Quick answer
A payslip batch is complete only when every expected approved record has an explained outcome. Reconcile the expected population against released records, held records and confirmed exclusions using employee IDs and the intended record structure. Check duplicates separately. Do not add “viewed” records to “sent” records as if they were different employees, and do not call an upload count proof that every worker received a payslip.
Define the unit you are counting
Before comparing totals, decide whether the batch expects one record per employee or another approved structure. The examples here assume one employee-level record for one salary period. If your payroll legitimately produces multiple records for an employee, use a composite key such as employee ID plus period and approved record reference.
A staffing agency’s client roster counts assignments, not necessarily unique payslip recipients. A worker moving from site A to site B may appear twice in assignment data but only once in the approved employee-level batch. Conversely, a departed worker may still be expected to receive a final record.
Use mutually exclusive reconciliation categories
| Category | Definition for this checklist | Include in completion arithmetic? |
|---|---|---|
| Released | Correct intended record released once | Yes |
| Held | Expected record awaiting a specific resolution | Yes, but still unresolved |
| Approved exclusion | Documented decision that this batch does not issue the record | Yes, with authority and reason |
| Unexplained missing | Expected record has no outcome | Yes, as an open exception |
| Viewed / acknowledged | Later events about a released record | No; do not add again |
| Duplicate | Extra record beyond the intended structure | Investigate separately; not extra completion |
“Released” in this table means your documented send action, not a guarantee of mailbox delivery. Keep downstream notification and access issues in a separate follow-up layer.
Simulated 100-record reconciliation
Assume an approved expected population of 100 employee-period records. A pre-release review finds 96 ready records, three held for verified contact updates and one held for a payroll decision. These are 100 accounted-for records, but the batch is not fully released.
After the first release, the log shows 96 released and four held. Forty viewed records are a subset of those released. The calculation is 96 + 4 = 100, not 96 + 4 + 40 = 140.
If a report shows 97 release rows but only 96 unique intended record keys, investigate the duplicate. Do not use the row count to hide a missing employee. If 95 expected keys appear as released and four as held, one remains unexplained regardless of how many times another record was sent.
This example demonstrates arithmetic and category design. It is not a product dashboard screenshot or a result from a Philippine customer.
A repeatable batch-check sequence
- Freeze the expected population from approved payroll and record the salary period.
- Define the unique record key before combining client-site lists.
- Match prepared records to the expected keys. Inspect missing and extra keys.
- Resolve duplicate keys according to the approved record structure, not by blindly deleting the later row.
- Verify contacts and separate held records with an owner and reason.
- Preview the employee-facing records and approve release.
- Reconcile released, held, approved exclusions and unexplained missing records after the action.
- Track delivery and access follow-up separately so later events do not distort the population totals.
Ant Payslip supports preparing and distributing individual records from approved Excel or CSV data. The reconciliation structure above is an agency control; do not assume the product automatically performs every join, category calculation or duplicate decision.
Keep the reconciliation log proportionate
Employee ID, record reference, period, outcome, owner and next action are usually more useful than copying full salary details into a shared completion report. The NPC rules address proportionate processing and appropriate personal-data safeguards. NPC: Data Privacy Act implementing rules.
Give client-site coordinators only the information needed for their authorised role. A reconciliation request does not automatically authorise sending the agency’s complete payroll to every client.
If an exception survives the cutoff, keep it visible with a follow-up date. Closing the main batch should not remove unresolved workers from the HR queue. When releasing a held record later, link the action to the original expected population and confirm that no previous record already satisfies that entry.
Reconciliation is not payroll evidence
The employer still needs the period, rate, regular/overtime amounts, deductions and actual payment in its relevant payroll records. Omnibus Rules, Book III, Rule X, section 6. Batch counts and online views do not replace required evidence. Never exclude a worker merely to remove an obligation; missing email requires a verified alternative, not wage withholding.
Frequently asked questions
Can I reconcile using employee names?
Use stable IDs and the intended record key. Names help human review but may be duplicated or spelled differently across client lists.
Should viewed records be added to sent records?
No. Viewing is a later event on a record already included in the release count.
Is a held record an approved exclusion?
No. A hold remains an open action. An exclusion needs a documented decision and should not be invented to make totals match.
Does a matching total prove correctness?
No. Compare unique keys and outcomes, not only totals. One missing record and one duplicate can cancel numerically while both remain wrong.
Related guidance
See preparing Excel for individual delivery and following up on missing emails.
Review your agency’s batch controls
Request an Ant Payslip demo using an anonymised sample with transfers, duplicate keys and held contacts to evaluate the release workflow alongside your reconciliation log.
Source checked: 8 October 2026. Counts are simulated, and operational completeness is not proof of receipt or legal acceptance.