Attendance Policy & Regularisation: A Practical Guide

Most payroll disputes in Indian companies do not begin in payroll. They begin three weeks earlier, when a field executive forgets to punch out, a supervisor swaps two week-offs on WhatsApp, and a hybrid designer works from a cafe without telling anyone. None of it is malicious. All of it becomes a salary argument on the 1st. A clear attendance policy — one that defines the workday, sets numeric thresholds, and explains exactly how corrections happen — is what stops those small events from becoming a helpdesk queue.

The usual failure is that the attendance policy is written once, pasted into a handbook, and never connected to the payroll cut-off. It says employees must be punctual. It does not say what happens when someone is twelve minutes late on four days, whether that is a half day, who approves the correction, by when, and what happens if approval lands after payroll is locked. That gap is where loss of pay (LOP) errors, arrears and lost trust live.

This is a build guide, not a lecture. We will cover defining the workday for office, hybrid, shift and field teams; designing an attendance regularisation process with reason codes, evidence standards and SLAs; building a cut-off calendar with a hard freeze; and handling the messy cases — night shifts crossing midnight, week-off swaps, comp-off, GPS for field staff, and corrections after payroll has run. Every number below is illustrative and shows method, not a rule to copy.

Why a Weak Attendance Policy Quietly Becomes a Payroll Dispute

Nobody complains about a vague policy. They complain about their salary.

The chain is predictable. The policy is ambiguous about what counts as a present day, so the system produces daily statuses using vendor defaults nobody reviewed. Managers see exceptions late and bulk-approve on cut-off day. A few get missed. Payroll runs, three people see an unexpected LOP, and HR must choose between an arrear next month or an off-system reversal.

The failure modes each have a design fix:

• Undefined thresholds. "Late coming will be viewed seriously" is not a rule. A rule states a grace window in minutes and a conversion count.
• No single source of truth. Leave in a sheet, attendance in a biometric export, WFH in email, comp-off in a manager's memory. LOP then gets calculated twice, differently.
• No cut-off discipline. Regularisations trickle in after the freeze, so payroll either delays or processes known-bad data.
• Approval without accountability. Managers clear queues by approving everything, so genuine cases get no more scrutiny than fabricated ones.
• Silent defaults. The HRMS was configured at onboarding with a 15-minute grace nobody in HR knows exists. The system is writing your policy.
• Undocumented team differences. Sales does not punch, warehouse does. Often correct — but unwritten, it reads as favouritism.

The cost is a steady drip of HR ops hours, manager time and payroll rework. The fix is mostly writing and configuration, not budget.

What Your Attendance Policy Must Define

Use one completeness test: if a rule cannot be encoded as a condition in your attendance management system, India-specific or otherwise, it is not a rule yet — it is a sentiment.

The Workday, Shifts and Core Hours

Define, separately for each employee category:

• Shift timings per location and team; daily and weekly hours; whether breaks are paid and count towards hours worked.
• Core hours — the window when everyone must be available. This is what makes flexi-time survivable.
• Minimum hours for a present day and for a half day.
• Whether marking is time-based (checks in/out times), hours-based (checks duration), or both. Many companies run hours-based for knowledge teams and time-based for shift and customer-facing teams.

Before you set any number: working hours, daily and weekly caps, spread-over, overtime, rest intervals, weekly off entitlement and attendance record-keeping are governed by the applicable labour codes and the relevant state Shops and Establishments rules, or the Factories rules where they apply. These vary by state and establishment type and are amended periodically. Verify the currently notified rules for every state you operate in and design at or above those minimums — a policy cannot contract below a statutory entitlement, and employee consent does not make it enforceable.

Grace, Late Coming and Half Day Rules

Grace absorbs ordinary friction — traffic, a lift queue, a slow reader. It does not move the shift start. Define the window in minutes, whether it is per instance or a monthly bank, what happens beyond it, whether early departure is treated symmetrically, and whether making up hours later cancels a late mark.

That last point decides a lot. Under time-based late coming and half day rules, someone who arrives 40 minutes late and works 40 minutes extra is still late. Under hours-based rules they are not. Both are defensible — pick one and make the system match.

Thresholds must be explicit and mutually exclusive:

Two cautions. Do not stack punitive rules — if three lates already convert to a half day, do not also apply an hours-based half day to the same days. And decide the order of operations, because ambiguity here produces different answers from different people reading identical data.

Weekly Offs, Holidays and the Sandwich Question

This is the most disputed area after LOP itself. Document the weekly off pattern including rotational shifts; the national, state and optional holiday list; how an unauthorised absence adjacent to a holiday is treated; and what happens when a holiday falls on a weekly off.

Then settle sandwiching — whether a weekly off or holiday falling between two leave days counts as leave. State it in plain language with examples, check it against applicable state rules and your establishment's settled practice, and configure it once in the system. Changing it mid-year without notice reliably generates grievances, and manual sandwich maths in a spreadsheet reliably generates inconsistency.

Attendance Status Codes and Payroll Treatment

Give every day-state a code. Payroll should never see a blank cell.

The MP code does quiet but important work. Without it a missed punch looks identical to absence, employees see "Absent" on their portal and flood HR. With it, they see precisely what is wrong — and your missed punch approval workflow has something concrete to attach to.

Finally, state scope explicitly: probation, interns, fixed-term, contract staff, and whether senior grades are exempt from punch marking while still required to apply for leave. Every category the policy is silent on gets decided on the spot, inconsistently.

Choosing Capture Methods

The capture method determines your error profile — which exceptions you will spend the month resolving.

Two rules. Assign one primary method per employee with one documented fallback — people who can punch three ways punch inconsistently. And pre-approve the fallback: if the device is down, the policy should already say the supervisor logs an outage, employees mark via app or signed muster, and HR regularises in bulk under one reason code. Improvised fallbacks create ghost entries.

Your biometric attendance policy needs its own paragraph, because biometric data is sensitive personal information. State what is captured (in most modern systems a mathematical template, not a stored fingerprint image), why, who can access it, retention period, and what happens at exit. Provide a genuine alternative for employees who cannot enrol — worn fingerprints from manual work are common and are not misconduct. Align all of it with your privacy notice and applicable data protection obligations.

What the Attendance Regularisation Process Is — and When It Is Abuse

Regularisation is the correction of a factually wrong attendance record. That is the entire definition, and holding to it resolves most arguments about it.

Legitimate: the employee worked or was on approved duty, but the system does not reflect it. The record is wrong; the correction makes it right. Abuse: the record is correct and the employee wants it changed — that is a leave application, an unpaid absence, or a conversation with a manager.

Put the test in the policy in one line — regularisation corrects a recording error for time actually worked; it is not a substitute for applying leave — then make legitimate cases effortless so illegitimate ones have nowhere to hide.

Watch for these patterns: regularising a full day when someone arrived after lunch; backdated bulk requests filed on cut-off day when scrutiny is lowest; the same weekday recurring; a manager approving everything within seconds; and "missed punch" claimed on days with no other digital footprint. You do not need to catch every case, only to make patterns visible.

Reason Codes and Evidence Standards

This table is the operational heart of the process. Adapt the codes; keep the columns.

Three principles keep it workable. Evidence should be proportionate — demanding documents for one forgotten punch-out costs more friction than the risk it manages; self-declaration plus a digital footprint is enough for routine cases. Codes must be mutually exclusive, or people pick the easiest one and your data stops meaning anything. And make the free-text note mandatory in the system — codes give you the category, notes give you the story.

Approval Chain, SLAs and Auto-Decisions

Regularisation dies of latency. A request that sits for nine days lands on cut-off day with no time for scrutiny.

Then the harder question: what happens to a request nobody acted on? Three choices, and you must write one down.

1. Auto-approve on breach. Keeps payroll clean, but removes managerial control — once managers learn it, the queue stops being read.
2. Auto-reject on breach. Firm, but it penalises the employee for the manager's inaction and is hard to defend when the claim was genuine.
3. Auto-escalate, then default at the freeze. Requests climb the chain on breach; anything still open at the freeze gets a documented default — approve for low-risk codes like missed punch and outage, hold for HR decision on high-risk codes.

Option 3 holds up in practice. Add guardrails whichever you pick: cap auto-approval by volume and by code, never auto-approve pattern-flagged items, log auto-decisions distinctly so "auto-approved on SLA breach" never looks identical to "approved by Priya on the 14th", and notify managers when auto-decisions occur on their team.

The Attendance to Payroll Cut-Off Calendar

This is where the policy survives or does not. Assume an illustrative cycle running the 1st to month-end with credit on the last working day.

Many companies instead run a 21st-to-20th attendance month to create more processing room, at the cost of a small lag between working and being paid. Either shape works; what matters is leaving payroll at least two clear working days after the freeze.

Why a Hard Lock Matters

A soft cut-off everyone knows is negotiable is not a cut-off. The freeze does four things:

1. Makes the payroll input immutable. Edit a record on the 28th evening and the LOP register silently diverges from processed payroll — you will learn about it from a complaint, not a report.
2. Creates a defensible audit position. You point to a frozen dataset with a timestamp and an approval trail, not a live table that has been edited since.
3. Forces the upstream behaviour you want. Managers clear queues when deadlines are real. Month one brings complaints; month two the queue clears on time.
4. Stops the "one more correction" spiral. Every late exception invites the next, and payroll never gets a stable base.

Make the lock technical, not procedural — the system should refuse edits for everyone except a named break-glass role. And publish the calendar at the start of the financial year, in the HRMS and as a recurring manager invite.

Corrections After Payroll Has Run

Some corrections arrive late and are genuinely valid. You need a written route that does not reopen a processed month.

• Never reopen a closed period. It breaks reconciliation with statutory filings, bank advice files and issued payslips, and makes every prior month feel provisional.
• Distinguish arrears from adjustments. An arrear pays money owed for a past period — an LOP reversal — as a separate labelled line naming the affected month. An adjustment corrects the current period's computation because of a past event, such as a restored leave balance. Show both distinctly; a silently larger net pay generates its own ticket.
• Define a correction window — illustratively, until the end of the following month, beyond which HR Head approval and a documented reason are required.
• Log the root cause every time: employee delay, manager delay, device failure or system error. If most trace to one department, that is a management conversation, not a payroll fix.
• Cap the break-glass. Two named approvers, a mandatory reason, and a monthly report reviewed by someone senior. Recurring volume means the calendar is set wrong, not that you need more exceptions.

Field, Hybrid and Remote Teams

"Present" means something different for a field executive than for an accountant, and the policy should say so.

For field roles, attendance is activity confirmation, not gate-crossing. Use route or beat plans where a completed beat is a present day; geofenced check-in with a realistic radius, since a tight radius in a dense urban area produces constant false exceptions; and activity corroboration — pairing a punch with a visit note or order entry is more robust than location alone, and less intrusive.

On privacy, be explicit. Continuous background tracking is a significant intrusion and a common source of resentment in field teams. Capture location at events, restrict it to working hours, state in writing what is captured and why, who sees it and how long it is retained, obtain informed consent with a documented alternative, and limit access to the manager and HR ops. Align this with applicable data protection obligations. A team that sees the tool as removing daily reporting calls will use it; a team that feels watched will find workarounds and your data quality collapses.

For hybrid and remote staff, do not reproduce office punch discipline at home — it produces theatre, not information. Define the WFH entitlement and booking method, keep marking light (web or app punch plus core-hours availability), and state what counts as unavailability, usually handled as a performance conversation rather than a deduction. Track office-day compliance as a separate metric; folding it into daily attendance status corrupts your LOP logic. Avoid keystroke or screenshot monitoring as an attendance proxy — disproportionate for most office roles, and the trust damage is hard to reverse.

Shift, Roster and Night Shift Teams

Shift teams break naive configurations in predictable ways.

• Night shifts crossing midnight. A 22:00–06:00 shift spans two dates. The system must attribute it to one date, conventionally the start date, and must not read the 06:00 punch-out as a stray next-day punch. Get this wrong and one correct shift produces two broken days — an absence and a missed punch. It is the single most common configuration bug; test it with real data before go-live.
• Roster publication. Publish rosters a defined period in advance (illustratively seven days) and lock them before cut-off. A roster changed after the fact makes that team's attendance unverifiable.
• Week-off swaps. Require a formal request naming both employees and both dates, approved and recorded before the swap. A WhatsApp message between two employees is not a roster change. Use a distinct reason code for after-the-fact approvals so you can measure how often it happens.
• Shift allowances must come from the same frozen dataset as LOP. Two sources for "how many night shifts did she work" guarantees a dispute.
• Overtime must be pre-approved and captured against a specific date and shift, computed per applicable statutory provisions and your policy — verify currently notified eligibility, rate and limits, and remember policy cannot reduce a statutory entitlement. Never infer overtime from a late punch-out; someone finishing personal work is not doing approved OT, and automatic OT on stray minutes creates cost and precedent problems.
• Comp-off. Define the trigger, the credit, a validity window (illustratively 60 days), the approval path, and treatment on expiry and at exit. Credit it automatically from attendance data — manual comp-off claims are a documented mess in every company that runs them.

Integrating Leave and Attendance So LOP Is Calculated Once

If leave and attendance live in separate systems, LOP gets calculated twice and the answers differ. Run this sequence once, in the system, at the freeze:

1. Build each employee's daily calendar: working days, weekly offs, holidays, roster.
2. Overlay approved leave; those days take their leave code.
3. Overlay attendance data for the remaining working days.
4. Apply regularisation outcomes, overwriting raw status.
5. Apply threshold rules: grace, late conversion, half day and full day.
6. Apply the sandwich rule if you have one.
7. Resolve unmarked working days with no approved leave to absent.
8. Consume available leave balance in a stated priority order where policy permits.
9. Whatever remains is LOP — one number, one place, one calculation.
10. Freeze and publish the LOP register to managers before processing.

Step 10 is the cheapest dispute prevention available. A one-page LOP register per department, sent on the 27th, catches most errors before they reach a salary slip; managers recognise their own team's anomalies in a way HR cannot.

Worked Examples (Illustrative Numbers)

All figures are illustrative and chosen for arithmetic clarity. Your LOP base — gross, basic plus DA, or another defined component — must be set by policy and checked against applicable statutory and contractual requirements.

Example 1 — full-day LOP. Monthly gross Rs 60,000, divisor 30 days, per-day value Rs 2,000. One unapproved absence with nil leave balance gives LOP of Rs 2,000. Note the divisor: calendar days, fixed 26 days, or actual working days each produce a different per-day value for the same salary. State your choice and never switch it between months.

Example 2 — half day from hours. Same per-day value. On the 12th the employee logs 4 hours 30 minutes, between the 4-hour and 6-hour thresholds, so the status is HD. With nil balance, LOP is Rs 1,000.

Example 3 — half day from late marks. Illustrative rule: 10-minute grace, three lates in a month convert to one half day.

Three lates convert to one half day, offset against leave balance if available, otherwise LOP of Rs 1,000. Do not also apply an hours-based half day to those same three dates — that double-counts the same lateness. Decide too whether the counter resets monthly; monthly reset is simpler to explain.

Example 4 — a mixed month with regularisation. Monthly gross Rs 45,000, 30-day divisor, per-day value Rs 1,500.

Four of six flagged days were recording errors, not attendance problems. That ratio is normal, and it is exactly why regularisation must be easy for genuine cases — make it painful and employees stop bothering, absorb wrong deductions silently, then escalate three months at once.

Example 5 — correction after the freeze. On the 3rd, the employee shows the 17th was an approved WFH day recorded in email but never entered in the system. The closed period stays closed. HR verifies, logs the correction under RG-05 with root cause "approval not entered", and next month's payroll carries a labelled line: "LOP reversal — previous month: Rs 1,500". The employee is made whole, the audit trail is intact, and the process failure is captured rather than absorbed.

Attendance Policy Template Outline (HR Policy Template India)

Use this skeleton for your own document. Keep each section to a paragraph plus a table where possible — a policy nobody finishes reading protects nobody.

1. Purpose and scope — objective; categories and locations covered and excluded; effective date, version, review cycle; links to leave, hybrid and conduct policies.
2. Definitions — working day, working hours, core hours, shift, roster, weekly off; present, half day, absent, short hours; grace, late arrival, early departure; regularisation, cut-off, freeze, LOP; comp-off, overtime, on-duty, field duty.
3. Working hours and shifts — timings by location and team; daily and weekly hours; break treatment; flexi-time; roster publication and lock timelines; night shift date attribution; a statement that all provisions operate at or above applicable statutory minimums.
4. Attendance capture — primary method by category and approved fallbacks; biometric purpose, access, retention and alternatives; GPS scope, consent and privacy safeguards; proxy-punching prohibition; device outage protocol.
5. Marking rules and thresholds — grace; late and early-departure treatment; accumulation and conversion; half day and full day thresholds; status code table; order of precedence.
6. Weekly offs, holidays and leave interaction — weekly off pattern; holiday and optional holiday rules; sandwich rule with examples; interaction with each leave type; leave-balance consumption priority before LOP.
7. Attendance regularisation — definition and legitimacy test; reason code and evidence table; submission window; approval hierarchy and SLAs; auto-escalation and default-at-freeze rules; monthly limits; misuse.
8. WFH, field duty and travel — entitlement, booking and marking; core-hours availability; field check-in requirements; geofence rules; travel-day treatment.
9. Overtime and comp-off — eligibility, pre-approval and computation per applicable statutory provisions; comp-off trigger, credit, validity, expiry and treatment at exit.
10. Payroll cut-off and freeze — monthly calendar; soft cut-off and hard freeze; post-freeze restrictions and break-glass approvals; arrears versus adjustments; correction window.
11. LOP computation — base and divisor stated explicitly; calculation sequence; partial-day treatment; payslip presentation of LOP and reversals.
12. Roles and responsibilities — employee, manager, shift supervisor, HR ops, payroll, and IT or facilities for device uptime.
13. Monitoring, audit and records — records and retention aligned to applicable statutory requirements; audit trail; periodic sampling; metrics and reporting lines.
14. Grievance and dispute resolution — channel, response SLA, escalation path, documentation and closure.
15. Non-compliance — habitual lateness; unauthorised and extended absence; falsification of records; link to the disciplinary process and principles of natural justice.
16. Annexures — status codes, reason codes, cut-off calendar, shift master, holiday list, FAQ.

Version the document with an effective date and change log, because a dispute about a period six months back needs the version that applied then. And write a one-page employee summary: shift timings, grace, thresholds, how to regularise, and the cut-off date. That is the version people actually read.

Audit Trail and Dispute Handling

Attendance data is a business record with statutory relevance; confirm the currently notified record-keeping requirements, retention periods and permissibility of electronic registers for your establishment and states.

Retain, for every record: raw punch events with timestamp, device identifier and location where applicable; the computed status and which rule produced it; every request with submitter, reason code, note and evidence reference; every decision with actor, timestamp and whether it was human or automatic; every post-freeze correction with approvers, root cause and settlement cycle; and configuration change history. That last item is routinely overlooked and routinely decisive — a dispute about a late mark from four months ago needs the grace setting that applied then, not today's.

For disputes, run one consistent process. The employee raises it through a defined channel, ideally a ticket. HR pulls the raw data and audit trail and compares the computed outcome against the rules in force on that date, then communicates the outcome in writing with specific reasoning. If the company was wrong, correct it via arrears or adjustment and log the cause; if it was right, check whether the policy was unclear — if two people made the same mistake, it was.

Reasoning matters as much as outcome. "Your 14th was a half day because you logged 4 hours 15 minutes against a 6-hour minimum" closes a dispute. "The deduction stands" opens a new one.

Common Mistakes

• Writing the policy without auditing the configuration. The policy says 10 minutes; the system says 15. The system wins, because it computes salary. Check line by line before publishing and after every vendor upgrade.
• Treating a missed punch as an absence. It is a data gap, not behaviour. Separate status, separate workflow, separate metric.
• Setting the cut-off too close to the pay date. A freeze on the 29th with credit on the 30th leaves no room to fix anything.
• Allowing unlimited backdating. Requests for dates two months old cannot be verified. Set a window and enforce it.
• Letting managers bulk-approve unread. Forty approvals in two minutes is not efficiency. Publishing per-manager statistics corrects it faster than any rule.
• Running LOP before checking leave balances. It produces deductions that then need reversal.
• Copying another company's policy. A 400-person factory muster does not fit a 60-person hybrid product team, and copying imports rules that may not match your states or shift patterns.
• Skipping night shift testing. The midnight boundary is where most configurations break.
• Penalising employees for device failures. An outage is the company's problem — one bulk regularisation, not forty individual requests.
• Making regularisation harder than necessary. Every extra field and approval level reduces compliance and pushes genuine corrections underground.
• Never reviewing the policy. Shift patterns change, offices move, state rules get amended. Review annually and after any material change in how work is done.
• Not deciding the divisor. Whether per-day value uses 26, 30, 31 or working days changes every LOP figure you compute.

Metrics to Monitor

Read these together. A rising regularisation rate is not automatically bad — it may mean employees have finally started correcting records they previously accepted in silence.

Rollout: Manager Training and Communication

Policies fail at rollout more often than at drafting.

Do not launch with a PDF. Launch with a one-pager, a short session and a live demo of the actual screens. Train managers separately on what regularisation is and is not, how to assess a request in under a minute, what the SLA is and what happens when they miss it, how to read their exception dashboard, and how to talk to someone regularising too often — as a working session with real cases from their own team, not slides.

Explain the why, especially the freeze: "the cut-off is the 25th because payroll needs two clear days, and after that a correction becomes an arrear next month" earns cooperation that a bare rule does not. Communicate in the languages your workforce uses; for plant, retail and field teams an English-only rollout is a non-rollout.

Then run a shadow month — generate exceptions, run the workflow, produce the LOP register, deduct nothing, and reconcile against your existing method. Every discrepancy found in the shadow month is a dispute you did not have in the live one.

A 30/60/90 Day Implementation Plan

Days 1–30 — define and configure. Audit the last three months of attendance and payroll data for exceptions, LOP instances, manual corrections and disputes; that is your baseline and your business case. Map every employee category to a work pattern, and verify the currently notified statutory position for each state and establishment type. Draft the policy from the outline above and fix the hard numbers: grace, thresholds, conversion counts, sandwich rule, LOP base and divisor, correction window. Build the reason code table and test each code against ten real historical scenarios. Design the cut-off calendar backwards from the credit date. Configure the system to match the policy exactly, documenting each setting against the clause it implements, then test edge cases with real data — night shift, week-off swap, half day next to a holiday, outage day, mid-month joiner and leaver. Get sign-off from HR, Payroll and Finance, plus legal review of the statutory and privacy sections.

Days 31–60 — pilot and train. Pilot with three teams covering different work patterns — one office, one shift, one field. Run a full shadow cycle through to a no-deduction LOP register, reconcile line by line, and investigate every difference; most will be configuration bugs. Train pilot managers and collect their friction points, because they will find what testing missed. Fix, re-test, and repeat the shadow run if the fixes were substantial. Finalise the communication kit and stand up the exception dashboard — start tracking metrics now, or you lose your baseline.

Days 61–90 — launch and stabilise. Roll out with a defined effective date and adequate notice, training by site in the local language on live screens. Enforce the first hard freeze, while routing genuine post-freeze cases through arrears so people see the safety net works. Review the first live cycle: pending at freeze, break-glass edits, disputes, root causes. Hold a manager review with per-team metrics visible, fix the three or four things that broke, and communicate the fixes. Then publish the annual calendar, set the cadence of monthly metrics and annual policy review, and compare against your day-one baseline.

Frequently Asked Questions

How many days should employees get to raise a regularisation request?

Three to five working days from the affected date suits most Indian companies — short enough that memory and evidence are fresh, long enough to cover short leave or travel. Whatever you choose must close comfortably before the soft cut-off, or a valid request can arrive with no time left for approval.

Should requests auto-approve if the manager does not act?

Auto-approval keeps payroll clean but removes oversight, and managers quickly learn to ignore a queue that resolves itself. Prefer auto-escalation to the skip-level on SLA breach, with a documented default at the freeze — approval for low-risk codes like missed punch and outage, HR decision for higher-risk ones. Log auto-decisions distinctly in the audit trail.

Can we deduct salary for late coming?

Any deduction must fall within what is legally permissible for your establishment, and the applicable provisions on permitted deductions under the relevant labour code and state rules govern this. Most companies avoid direct monetary fines and instead convert accumulated late marks into a half day charged against leave balance, with LOP only if no balance exists. Have any mechanism reviewed against currently notified requirements and your employment contracts before adopting it.

How do we handle attendance when the biometric device fails?

Treat it as a company-side event. The supervisor logs an outage with start and end times, employees mark via the pre-approved fallback, and HR applies a bulk regularisation under a dedicated outage code. That resolves in one action instead of dozens and gives a clean audit explanation. Track outage frequency by device — one that fails monthly needs replacing, not more regularisations.

What is the right attendance cut-off date in an Indian payroll cycle?

It depends on your credit date and how much validation time payroll needs. A workable illustrative pattern for month-end credit is a soft cut-off on the 25th, approvals closed on the 26th, hard freeze on the 27th evening, and processing on the 28th and 29th. A 21st-to-20th attendance month is the common alternative. What matters is at least two clear working days after the freeze.

How do we manage field attendance without invading privacy?

Capture location at discrete events rather than continuously, restrict it to working hours, and pair it with a work artifact such as a visit note or order entry, which is more reliable and less intrusive than location alone. Tell employees in writing what is captured, why, who sees it and for how long; obtain informed consent; and restrict access to the manager and HR ops.

What is the difference between an arrear and an adjustment?

An arrear pays money owed for a closed past period, such as reversing an incorrect LOP, and appears as a separate line naming that period. An adjustment changes the current period's computation because of a past event, such as a restored leave balance being consumed differently. Label both on the payslip.

Do hybrid employees need different attendance rules?

They need different marking rules, not a different philosophy: a defined WFH entitlement, a light marking method, and a clear core-hours availability expectation. Track office-day compliance as a separate metric rather than folding it into daily attendance status, or you will corrupt your LOP logic.

Conclusion

A good attendance policy is not a document about punctuality. It is the specification for how time worked becomes money paid, and it has to be precise enough that a system can execute it without a human interpreting it differently each month.

The pieces are unglamorous and mutually dependent. Define the workday and thresholds in numbers. Give every day-state a code and every code a payroll treatment. Make regularisation genuinely easy for real errors, with reason codes, proportionate evidence and real SLAs so abuse becomes visible. Run the leave and attendance sequence once so LOP is calculated once. Publish a cut-off calendar and enforce a hard freeze. Handle late corrections as labelled arrears instead of reopening closed months. Keep an audit trail that includes configuration changes, and watch a small set of metrics.

Do that and the 1st stops being a support queue. Managers know what to do and by when, payroll works from one frozen dataset, and employees can see their record, correct it within a defined window, and understand why their salary is what it is.

If you are rebuilding this, it helps to see it working end to end. CozyHR handles configurable shift and grace rules across office, hybrid, shift and field teams, with self-service regularisation routed through your own approval chains and SLAs. Attendance syncs straight into payroll with a hard cut-off lock and exception dashboards that surface what is pending before it becomes a deduction dispute — worth a look while you are drafting.