Reconcile warehouse attendance exceptions before payroll export

The payroll cut-off arrives before the exit record does

An overnight warehouse shift ends in the morning. The attendance system has the entry record but not the exit, and the payroll export cut-off is approaching. A rule marks the employee absent or assigns fewer hours, although the exit event is still waiting to synchronise.

The workflow should expose the evidence gap and route it to the authorised attendance reviewer. It should not turn incomplete device data directly into a wage decision. Payroll rules, employment terms and applicable requirements need their own authorised configuration and review.

A bounded pilot can improve this handoff by connecting the roster, check-in evidence, exception decision and export version. The purpose is to produce a reliable approved input to payroll, not to infer pay from an unreviewed timestamp.

Understand the shift and synchronisation boundary

The Frappe HR — Auto Attendance documentation explains that attendance processing depends on assigned shifts and the last synchronisation of check-in data passing the configured actual shift end, including any allowed check-out window. In the proposed workflow, that synchronisation marker must represent actual completed data coverage. Advancing it to the current clock time without evidence can make incomplete records look complete.

The Shift Type documentation supports overnight shifts and configurable methods for interpreting check-in pairs or working hours. These are software settings to verify in the deployed version. They do not determine the legally or contractually payable hours for every employee.

Record the shift identifier, effective dates, site time zone and applicable attendance-policy version. An employee’s current shift assignment must not overwrite the assignment that applied to the historical night being reviewed.

The data source also needs an identity and coverage record. A device that is online now may still have an earlier unsynchronised interval. Connectivity status and completeness of historical records are different facts.

Illustrative overnight exception

Assume a fictional shift runs from 22:00 on 15 October to 06:00 on 16 October, all in IST. The device records entry at 21:55 and exit at 05:55. The exit event reaches the central system only at 08:20, after an illustrative 08:00 export cut-off.

At 08:00, the system’s verified device coverage extends only to 05:00. The missing exit should therefore remain an attendance exception. The fact that the wall clock reads 08:00 does not prove that all events through shift end have been received.

Once the exit arrives, the raw elapsed time between 21:55 and 05:55 is eight hours. That interval is not automatically eight payable hours. Breaks, approved attendance interpretation and payroll rules must be applied through the appropriate process.

The attendance reviewer examines the late event and relevant evidence, then approves or otherwise resolves the record at 08:40. The resulting payroll input must reflect that authorised decision and the correct export or correction process. It should not silently rewrite an earlier file already consumed by payroll.

Separate source evidence from approved payroll input

A clear workflow preserves several layers rather than repeatedly editing one attendance row.

The reviewer should be able to correct an association or accept authorised supporting evidence without deleting the original device event. The audit trail should explain why the approved attendance differs from a raw first-in/last-out calculation.

Access must be proportionate. A shift supervisor may need to resolve an attendance exception without seeing bank details or unrelated payroll information. The pilot should use the minimum authorised employee data and avoid collecting biometric templates or other unnecessary sensitive material.

Compare native attendance processing with an exception layer

A controlled manual review can be sufficient for a small shift group with few exceptions. It needs a consistent cut-off, an accountable reviewer and a reconciliation to the exported payroll input.

An existing HR system may already support shifts, check-ins and attendance approvals. Test the exact overnight and synchronisation behaviour. A configuration issue should be resolved there before adding another interpretation engine.

An exception layer is useful when device feeds, roster changes and payroll cut-offs are managed in separate systems. Its job is to expose unresolved evidence and preserve the approved handoff. It should not implement a second set of wage rules that competes with payroll.

Use the overnight example to evaluate each option. The expected result at 08:00 is a visible unresolved exception, followed by an authorised resolution when the evidence arrives. Automatic absence or a silently fabricated exit fails the intended test.

Bound the pilot and establish its baseline

Choose one warehouse shift group, one attendance source and one payroll handoff. Include overnight work if that is the problem being addressed. Define the period, reviewers and exclusions before examining results.

Begin with historical replay using authorised records, then supervised processing under the existing attendance and payroll controls. The pilot should not initiate payments or deductions independently.

Establish the baseline from comparable periods: missing-event cases, late uploads, roster mismatches, time to review, exports requiring correction and manual reconciliation effort. Keep pending cases in the denominator so that unresolved records do not disappear from the success rate.

For an illustrative set of ten shift records, eight may be approved and two unresolved at cut-off. The export process must account for all ten records, clearly distinguishing the eight eligible inputs from the two pending decisions. How pending cases are handled for payroll is an authorised payroll-policy decision, not a default invented by the integration.

Test duplicate, delayed and changed records

Replay a duplicated device event with the same source identifier. It should not create another shift or extra attendance time. Then test two genuinely different events with similar timestamps so that deduplication does not discard valid evidence.

Replay a roster correction after the device data arrives. The reviewer should see the old and proposed assignments and determine the correct historical association. A later roster change must not automatically reclassify already approved periods without review.

Test a missing exit that never arrives. The system should retain the exception and support the organisation’s approved evidence-review process. It must not insert a convenient shift-end timestamp and label it as a device observation.

Finally, replay an event received after export. The workflow should identify whether the earlier batch was generated, transmitted, acknowledged or already posted in payroll. The correction path depends on that state.

Make exports reproducible and safe to repeat

Each export needs a batch identifier, version, included approved records and a record of the receiving system’s acknowledgement. Re-sending the same batch should not create a second set of payroll inputs. Verify this behaviour in a test environment with the actual receiving process.

If an unconsumed batch is replaced, preserve the superseded version and its status. If payroll has already posted the earlier input, use the authorised correction process rather than overwriting history. The integration owner and payroll authority must agree how the adjustment is represented.

A lost acknowledgement should trigger reconciliation against the batch reference. Do not assume failure and send a newly numbered duplicate batch containing the same attendance records.

The pilot should also prove that another site’s employee records are inaccessible to an unauthorised reviewer. Test permissions with representative roles rather than using an administrator for every action.

Stop and recover before the next payroll handoff

Pause the pilot if it marks incomplete records as final, associates attendance with the wrong employee or shift, duplicates an export or changes a posted input silently. Return to the approved manual review and payroll handoff.

Reconcile source events, attendance decisions and export acknowledgements for the affected period. Identify which records are unresolved and which corrections have been accepted. A successful data import does not prove that payroll received the right version.

Restart requires the attendance and payroll owners to review the failure and the corrected replay. Confirm the cut-off, synchronisation coverage and open exceptions before the next handoff.

For a focused pilot assessment, prepare a redacted overnight-shift case, a late event and the current export specification. The review should produce a defined exception queue, approval handoff and correction test that both attendance and payroll owners can inspect.

Reliable attendance reconciliation preserves evidence and routes uncertainty to an authorised decision. That gives payroll a traceable input without allowing a missing device event to decide someone’s pay.

Next actions

1. Confirm overnight-shift interpretation and actual source-data coverage.
2. Separate raw events, attendance approval and payroll interpretation.
3. Test late events, duplicates, roster changes and export acknowledgement loss.
4. Review the bounded pilot and correction process with attendance and payroll owners.