same email thread that requested the change is weak evidence because it does not establish an independent contact path. A callback to the number on file or a confirmation through an approved employee portal is stronger. Grade the identity gate as documented only when the record names the channel and the reviewer. If the employee could not be reached, record that outcome and treat the change as unverified rather than assigning a partial pass that hides the missing confirmation.

Authorization chronology

A change can be complete and still lack a valid approval sequence. Record when the request arrived, when identity was confirmed, when the payroll owner approved the change, and when the record was updated. Compare those timestamps with the payroll calendar and the published freeze point. A change approved after the freeze should appear as deferred to the next cycle unless a documented exception applies. The chronology makes it possible to distinguish a late request from a late approval, which are different process failures even when both produce the same delayed payment.

Bank validation records

Record what was checked in the bank file, not only what was entered in the payroll system. Capture the employee identifier, the account ending, the payment amount, the effective date, and the reviewer who compared the change log with the file. If the provider supports a prenote or test deposit, record the result and the date. A successful test is not proof that the live file is correct, but it is useful evidence about the account path. A failed or missing test should remain visible in the coverage result rather than being treated as an administrative detail outside the study.

Payroll-cycle linkage

Every approved change needs a link to the first payroll cycle that used it. Without that link, the team cannot tell whether the employee was paid to the old account, the new account, or a mixture across two runs. Record the pay period, the pay date, the file reference, and the release confirmation. If the change was deferred, record the cycle in which it actually applied and the reason for the delay. Count each change only once in the coverage result, using the cycle that first carried the new account, so a repeated request cannot inflate the apparent completion rate.

Replay and exception review

Give a second reviewer a de-identified subset of changes and the gate definitions without the first reviewer's notes. Ask the reviewer to grade each gate and cite the evidence that supports the grade. Compare the two sets and resolve disagreements by inspecting the underlying record. Track the reasons for disagreement, because repeated confusion about one gate usually means the definition is unclear. Treat deferred and rejected requests as part of the review, not as failures to be removed from the denominator. A complete study explains what happened to every request in the population.

Limits and decision link

Coverage depends on the quality of the request record, the identity process, and the provider's file data. It does not establish that a bank account belongs to the employee, that a payment was legally correct, or that the employer followed every applicable rule. Repeat the study across pay groups and channels before treating one result as a pattern. Use the finding to repair one gate, such as requiring a documented second-channel check before approval. Then name the decision that would change if the result is wrong, such as whether a late request should use the current cycle or the next one.

Request channels and timing

Record the channel through which each change arrived and the time it was received. A portal submission, a signed form, a phone call, and an email carry different evidence properties, and one blended coverage number can hide a weak channel. Compare requests that arrived before the payroll freeze with those that arrived after it. A late request is not automatically a failure if the team deferred it correctly and told the employee when the change would apply. Report timing separately from evidence completeness so a prompt but unverified request does not appear stronger than a late but well-documented one. If one channel consistently lacks required fields, repair the form instead of adding another reminder after submission.

Approval and substitution evidence

Record who approved each change and whether that person had current authority for the pay group. A named approver is not enough if the approval came from someone covering an absence without a documented delegation. Capture the approval date, the scope, and any condition attached to it. If the primary owner was unavailable, record the substitution and the evidence that the backup could act. Compare approvals by approver and pay group to find patterns, such as one group relying on a single person for every change. A substitution without a stated scope is not evidence of authority, even when the person had technical access to the payroll system.

Data quality checks

Before scoring coverage, run a small set of checks on the extract. Confirm that every request has a unique identifier, that each employee link resolves to one payroll record, and that the same change is not counted twice because it appeared in two channels. Check for impossible timestamps, missing pay periods, and account endings that do not match the request. Record each check and its result rather than silently cleaning the data. A high coverage score on a duplicate-heavy extract is misleading, and the report should show the cleaning decisions so another reviewer can reproduce the denominator. Keep the raw extract separate from the working copy.

Reporting the coverage result

Report coverage by gate, by channel, by pay group, and by disposition. Show the denominator and the number excluded, with the reason for each exclusion. Do not publish only a single percentage, because it hides whether the weak point is intake, identity, approval, or file matching. Include a short note on what the study does not measure and which decision the result supports. A useful report ends with one repair, an owner, and a retest date. The next cycle should be able to show whether the repair changed the gate that prompted it, rather than merely producing a new number.

Sampling the request population

Use a fixed sampling rule when the population is too large to review in full. Record the period, the number of requests, the selection method, and the number reviewed. If the team samples only requests that reached the bank file, the study will miss deferred and withdrawn requests that may carry the weakest evidence. Include those dispositions in the frame so the coverage result reflects the full intake rather than only the successful changes. A convenience sample chosen after the exceptions are known cannot measure coverage, and the report should label it as illustrative rather than representative. Keep the selection list so a second reviewer can reproduce the same denominator.

Retest after the repair

Retest the same gates after the repair with a new, comparable period. Keep the original definitions and the population rule so the comparison is meaningful. Report the before and after result by gate, not just an overall change. If the repair improved identity evidence but left approval unchanged, say so instead of reporting a single blended improvement. A retest that changes several definitions at once cannot show which change helped. Record the date, the sample, and the reviewer, then decide whether to keep, adjust, or abandon the repair before starting another one.

Sources

  1. IRS Publication 15 (Circular E)
  2. SSA Employer Information
  3. NIST Cybersecurity Framework 2.0
  4. Consumer Financial Protection Bureau: Bank accounts

FAQs

Does this study prove a bank account is valid?

No. It measures whether the process produced evidence for the request, identity check, approval, and file match. Account validity is a separate question.

Should deferred changes be excluded?

No. Keep them in the population with a disposition so the result shows what happened to every request.

Who decides whether a late change is approved?

The authorized payroll owner or a named delegate decides under the exception rule. Support staff prepare and flag the evidence.

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

Related research