EWA for Multi-Shift Manufacturing Companies: How to Implement It to Calculate Worked Hours Correctly?

To implement EWA at a multi-shift manufacturing company, the system needs to distinguish scheduled shifts from actual work; regular hours from overtime; pending data from approved data; and transactions within the period from adjustments made after cutoff. Every record must be linked to an employee ID, work date, shift, plant, approval status, update time, and version. The company should pilot at one plant with relatively stable data, measure on-time attendance approval, limit freshness, successful transactions, and payroll discrepancies before expanding.

> Note: This article is a reference business and technical framework. Salary formulas, components included in the limit, limits, approval processes, and settlement timing must be confirmed by the company, EWA provider, Payroll, legal, and accounting teams based on the actual records.

> Terminology: EWA (accessing wages based on days worked) · payroll (salary calculation) · HRIS (human resources information system) · ERP (enterprise resource planning) · cutoff (period closing point) · pilot (trial implementation) · UAT (user acceptance testing) · KPI (key performance indicator) · workflow (work process) · wave (expansion phase) · dashboard (monitoring board) · go-live (official production launch).

Why is the factory environment a separate EWA problem?

Manufacturing companies often have large workforces, multiple shifts, order-driven overtime, and data passing through multiple levels: time clocks, team leaders, supervisors, HR, payroll, accounting, and banks. Having a scheduled shift does not mean a person completed the shift. A clock-in/out record does not mean it has been recognized. Registered overtime is not necessarily overtime that was completed and approved.

In this environment, the biggest challenge is not displaying an "Access wages" button, but answering accurately:

• whether the worker is currently working and belongs to the program;

• which shift has been completed;

• which worked hours are eligible for calculation;

• which income amount is sufficiently certain;

• which amount must be withheld under policy;

• how previous transactions were transferred and settled.

If any of these questions is answered incorrectly, the available limit may be higher or lower than reality, increasing complaints and end-of-period adjustment workload.

1. Data map from shift work to an EWA transaction

``` mermaid
flowchart TD
A["Shift schedule"] --> B["Clock-in/out"]
B --> C["Exception handling"]
C --> D["Attendance and overtime approval"]
D --> E["Calculate eligible wage portion"]
E --> F["EWA limit"]
F --> G["Transaction and payment"]
G --> H["Payroll, accounting, and reconciliation"]
```

Each step must have a standard data source and an accountable owner. The EWA platform should not interpret a single clock swipe as wages if the attendance process has not yet confirmed it.

------------------------------------------------------------------------------------------
Data domain Proposed authoritative source Business owner
------------------------------ ------------------------------- ---------------------------
Employee profile and status HRIS HR

Shift schedule Shift scheduling system Production/HR

Clock-in -- clock-out Timekeeping machine/app HR Operations

Exceptions and approval Attendance workflow Team leader/supervisor/HR

Period and pay rules Payroll Payroll

Limits and transactions EWA platform EWA Operations

Transfer result Payment partner Payment/Finance

Reconciliation and recording Payroll/ERP Payroll/Accounting
------------------------------------------------------------------------------------------

2. Distinguishing shift schedules, timekeeping data, and approved worked hours

(Core concept: see What are approved worked hours?.)

Shift schedule

The shift schedule indicates when a worker is expected to work. It is used to detect lateness, early departure, absence, shift changes, or overlapping shifts, but it does not by itself prove that the person worked.

Clock-in -- clock-out data

The machine records events. Data may be missing because the worker forgot to clock in/out, the device failed, the network was interrupted, the wrong machine was used, or the worker worked outside their usual location.

Approved worked hours

This is the business result after rules have been applied and exceptions processed. Depending on policy, only this status may be eligible to enter the limit calculation engine.

Why should "provisional worked hours" not be used without clear labeling?

If the company wants to display a limit based on provisional data, it must have risk mitigation, status labels, a holdback level, and a recalculation method when data changes. Workers need to understand why the available amount may change. A number that appears certain should not be displayed when the source is still awaiting approval.

3. Minimum data fields for a multi-shift factory

Employee profile

-----------------------------------------------------------------------------------------------------------
Field Purpose
--------------------------------------- -------------------------------------------------------------------
employee_id Unique identifier that is never reused

employerid / legalentity_id Employing legal entity

plant_id Plant or location

departmentid / lineid Department or production line if used by policy

payroll_group Pay-period and payment-rule group

employment_status Active, temporarily inactive, terminated, or corresponding status

effectivefrom, effectiveto Effective dates

ewa_eligibility Program participation eligibility

sourceupdatedat, record_version Control of stale/current data
-----------------------------------------------------------------------------------------------------------

Shifts and worked hours

----------------------------------------------------------------------------------------------------
Field Purpose
---------------------------------- -----------------------------------------------------------------
work_date Business date used to calculate worked hours

shift_id Shift ID

shiftstart, shiftend Start/end time with time zone

checkin, checkout Timekeeping events

regular_minutes Eligible regular-work minutes

overtime_minutes Identified overtime minutes

leave_code Leave type if relevant

attendance_status Full attendance, incomplete attendance, absent, exception, etc.

approval_status Pending, approved, rejected, adjusted, locked

approvedby, approvedat Approval audit trail

record_version Version after modification
----------------------------------------------------------------------------------------------------

Payroll and transactions

• payperiodid;

• eligible income component code;

• formula version;

• cutoff time;

• payroll-period status;

• transaction_id;

• idempotency_key;

• requested amount, fee, and transferred amount;

• EWA and payment status;

• payment_reference;

• limit before/after transaction;

• data version used for calculation.

4. Which date should a night shift belong to?

A shift that starts before midnight and ends the next day is a common source of discrepancies. The timekeeping system may assign events by calendar date, while payroll may assign the entire shift to the start date or the business work date.

The company needs to define:

1. whether the work_date of a night shift is the start date or end date;

2. how regular hours and night allowances are separated;

3. which date overtime after the shift belongs to;

4. how rest/public holidays crossing a shift are handled;

5. what the standard time zone is;

6. whether cutoff can split one shift across two periods;

7. whether callbacks and late-arriving data trigger recalculation.

Illustrative example

A shift starts at 22:00 on day 10 and ends at 06:00 on day 11. If timekeeping uses day 11 while payroll uses day 10, EWA may undercount or double-count if shiftid and workdate are not standardized.

This should not be solved simply by comparing total monthly hours, because EWA needs to know, at each point in time, which portion of worked hours has become eligible.

5. When should overtime be included in the limit?

Overtime commonly has multiple statuses:

1. planned;

2. registered or accepted by the worker according to the process;

3. actually attended;

4. confirmed by the supervisor;

5. approved by HR/Payroll;

6. payroll period locked.

The company must define which status is eligible for EWA. Overtime can make the limit more attractive but can also fluctuate more than regular work.

Three reference policy options

------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Option Approach Advantage Risk/trade-off
---------------------------------- ------------------------------------------- ---------------------------- ------------------------------------------------------------
Do not include overtime Use only approved regular hours Simple, fewer adjustments Lower limit than expected income

Include only approved OT Use completed and approved overtime Balances value and control Depends on approval speed

Include a portion with a reserve Use provisional data with a holdback rate Earlier limit updates More complex; requires explanation and adjustment handling
------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Every option must be approved by Payroll, HR, legal, and risk management. EWA should not automatically treat every overtime_minutes value as certain wages.

6. Paid leave, unpaid leave, and incomplete attendance

These situations affect the limit differently:

• paid leave may be calculated according to policy after approval;

• unpaid leave does not create the corresponding wage amount;

• compensatory leave may relate to data from another period;

• missing check-in/check-out requires an exception;

• late arrival/early departure may have rounding rules;

• work stoppage or transfer may have separate handling;

• business trips/training may not appear on the timekeeping machine.

The company should establish a status-code table rather than allowing each plant to interpret statuses differently.

--------------------------------------------------------------------------------------------------------
Attendance code Name Included in EWA? Condition Approval owner
----------------- --------------------- ------------------ ------------------------ --------------------
WORK Regular work Per policy Approved Supervisor/HR

OT Overtime Per policy Completed and approved Supervisor/Payroll

AL Paid leave Per policy Approved leave request HR

UL Unpaid leave No Confirmed HR

MISS Missing timekeeping No/hold Awaiting completion Supervisor
--------------------------------------------------------------------------------------------------------

Values in the table must be confirmed by the company; this is only a structural template.

7. Late attendance adjustments and versioned limits

In a factory, worked-hour records may be changed after a worker has already made a transaction. The system needs to know:

• which record changed;

• the before and after values;

• who changed it and who approved it;

• which version was used to calculate the limit;

• which transactions are affected;

• when the difference will be handled;

• whether subsequent transactions need to be temporarily suspended.

Do not delete old records

Create a version or adjustment event instead. If the old record is overwritten directly, the company cannot reconstruct why the limit had that value at the time of the transaction.

Sample traceability fields

``` json
{
"employee_id": "EMP-000123",
"work_date": "2026-08-18",
"shift_id": "NIGHT-A",
"approvalstatus": "ADJUSTEDAPPROVED",
"regular_minutes": 480,
"overtime_minutes": 60,
"record_version": 4,
"sourceupdatedat": "2026-08-20T03:20:15Z"
}
```

This is mock data for illustration, not an official specification of Lương Ngày.

8. On-time attendance approval is a leading KPI

A project can have a good application but still fail because attendance is approved late. Workers who do not see a limit may assume that EWA does not work.

Proposed KPIs

$$
\text{On-time attendance approval rate} = \frac{\text{Records requiring approval that are approved before the deadline}}{\text{Total records requiring approval}} \times 100\%
$$

It should be tracked by:

• plant;

• production area;

• shift;

• team leader/supervisor;

• exception type;

• day within the period;

• time from shift end to approval.

The goal is not to turn the KPI into a tool for penalizing supervisors. The dashboard should identify causes: device failures, incorrect employee lists, too many exceptions, missing approval permissions, or an unsuitable process.

9. How should the limit be calculated so it does not fluctuate unexpectedly?

A conceptual formula can be represented as:

$$
\text{Available limit} = \text{Approved eligible income portion} \times \text{Allowed rate} - \text{Holdback amount} - \text{Amount already received in the period}
$$

The components must be defined:

• which income is eligible;

• which attendance status qualifies;

• who determines the allowed rate and for which group;

• what the holdback is intended for;

• whether pending transactions reserve the limit;

• how attendance adjustments change the limit;

• when the limit is locked for the payroll period.

Detailed formulas or risk thresholds should not be disclosed until approved. Workers need enough explanation to understand the displayed amount, but they do not necessarily need to know internal anti-fraud logic.

10. Integration with timekeeping and payroll systems

(Data and architecture requirements: see Integrating EWA with timekeeping, payroll, and ERP.)

Near-real-time API

Suitable when the factory has modern systems and needs rapid updates after approval. Authentication, versioning, retries, out-of-order data, and error monitoring must be controlled.

Batch file/SFTP

Suitable for legacy systems or processes that close on a schedule. Files need a batch ID, total record count, checksum, version, naming rules, duplicate prevention, and row-level error reporting.

Controlled manual synchronization

Can be used for a small pilot. It requires a standard template, preparer/approver, logs, total checks, a secure file storage area, and a plan to eliminate manual steps when scaling.

Do not connect every source system directly

A multi-plant company may have many timekeeping machines or software systems. A standardized integration layer allows EWA to receive the same data model instead of writing separate logic for each device.

11. Four-way reconciliation in a manufacturing environment

(Details: see Reconciling EWA transactions with payroll and accounting.)

Reconciliation should link:

1. attendance and limits;

2. EWA transactions;

3. payment results;

4. payroll/ERP.

Common discrepancies to identify

• attendance was adjusted but the limit was not updated;

• transaction succeeded but payroll is missing it;

• payment succeeded but EWA still shows processing;

• transaction was entered into the wrong period;

• one transaction appears twice;

• a terminated worker still has a transaction;

• a refund did not restore the limit according to the process;

• employee ID is correct but the legal entity/plant is wrong;

• total amount matches but individual transactions contain over/under amounts requiring offsetting.

Each discrepancy needs a case, owner, internal deadline, evidence, and approver for closure.

12. Organizing worker support at the factory

Shift workers may encounter problems outside office hours. The support channel needs to match actual usage times.

Four support levels

--------------------------------------------------------------------------------------------------
Level Issue Contact point
-------------------- --------------------------------------------- -------------------------------
Level 0 Guidance, FAQ, self-service status App/documents

Level 1 Activation, usage, attendance not displayed HR/factory contact

Level 2 Transactions, payments, integration data EWA Operations/IT/Payment

Level 3 Critical incidents, fraud, payroll Risk/Security/Finance/Payroll
--------------------------------------------------------------------------------------------------

Tickets should include

• controlled employee ID;

• plant and shift;

• issue type;

• transaction ID if applicable;

• timestamp;

• attendance/limit status;

• action already taken;

• next contact point;

• deadline and outcome.

Support staff must not ask workers to provide passwords or OTPs.

13. Communication on the shop floor should be simple but complete

Messages should explain:

• what EWA is;

• which portion of money can be accessed;

• why the limit changes;

• fees, if any;

• how transactions are settled;

• what to do when attendance has not been approved;

• what to do when changing the phone number/receiving account;

• how to report an unusual transaction;

• EWA does not replace checking the payslip.

Communication channels

• onboarding;

• pre-shift meetings;

• posters with QR codes;

• short videos;

• app/SMS;

• team leaders or HR at the factory;

• bilingual materials when the workforce needs them.

Do not train only team leaders and assume every worker understands. Measure reach, activation, and repeated questions.

14. Security and privacy at the production site

(Full framework: see Data security and privacy when implementing EWA.)

Common risks include shared phones, SIM changes, counter support, screens exposing another person's data, and Excel files transferred through inappropriate channels.

Controls should include:

• verification during activation and sensitive transactions;

• no shared accounts;

• masking account numbers and amounts when displayed in public areas;

• no photographing/sending payslips through group chats;

• access control by plant and task;

• support-action logs;

• device/phone-number change procedures;

• simulated or masked test data;

• retention and deletion periods for intermediary files;

• a channel for reporting lost accounts.

The Personal Data Protection Law No. 91/2025/QH15 and Decree 356/2025/ND-CP took effect on January 1, 2026. Companies should review roles, purposes, processing scope, and worker rights against the actual architecture.

15. How should an EWA pilot at one factory be designed?

(Standard roadmap: see 90-day EWA pilot plan for businesses.)

Choose the scope

Choose one production area or shift group with:

• confirmed demand;

• relatively good data;

• supervisors ready to approve attendance;

• a representative payroll process;

• sufficient on-site support;

• no excessive differences from the next expansion site.

Cover at least the critical life cycles

The pilot must validate:

• activation;

• regular work and overtime;

• night shifts;

• attendance adjustments;

• successful/failed/unknown transactions;

• daily reconciliation;

• one complete payroll period;

• complaints and exception handling.

Pilot KPIs

• percentage of eligible workers with valid data;

• on-time attendance approval rate;

• time from attendance approval to limit update;

• activation rate;

• successful transaction rate;

• time to receive money;

• tickets per 1,000 transactions;

• automatic reconciliation rate;

• discrepancy rate by cause;

• unknown transactions;

• false-block rate;

• operating cost per user/transaction.

Numeric targets should be based on the baseline and factory capability, not copied from another project.

16. UAT checklist for production shifts

Shifts and attendance

• [ ] Full attendance on day shift.

• [ ] Night shift spanning two dates.

• [ ] Shift change before and after cutoff.

• [ ] Missing check-in or missing check-out.

• [ ] Late arrival, early departure, and rounding rules.

• [ ] Paid leave and unpaid leave.

• [ ] Business trip/training without timekeeping-machine records.

Overtime

• [ ] OT planned but not performed.

• [ ] OT performed but awaiting approval.

• [ ] OT approved.

• [ ] OT adjusted after approval.

• [ ] OT crossing a rest/public holiday according to company procedures.

Employees

• [ ] New employee before effective date.

• [ ] Temporarily inactive employee.

• [ ] Employee terminated mid-period.

• [ ] Employee transferred between plants/legal entities.

• [ ] Duplicate or incorrectly mapped employee ID.

Transactions

• [ ] Request within limit.

• [ ] Request exceeding limit.

• [ ] Repeated submission with the same idempotency key.

• [ ] Timeout and unknown result.

• [ ] Payment failure.

• [ ] Refund transaction.

Payroll and reconciliation

• [ ] Transaction posted to the correct period.

• [ ] Duplicate import file is blocked.

• [ ] Late attendance adjustment creates a traceable adjustment.

• [ ] EWA -- payment -- payroll -- ERP match.

• [ ] Discrepancy creates a case and is approved for closure.

17. Expanding from one factory to multiple factories

Do not copy the configuration unchanged if factories differ in shifts, equipment, payroll, or legal entities.

Group by wave

Group factories with:

• the same timekeeping system;

• the same shift-code set;

• the same payroll policy;

• the same legal entity;

• the same support capability;

• the same level of attendance-approval readiness.

Gate before each wave

• employee and shift mappings have been checked;

• work/overtime rules have been approved;

• UAT has been completed for the factory;

• supervisors have been trained;

• dashboard and reconciliation cover the new scope;

• access rights are opened for the correct units;

• previous-wave issues have been resolved;

• rollback is ready.

Track KPIs by factory to detect performance degradation during expansion.

18. Common mistakes

Using the shift schedule as actual worked hours

Create a limit before there is evidence of work.

Treating every clock swipe as valid attendance

Ignore missing punches, overlapping shifts, wrong people, and exceptions.

Including all unapproved overtime

Make the limit fluctuate and increase end-of-period adjustments.

Failing to standardize the date of night shifts

Cause missing/duplicate worked hours and incorrect periods.

Measuring only transactions, not attendance approval

Fail to see the bottleneck at the supervisor level.

Making HR handle every ticket

HR cannot independently resolve payment, API, reconciliation, or fraud issues.

Expanding while the pilot is manually supported

The results look good but do not reflect operational scalability.

Overwriting late attendance adjustments

Lose the ability to reproduce the limit at the time of the transaction.

Conclusion

EWA has particular potential in multi-shift manufacturing companies, but its value appears only when worked hours are confirmed correctly and on time. The platform needs to distinguish shift schedules -- timekeeping -- approved attendance, handle night shifts and overtime clearly, manage data versions, and reconcile every transaction.

Companies should start at one plant with relatively stable data, treat on-time attendance approval as a leading KPI, and expand in waves only after completing one full payroll period. Learn more about Lương Ngày for businesses to discuss a timekeeping-data assessment and the scope of a Lương Ngày factory pilot.

References

• Labor Code No. 45/2019/QH14

• Personal Data Protection Law No. 91/2025/QH15

• Decree 356/2025/ND-CP guiding the Personal Data Protection Law

------------------------------------------------------------------------

Author: Nguyễn Tấn Lộc --- Strategy Department Specialist, Nhân Kiệt Human Resources Supply Co., Ltd.

Consulting on Lương Ngày solutions for businesses: Hotline 0937.022.655 · Email info@nhankiet.vn · Lương Ngày for businesses