Attendance Policy & Regularisation: An HR Setup Guide

An attendance policy is the quietest document in your HR handbook and the one that touches the most money. Every paid day, every loss of pay deduction, every overtime rupee, every leave balance and every statutory register traces back to a single question: was this person at work, and for how long? Get the attendance policy and the regularisation workflow right, and payroll becomes a calculation. Get them wrong, and payroll becomes a negotiation — every month, with a different set of exceptions.

This guide is a practical setup manual for HR managers, operations leads and payroll teams in India. It covers how to write an attendance policy that people actually follow, how to design a regularisation workflow that closes on time, how to configure the whole thing in an HRMS, and how to keep the records an inspection will ask for.

Why attendance policy design deserves more attention than it gets

Most companies inherit their attendance policy. Somebody copied a template in year one, the business changed shape three times since, and the policy now describes an organisation that no longer exists. The symptoms are recognisable:

• Payroll cut-off slips every month because approvals are pending
• The same twenty employees appear in the exception list every cycle
• Managers approve regularisations without looking, because the alternative is a payroll escalation
• Nobody can explain, from the data alone, why a specific employee lost half a day's pay
• Field and shift employees are governed by a policy written for a desk job

Each symptom has the same root cause: the policy describes intent, not mechanics. A good attendance policy is mechanical. It states what counts as present, what counts as a shortfall, what happens automatically, what requires approval, who approves it, by when, and what happens if nobody acts.

The other reason to take it seriously is statutory. Attendance data is the source for the attendance register, the overtime register, weekly off compliance, working hours limits, leave accrual and loss of pay. Under India's labour codes, hours, rest days and overtime carry sharper consequences than they used to, and the records must be producible. A policy that leaves attendance as a monthly negotiation cannot produce clean registers.

Verify before you apply: working hours limits, overtime rates, spread-over rules and weekly off requirements are set by the applicable central and state rules and vary by state, establishment type and employee category. Confirm the current position for your locations with the relevant notification or your labour advisor before configuring your system.

The building blocks of an attendance policy

Before writing anything, settle nine design decisions. Everything else follows from them.

1. Who is covered, and in which category

Not everyone should be governed identically. Define employee categories explicitly and state which attendance rules apply to each:

Write the category into the employee master, not into a manager's memory. Every downstream rule keys off it.

2. What defines a working day

Choose and state one of the common models:

• Fixed timing: in by a stated time, out after a stated time, with a defined grace period.
• Core hours: mandatory presence during a window (say, 11:00 to 16:00), with flexibility around it and a daily hours minimum.
• Total hours: a daily or weekly hours target with no fixed window.
• Roster-driven: the published shift defines the expected timing, and the roster is the source of truth.
• Output-based: attendance is marked present on any working day the employee is not on approved leave; used for senior and client-facing roles where hours tracking adds no value.

State the minimum hours that qualify for a full day and for a half day. Ambiguity here is the single most common source of payroll disputes.

3. Grace, late marks and their consequences

Decide the grace period (commonly ten to fifteen minutes), how many late marks are tolerated in a month, and what happens beyond that. Make the consequence proportionate and, importantly, make it automatic and visible: a late mark that only surfaces in the payslip is experienced as a trap.

A workable structure:

• Grace of 15 minutes, no consequence
• Up to 3 late arrivals in a calendar month, no consequence
• The 4th and subsequent late arrivals convert to half-day short leave, deducted from leave balance if available and from pay if not
• Repeated patterns over three consecutive months are handled as a performance conversation, not an escalating deduction

4. Missed punches and their default treatment

Every system has missed punches. What matters is the default. Two philosophies:

• Default absent, regularise to present. Rigorous, produces clean data, requires an efficient workflow. Best where attendance drives pay directly.
• Default present, flag for review. Friendlier, faster payroll, weaker data. Acceptable in low-variance office environments where LOP is rare.

Pick one, state it, and apply it consistently. Mixed defaults across teams destroy the integrity of the register.

5. Half-day and short-leave rules

Define the hours thresholds for a half day, whether half days are counted in first or second half, and how short leave (a permitted absence of one to two hours) is requested and limited. Short leave is a useful pressure valve — it stops people burning a full leave day for a bank appointment — but it needs a monthly cap.

6. Weekly off, holidays and week-off working

State the weekly off pattern for each category, how weekly offs rotate for shift teams, the annual holiday list per location, and what happens when someone works on a weekly off or a declared holiday. The last one is important: the treatment is usually either compensatory off within a defined window or premium pay, and the applicable statutory position depends on your establishment type and state. Choose, document, and configure it — do not leave it to case-by-case discretion.

7. Overtime eligibility and rate

Define which categories are eligible, the minimum block that counts (for example, thirty minutes), whether overtime requires pre-approval, the daily and weekly limits, and the rate. Pre-approval is worth the friction: unapproved overtime is both a cost control problem and an hours-limit compliance problem.

8. Remote, hybrid and travel days

If you have a hybrid model, state how a work-from-home day is recorded, whether there is an in-office minimum, how it is measured (per week, per month, per quarter), and what happens when the minimum is missed. Define travel days too — a full day of client travel should not require the employee to prove a punch.

9. Absence without information

Define the escalation: same-day manager contact, written communication after a stated number of days, and the point at which the case moves to the absconding process. Cross-reference your absconding policy rather than duplicating it, and never treat unreported absence as an automatic resignation without following a documented process.

Writing the policy: structure and sample clauses

Keep the document under six pages. Structure it as follows.

1. Purpose and scope. One paragraph. Who it applies to, which locations, effective date, and what it supersedes.

2. Definitions. Working day, half day, grace period, late mark, short leave, weekly off, regularisation, cut-off date, roster. Define terms once and use them consistently — the majority of policy confusion is vocabulary drift.

3. Attendance capture. The approved capture mechanisms per category, and a statement that the system record is the primary record.

4. Working hours and shifts. The models by category, roster publication timelines, and change notice.

5. Late arrival, early departure and short leave. Thresholds and consequences.

6. Regularisation. The workflow, timelines and approval matrix. (Covered in detail below.)

7. Overtime. Eligibility, approval, limits, rate.

8. Weekly off and holiday working. Treatment and compensatory off window.

9. Absence and unreported absence. Notification duties, escalation.

10. Payroll linkage. How attendance translates to paid days, LOP and overtime; the cut-off date; and how post-cut-off corrections are handled.

11. Data and privacy. What attendance data is captured, especially location data for field teams, why it is captured, how long it is retained and who can see it.

12. Exceptions and review. Who can approve a deviation, and when the policy is reviewed.

A sample clause on the payroll linkage, which is the clause most often missing:

Attendance for a payroll month is locked at 18:00 on the 25th of the month. Paid days for that month are computed from the locked record. Regularisation requests and approvals received after the lock will be processed in the following payroll month as an adjustment. No manual override of paid days will be made after the lock except with the written approval of the HR head, and every such override will be logged with the reason.

That single clause eliminates the most common payroll delay in Indian SMBs.

Designing the regularisation workflow

Regularisation is the process by which an incomplete or incorrect attendance record is corrected — a missed punch, a client visit, an approved late arrival, a system failure. It is where most attendance systems succeed or fail operationally.

Principles

Employee raises, manager approves, HR audits. Do not let HR become the default approver; managers know whether the person was working. HR's role is to audit patterns, not to adjudicate individual days.

Reason codes, not free text. A structured set of reasons produces analysable data. Free text produces "as discussed."

Suggested reason codes: missed punch (device unavailable), missed punch (forgot), client or site visit, official travel, work from home approved, biometric failure, power or network outage, medical appointment with prior approval, approved late arrival, system error.

Hard windows. A request must be raised within a set number of days of the date, and approved before the monthly lock. Requests raised late are not impossible, but they route to a higher approver and process in the next cycle.

Caps. Limit regularisations per employee per month (four is a common figure). Beyond the cap, the request needs skip-level approval. Caps convert a data-quality problem into a visible management conversation.

Auto-escalation, never auto-approval. If a manager does not act within the window, escalate to the skip-level manager. Never auto-approve on timeout — that guts the control entirely. Auto-rejection with the option to re-raise is acceptable in strict environments.

The workflow, step by step

1. Detection. System flags an incomplete day the morning after — missing punch, hours below threshold, absence on a rostered day. The employee gets a notification. Early detection is what makes the rest work; a flag on the 24th about the 3rd is useless.
2. Request. Employee submits a regularisation with date, reason code, corrected in/out times where relevant, and a comment. Attachment optional, mandatory for selected reason codes.
3. Validation. System checks the request against policy: is it within the window, within the monthly cap, consistent with an approved leave or holiday, and does it conflict with another request?
4. Approval. Reporting manager approves or rejects with a comment. Rejection requires a comment.
5. Escalation. No action within the stated period escalates to the next level with a reminder.
6. Posting. Approved regularisation updates the attendance record and recalculates paid days, LOP and overtime.
7. Audit. HR reviews an exception report: employees exceeding caps, managers with 100% approval rates, reason codes trending upward, and repeated same-reason patterns.

Approval matrix

That last row on the manager's own attendance matters more than it looks. Self-approval, even where technically prevented, tends to reappear as informal arrangements unless the matrix names an approver explicitly.

Capture mechanisms and when to use each

Two rules govern all of them. First, retain raw logs, not just monthly summaries — the raw log is what reconstructs a disputed day. Second, whatever the mechanism, the employee must be able to see their own record daily. Attendance disputes are almost always visibility failures: nobody argues about a record they have been looking at all month.

A note on biometric and location data

Attendance data of this kind is personal data, and biometric and precise location data are more sensitive than most. Practical obligations to build in:

• Collect the minimum necessary. If a geofence check-in answers the question, continuous background tracking does not need to happen.
• Tell employees plainly what is captured, why, how long it is kept, and who can see it. Put this in the policy, not buried in a system consent screen.
• Restrict access by role. A reporting manager needs their team's attendance; they do not need location history for other departments.
• Set a retention period and enforce it.
• Where a vendor processes the data, have the contractual terms and security commitments in place.

Configuring attendance in an HRMS

Configuration is where policy either becomes real or becomes decorative. Work through this sequence.

Step 1 — Set up the calendar. Location-wise holiday lists, weekly off patterns, and the financial year. Multi-state employers need separate holiday lists per location; a single national list is almost always wrong.

Step 2 — Define shifts. Each shift needs a start, end, break duration, grace, half-day threshold, full-day threshold, and whether it crosses midnight. Night shifts crossing midnight are the classic configuration bug — verify that a punch at 23:50 and one at 08:10 the next morning resolve to a single shift.

Step 3 — Build the attendance policy objects. Most systems let you create named policies with the rules from the design decisions above, then assign them to categories or departments. Create as few as you can justify; every extra policy is future maintenance.

Step 4 — Configure regularisation. Reason codes, caps, windows, approval matrix, escalation timers, mandatory attachments.

Step 5 — Configure the payroll bridge. Map attendance outcomes to payroll inputs: present days, paid leave days, LOP days, half days, overtime hours, weekly off worked, holiday worked. This mapping is the single highest-risk configuration in the system. Test it against hand calculations before go-live.

Step 6 — Set the lock. Cut-off date and time, who can unlock, and whether unlock is logged.

Step 7 — Build the reports. At minimum: daily attendance summary, monthly muster, exception report, overtime report, regularisation audit, late-mark report, and the statutory attendance register format.

Step 8 — Parallel run. Run one full month in parallel with your existing method. Compare paid days employee by employee. Investigate every difference. Do not skip this — go-live differences discovered in a payslip cost far more to fix than differences discovered in a spreadsheet.

Test cases to run before go-live

Build these as a checklist and verify each one:

• Employee punches in and out normally on a regular day
• Employee punches in only; regularisation raised and approved
• Employee punches in only; no regularisation raised by cut-off
• Night shift crossing midnight
• Employee works on a weekly off, then claims compensatory off
• Employee works on a declared holiday
• Half day in the first half, and half day in the second half
• Approved leave on a day with a stray punch
• Leave applied for a period spanning a holiday and a weekly off
• Employee joins mid-month; employee exits mid-month
• Overtime approved and unapproved on the same day
• Late arrival crossing the monthly late-mark threshold
• Regularisation raised after the cut-off
• Employee transferred between locations mid-month with different holiday lists

The attendance-to-payroll bridge

The bridge is deceptively simple to describe and easy to get wrong.

Paid days = calendar days in the month, minus unpaid absences, adjusted for mid-month joins and exits.

Decide and document your convention on two points that cause most disputes. First, whether weekly offs and holidays adjacent to unpaid absence are paid — many organisations follow a "sandwich" rule where a weekly off falling between two unpaid days is also unpaid, and this must be stated explicitly because employees rarely anticipate it. Second, whether your monthly divisor is calendar days, fixed 30, or actual working days. All three are used in practice; consistency and disclosure matter more than the choice.

Loss of pay = per-day rate × unpaid days, where the per-day rate follows your stated divisor.

Overtime = approved overtime hours × the applicable rate, subject to daily and weekly limits.

Build a monthly reconciliation as a standing control:

Five checks, ten minutes, and they catch the overwhelming majority of payroll errors before employees do.

Rolling out a new attendance policy without a revolt

Attendance changes are read as trust signals. A tighter policy introduced without explanation will be interpreted as surveillance, regardless of intent. Sequence the rollout deliberately.

Weeks 1–2: design and consult. Draft the policy. Show it to a small group representing each category — an operations supervisor, a field sales rep, an office manager. They will find the cases you missed within twenty minutes.

Week 3: communicate the why. Explain the reasons that are actually true: accurate pay, fair overtime, statutory registers, fewer month-end disputes. Do not claim it is "for employee convenience" if it is not; people can tell.

Weeks 4–7: soft launch. Run the system with no financial consequence. Employees see their record, raise regularisations, learn the workflow. Publish weekly adoption numbers by team. Managers whose teams are not regularising need a nudge now, not in month three.

Week 8: go live. Financial consequences begin from a stated date, communicated at least two weeks in advance.

Weeks 9–12: stabilise. Weekly exception review. Fix the policy where reality proves it wrong — a grace period that is too tight for a team dependent on public transport, for example, is a policy problem, not a discipline problem.

Month 4 onwards: audit rhythm. Monthly exception report to function heads. Quarterly policy check. Annual review.

Manager enablement

Managers are the workflow. Give them a one-page brief covering: how to approve, what the SLA is, what happens on escalation, which patterns to raise with HR, and — critically — the difference between an attendance problem and a performance problem. A manager who uses attendance deductions as a discipline tool will create disputes; a manager who never questions a pattern will create cost.

Attendance for specific workforce types

A single policy rarely survives contact with a mixed workforce. Here is how the core rules bend for each common population, without breaking the register.

Retail and multi-store teams

Store teams run on rosters built around footfall, not on a fixed nine-to-six. The practical requirements are a roster published at least a week in advance, a device or app at each store, a clear rule for store-transfer days, and a mechanism for shift swaps between employees. Shift swaps deserve their own small workflow: employee A requests, employee B accepts, store manager approves, and the roster updates so that the register reflects who actually worked. Swaps handled over messaging apps are invisible to the system and reappear as regularisations three weeks later.

Store openings and closings also create a spread-over question — the gap between the first punch and the last punch on a split-shift day. Track spread-over explicitly if your operations use split shifts, because the statutory limits apply to it separately from total hours worked.

Manufacturing and warehouse operations

Here the register is closest to the statutory model: fixed shifts, defined rest between shifts, overtime as a genuine cost line, and contract workers alongside permanent staff. Three configuration points matter. Rest between consecutive shifts must be enforced by the system, not by supervisor goodwill, because rotating a worker from night to morning without adequate rest is both a safety and a compliance issue. Overtime must be pre-approved and capped against daily and weekly limits. And contract workers' attendance must be captured by you, at your gate, rather than accepted from the contractor's own sheet — your record is what protects you if the contractor's numbers are challenged.

Field sales and service technicians

Field roles need proof of work, not proof of desk. Geo-tagged check-in at the customer location, with the visit itself as the unit of record, produces better data than a clock-in at a branch the employee never visits. Set the geofence radius generously — GPS drift in dense urban areas routinely exceeds a hundred metres, and a tight radius produces false failures that erode trust in the system fast.

Travel days need their own treatment. An employee on a train between two cities should be marked present through a travel reason code, not required to regularise a missed punch. And offline capture matters: the app should queue a check-in made in a low-network area and sync it later with the original timestamp.

Hybrid office teams

The question here is not hours but presence pattern. Define the in-office expectation in a measurable unit — days per week or days per month — and report against it transparently. Measure at the team level first and the individual level second; a team that consistently misses the target usually has a workload or space reason, not a compliance one. Keep the consequence proportionate. Most organisations that attached immediate financial consequences to hybrid attendance targets found themselves reversing the decision within two quarters.

Employees on client sites

Employees deployed at a client's premises often punch into the client's system, not yours. Decide up front whose record governs pay. The workable arrangement is that your system remains the record for payroll, with a monthly reconciliation against the client's log, and any discrepancy resolved before your cut-off. Never let a client's attendance system become the direct input to your payroll — you lose both control and auditability.

Troubleshooting the workflows that stall

Even a well-designed workflow degrades. These are the four failure modes worth naming, with the fix for each.

Managers approve everything without reading. The signal is a manager whose approval rate is 100% with a median response time under a minute. The fix is not a lecture; it is a monthly report showing each manager's team regularisation volume against the company median, reviewed by their function head. Visibility corrects this faster than policy.

Regularisations pile up in the last three days. Employees wait until the cut-off looms. Fix it with daily notification of an incomplete day, a weekly digest of open items, and a hard rule that requests older than a stated number of days route to a higher approver. The friction should sit on lateness, not on the request itself.

The same employees appear every month. Before treating this as indiscipline, check the mechanics. A device sited at the wrong entrance, a shift that starts before the first bus arrives, a geofence set too tight, or a role that genuinely begins at a customer site will each produce a persistent "offender" list. Fix the mechanics first; only then have the conversation.

Payroll keeps overriding the locked record. Every override is a signal that the workflow closed too late or the policy has an unhandled case. Log the reason for each override, review the log monthly, and convert the top recurring reason into a policy rule or a system configuration. An override log that shrinks quarter on quarter is a healthy system; one that stays flat means the lock is theatre.

Metrics worth watching

• Attendance compliance rate: days captured cleanly without regularisation, as a percentage. Below 90% suggests a capture problem, not an employee problem.
• Regularisation volume per employee per month: rising volume usually means a device, network or roster problem.
• Approval SLA adherence: percentage of requests actioned within the window, by manager.
• Late-mark distribution: concentrated in one team suggests a shift-timing or commute reality, not indiscipline.
• Overtime hours by team: flags both cost and hours-limit compliance risk.
• Unapproved absence rate: the leading indicator for attrition and disengagement.
• Post-cut-off adjustments: the count should trend toward zero; if it does not, your cut-off or your workflow is wrong.

Review these monthly with function heads. Attendance data is one of the few HR datasets that is timely, complete and objective — it is worth more than the average engagement survey as an early warning system.

Frequently asked questions

What is attendance regularisation, in plain terms? It is the correction of an attendance record that the system captured incompletely or incorrectly — a missed punch, an off-site meeting, a device failure. The employee raises a request with a reason, the manager approves it, and the corrected record flows into payroll.

How many regularisations per month should be allowed? Three to five per employee per month is the common range. The number matters less than having a cap with an escalation path beyond it. If a large proportion of your workforce hits the cap routinely, the problem is your capture mechanism or roster, not your employees.

Can attendance be marked as absent automatically for a missed punch? Yes, if your policy states that default clearly, employees are notified promptly, and there is a working regularisation window before the payroll lock. Defaulting to absent without notice and without a correction window is unfair and will produce disputes and, potentially, wage-payment questions.

Is biometric attendance mandatory in India? No. Employers may choose their capture mechanism. Whatever you choose must produce a reliable attendance record capable of supporting the statutory registers. If you do use biometrics, treat the data as sensitive personal data with corresponding notice, access control and retention discipline.

How should attendance be handled for fully remote employees? Most organisations move remote knowledge workers to an exception-based model: present by default on working days, with leave and holidays recorded, and no hour-by-hour tracking. This produces a clean register while avoiding surveillance that adds no value. Where hours genuinely matter — support rotas, billable services — use a shift or timesheet model instead.

What is a sandwich leave rule and is it permissible? It is a convention where weekly offs or holidays falling between unpaid absences are also treated as unpaid. Practice varies and the permissibility depends on the applicable rules and your employment terms. If you apply it, state it explicitly in the policy and in the appointment letter; do not apply it as an unwritten practice.

When should the attendance cut-off be? Far enough before the pay date for payroll to process, and late enough that only a few days need estimation. The 25th of the month for a month-end pay date is a common landing point. Whatever you pick, hold it — a cut-off that moves is not a cut-off.

How long should attendance records be retained? Retention periods are set by the applicable labour rules and generally run for several years. Retain raw logs as well as processed registers, keep them digitally where permitted, and align the retention schedule with your data protection policy. Verify the current requirement for your establishment type and state.

Should managers be able to mark attendance on behalf of employees? For teams without devices, supervisor marking is often the only practical mechanism, and it is acceptable if it is logged, auditable and reviewed. It should not be a general capability for teams that have a capture mechanism — it dilutes the record and creates an obvious avenue for error and favouritism.

Conclusion: make the boring part automatic

The best attendance policies are unremarkable. Employees know what counts as a working day, see their own record daily, and correct problems within a day or two. Managers approve a handful of requests a week inside a clear SLA. Payroll receives a locked, reconciled dataset on the same date every month and does not renegotiate it. HR looks at patterns rather than individual days.

Getting there is not a technology project so much as a design project — nine decisions, one clear document, one workflow with real windows and caps, and a configuration that has been tested against actual edge cases. The technology then does what technology is good at: applying the same rule to everyone, every day, without getting tired.

CozyHR links attendance, shifts, leave and payroll in a single system, so regularisations route to the right approver, registers generate themselves, and paid days land in payroll already reconciled. If your month-end still involves chasing approvals in a spreadsheet, take CozyHR for a spin and see what a locked, clean cut-off feels like.