Common Payroll Errors and How to Prevent Them: The Ones That Look Correct

Most payroll errors are not wrong in a way anybody can see. A wrong salary is still a plausible salary. That is the whole problem.

Ask a payroll team what went wrong last quarter and you will hear about a missed deadline or a rate that changed. Ask them to reconcile three months of arrears runs and you will find something different: proration computed one way for joiners and another way for leavers, a loss-of-pay divisor that makes February expensive, an increment backdated to April that never reached the overtime rate, and one employee who has been paid on a wage base that stopped being correct in November 2025.

This article is about that second category. It is not a compliance checklist, because we already have one, and the two failure modes are genuinely different. Compliance errors are about what the law requires and are covered in common payroll compliance mistakes Indian businesses make. What follows is about the run itself: the arithmetic, the sequence, and the places where two correct-looking methods produce two different numbers.

Positions are current at August 2026 and reflect the Labour Codes as brought into force on 21 November 2025, the Central Rules notified on 8 May 2026, and the Income-tax Act, 2025, which governs salary paid from 1 April 2026. State rules under the codes are still being notified. Worked examples state their assumptions and should be re-derived rather than adjusted if you change one. Nothing here is legal or tax advice.

The short version, before the detail.

The most common payroll errors in India are not statutory failures but method errors: a defensible calculation applied inconsistently, which produces a number nobody can distinguish from a correct one. Four recur. Proration, where the calendar-day and working-day methods differ by about six per cent for the same mid-month joiner and most payrolls use one for joiners and the other for leavers. Loss of pay, where an actual-days divisor makes the same two days of leave cost ₹4,286 in February and ₹3,871 in March. Backdated arrears, which change provident fund, ESI, the overtime rate and the TDS estimate for the months they cover rather than only the current one. And a wage base that stopped being correct on 21 November 2025, when the Code on Wages redefined what counts as wages. Each is systematic, so it affects every employee it touches until somebody reconciles.

The Errors This Article Is About

Payroll errors fall into three groups, and they need different fixes.

• Statutory errors. A wrong rate, a missed registration, a late remittance. These are knowledge failures and they are fixed by knowing the rule.
• Input errors. A wrong date of joining, an unrecorded leaver, a bank account that belongs to somebody else. These are process failures and they are fixed by controlling who can change what.
• Method errors. An entirely defensible calculation applied inconsistently. These are design failures, they produce numbers nobody can distinguish from correct ones, and they are the subject of most of this article.

Method errors are the expensive category because they are systematic. A statutory error affects one thing you got wrong. A method error affects every employee it touches, every month, in the same direction, until somebody reconciles.

Proration, the Single Biggest Source of Wrong Numbers

An employee joins on 18 September on a monthly salary of ₹60,000. What do you pay them for September?

There are three common answers, all three are used in Indian payrolls, and they differ by about six per cent for the same employee in the same month.

The spread is ₹1,455 on one employee in one month, about six per cent. Across a year of joiners and leavers it is a number worth having decided deliberately.

No method here is wrong. What is wrong is using more than one. The pattern we find most often is calendar days for joiners and working days for leavers, or the reverse, usually because the two were configured by different people at different times. That is not a rounding difference, it is a policy that says arriving and leaving are worth different amounts.

The failure that costs the most, though, is the divisor mismatch. If salary is prorated on a fixed 30-day divisor and the month has 31 days, an employee present for all 31 days is paid 31/30 of their salary. It is a rare configuration and when it exists it has usually been running for a long time, because nobody investigates a payslip that is slightly too high.

How to prevent it. Write the method down as a policy statement, one line, covering joiners, leavers, mid-month structure changes and unpaid leave, and check that the system applies the same one to all four. Then run the test that finds it: pick any month with 31 days, filter for employees with full attendance, and confirm every one of them was paid exactly their monthly salary.

Loss of Pay, and Why Two Systems Disagree

Loss of pay uses the same divisor question and adds a second one, which is what the deduction sits on.

Two days of unpaid leave on a ₹60,000 salary, where the divisor alone decides whether the employee loses ₹3,871 or ₹4,286:

An actual-days divisor means the same absence costs an employee ₹415 more in February than in March. That is defensible and it is also the sort of thing that generates a query every February, so it is worth being able to explain rather than being caught by.

The second question matters more. Does loss of pay come off gross, or off basic, or off gross excluding reimbursements? If a component is a reimbursement against a bill it should not be prorated for absence at all, because the bill does not change. If it is an allowance it usually should. Systems that deduct LOP from a flat gross figure quietly reduce reimbursements the employee has already spent, which is a query you will get.

Then the consequence people miss: LOP changes the wage base, so it changes PF and it can change ESI. An employee whose wages fall below the ESI threshold in a month because of unpaid leave is not thereby uncovered, because coverage is tested by contribution period rather than by month. Dropping them out for that month is a common and entirely avoidable error.

How to prevent it. Define, per component, whether it is prorated for LOP. Do it once, in the salary structure, rather than as a rule in the payroll engine, so the answer is visible to whoever looks at the structure next.

Arrears Change the Past, Not Just the Present

An increment is approved in September, backdated to April. Five months of arrears are paid in the September run. Most payrolls handle the salary difference correctly and then stop, and that is where the error is.

A retrospective increase changes the wage figure for the months it covers. Everything computed on that figure for those months is now understated:

• Provident fund. PF wages for April to August were higher than what was contributed on, unless the employee was already at the ceiling. The arrears attract their own contribution and it belongs to those months.
• ESI, where applicable, on the same logic, and with the contribution period rule deciding whether coverage itself changes.
• Overtime. This is the one that is almost always missed. Overtime is paid at not less than twice the normal rate of wages. If the normal rate went up in April, every overtime hour worked since April was paid at the old rate and is short. Nobody recomputes overtime when they process arrears, and in an establishment with a lot of overtime this is the largest number in the whole exercise.
• The TDS estimate. An arrears payment raises the projected annual income, which raises the tax for the year, which is spread across the remaining months. Treating arrears as a one-off deduction at slab rate produces a different, wrong answer.
• Leave encashment and gratuity, both of which sit on the current wage, and both of which will be computed on the higher figure at exit whether or not the arrears were processed properly.

How to prevent it. Make the arrears report show the period the arrears relate to, not just the amount. A report that says "₹42,000 arrears" cannot be checked by anybody. A report that says "₹8,400 per month, April to August" can, and it makes the four consequential recomputations obvious.

The Wage Base That Stopped Being Correct in November 2025

This one deserves its own section because it affects a large share of Indian salary structures and produces no visible symptom at all.

The Code on Wages introduced a uniform definition of wages across all four codes. Basic pay, dearness allowance and retaining allowance are wages. The excluded allowances are capped: where they exceed one half of all remuneration, the excess is added back into wages.

Structures built with a low basic and a large allowance stack, which was extremely common, now have a higher statutory wage figure than the basic sitting in the payroll system. Everything computed on wages is affected: provident fund, gratuity, leave encashment and the overtime rate. If your payroll is still computing on the old basic, it has been under-computing since November 2025, per person, per month.

The symptom is that there is no symptom. Payslips look normal, remittances go out on time, nothing is rejected. It surfaces at an exit, when a gratuity figure is challenged, or at an inspection.

How to prevent it. Take one employee with a typical structure and add up the excluded allowances as a share of total remuneration. If they exceed half, your wage base needs the add-back and probably your structure needs revisiting. The component-by-component version of this is in our CTC breakup guide, and the salary allowances piece covers which allowances sit on which side.

Master Data Errors That Surface Months Later

These are unglamorous and they account for most of the queries HR actually receives.

The duplicate bank account row is worth pausing on. It is the only error on this list that can be deliberate, it is invisible in every report that looks at salary rather than at destination, and a uniqueness check takes an afternoon to build. Run it once against your existing master even if you never run it again.

Two more that belong here. A mid-year joiner who does not submit the declaration of previous employment leaves both employers applying the basic exemption and the rebate to their own slice of the year, which under-deducts across the year and produces a bill at filing. And an employee who changes state mid-year needs the professional tax registration in the new state to exist before their first payslip there, not after.

The Full and Final That Goes Wrong

Exit settlements concentrate every error in this article into one calculation, and they are usually done under time pressure by whoever is available.

• Proration on the wrong method, per section 2, and this is where the joiner-versus-leaver inconsistency shows up.
• Leave encashment on the wrong base. Check what your policy says the encashment rate is, and check it against the statutory wage definition rather than against basic by habit. Our leave encashment calculator runs it.
• Notice pay recovery exceeding the deduction cap. Total deductions in any wage period may not exceed fifty per cent of wages. A recovery that takes more than half of the final settlement has to be handled as a recovery outside payroll, not as a deduction inside it. The notice period calculator works through the shortfall arithmetic.
• Gratuity computed on the wrong wage or the wrong service. Both inputs come from the master and both are commonly wrong there. Our gratuity calculator and the fuller gratuity guide cover eligibility and rounding of service.
• TDS on the settlement. The final payment is part of the year's income and belongs in the annual estimate. Settling it at a flat rate, or not at all, moves the problem to the employee's return.
• A negative net settlement processed as a zero rather than as a receivable, which quietly writes off money and leaves no record of the write-off.

The Ten-Minute Pre-Release Check

Run these before the bank file goes out. Together they take about ten minutes on a payroll of a few hundred and they catch most of what this article describes.

• Headcount paid against the HR master. Any difference at all is investigated, not explained.
• Net pay movement per employee against last month. Anything above fifteen per cent gets a reason recorded next to it.
• Employer PF as a percentage of PF wages. Anything other than 12.00 per cent means a member is configured wrongly.
• Any negative net pay, and any deduction above half of wages. Both are structural, not arithmetic.
• Full-attendance employees paid exactly their monthly salary. This is the divisor test from section 2.
• Bank file total against the register's net pay total. Exact match, and the payee count matches too.
• Overtime hours in payroll against the muster roll for the same period. Both directions.
• Every arrears line has a period, not just an amount.

Note what is not on the list. Checking individual salaries for correctness is not a control, because a wrong salary looks like a right one. Every item above compares two independent records and looks for a difference, which is the only kind of check that finds a plausible-looking error.

Questions People Ask

What is the most common payroll error in India?

Inconsistent proration. Two defensible methods, calendar days and working days, applied differently to joiners and leavers, usually because they were configured at different times by different people. For an employee joining on 18 September on ₹60,000 a month, the calendar-day method pays ₹26,000, a fixed 26-day divisor pays ₹25,385 and the working-day method pays ₹24,545. The spread is ₹1,455 on one person in one month, about six per cent, and neither number looks wrong on the payslip. The related failure is a divisor mismatch: prorating on a fixed 30-day divisor in a 31-day month pays a fully present employee 31/30 of their salary, which nobody investigates because the payslip is slightly too high rather than too low. Pick one method, write it down, and apply it to joiners, leavers, structure changes and unpaid leave alike.

How do you calculate salary for a mid-month joiner?

Pick one method and apply it to joiners, leavers, structure changes and unpaid leave alike. The calendar-day method, days available divided by days in the month, is the most common in India and the easiest to explain to an employee. The working-day and fixed-26-day methods are equally defensible. What is not defensible is using different methods for arriving and leaving, which is what most payrolls do without intending to.

Should loss of pay be deducted from gross or basic salary?

Component by component rather than from a single figure. Allowances that compensate for attendance or presence are normally prorated; reimbursements against an actual bill are not, because the bill does not shrink when the employee takes leave. Decide it once in the salary structure so the answer is visible where the structure is defined rather than buried in payroll configuration.

What do I need to recompute when I pay backdated arrears?

Five things beyond the salary difference: provident fund for the arrears months, ESI where applicable, the overtime rate for hours already worked at the old rate, the TDS annual estimate, and any leave encashment or gratuity that has been computed since. Overtime is the one almost always missed, and in an establishment with regular overtime it is often the largest of the five.

How do I check my payroll before releasing salaries?

Compare independent records rather than reviewing figures. Headcount against the HR master, net pay against last month, the bank file total against the register total, overtime hours against the muster roll, and employer PF against 12 per cent of PF wages. Each of these compares two sources that were produced separately, which is what makes a plausible-looking error visible.

Why does the same leave cost more in February?

Because a divisor based on actual days in the month makes each day of a 28-day month worth more. On a ₹60,000 salary, two days of unpaid leave costs ₹4,286 in February and ₹3,871 in March. It is correct arithmetic on that method and it is a question you will be asked, so it is worth having the answer written down. A fixed 30-day divisor removes the variation at the cost of a small mismatch with the calendar.

Who should be allowed to approve payroll changes?

Separate the person who enters a change from the person who approves the run. That single separation prevents the two errors on this list that can be deliberate rather than accidental, a duplicate bank account and an unstopped leaver, and it costs nothing to implement. Keep a record of who changed what and when, because a payroll that cannot answer that question cannot be audited.

Where This Leaves You

Three things are worth doing this month, in this order.

Run the divisor test. Pick a 31-day month, filter for full attendance, and confirm everybody was paid exactly their monthly salary. It takes five minutes and it either clears the largest method error on this list or finds it.

Check your wage base against the definition in the Code on Wages. If the excluded allowances in a typical structure exceed half of total remuneration, your PF, gratuity, leave encashment and overtime rate have all been computed low since November 2025, and the arrears accumulate quietly.

Reconcile one month of overtime against the raw punch data. In the migrations we run, this is the check that finds money, and it finds it in both directions.

Method errors persist because the payroll produces a plausible number and nothing objects. The fix is structural rather than diligent: keep the attendance record, the register and the filing on one record so the comparisons in section 8 are queries rather than exercises. That is what our payroll management system and attendance management software are built to do together. Book a free demo and bring a month with a lot of joiners in it, because that is the month that shows the most.

Related reading: payroll reports every HR team needs, overtime pay calculation in India, the payroll compliance checklist, and payslip format and mandatory components.

Sources

• Code on Wages, 2019, enacted text on India Code. Section 2(y) on the definition of wages and the one-half proportion rule, section 14 on overtime, and section 18(3) capping total deductions at fifty per cent of wages in a wage period.
• Employees' Provident Fund Organisation, epfindia.gov.in. Contribution rates, the wage ceiling and the treatment of arrears in the electronic challan cum return.
• ESIC contribution rates, esic.gov.in. Contribution rates and the contribution-period rule that governs an employee crossing the wage threshold mid-period.
• Ministry of Labour and Employment, labour.gov.in. Commencement of the four Labour Codes on 21 November 2025 and the Central Rules notified on 8 May 2026.

Worked examples state their assumptions and were computed rather than estimated. State rules under the Labour Codes are still being notified and several positions above are state-specific. Verify the current position on the relevant portal before configuring a payroll run.