Short answer: stay on Excel below about 25 to 30 employees if you run a single-state payroll with uniform salary structures, and move to HR software above that, or earlier if your complexity is high. The trigger is not headcount by itself. It is the number of moving parts: states, PT slabs, shift patterns, variable pay components, and how badly your business breaks if the one person who owns the payroll sheet resigns tomorrow.
I watched a 34-person design studio in Bengaluru learn this the expensive way. They had a beautiful payroll sheet. Genuinely beautiful, built over four years by their ops lead, with nested formulas for PF ceilings and a separate tab for professional tax. Then she moved to Berlin. The founder opened the file, found nine tabs, three of them feeding a tenth through cell references nobody had documented, and paid a consultant ₹60,000 to reverse-engineer his own payroll. That is the Excel bill. It just does not arrive monthly.
The case for staying on Excel
Excel is not a bad payroll tool. It is a bad payroll tool at scale, which is a different claim. For a ten-person agency paying ten fixed salaries in one state with no overtime, a spreadsheet is faster than any software, costs nothing extra, and bends to whatever odd arrangement you have with your co-founder about her salary.
The honest test is four questions. Under 25 employees? One state? Salary structures that are broadly the same shape for everyone? And is there a second person who could run payroll next month if the first one vanished? Four yeses and you should keep your money.
Most Indian SMEs answer yes to the first three and no to the fourth. That fourth one is the quiet problem.
What actually breaks first
It is never the maths. Excel computes fine. What breaks is everything around the computation.
Attendance comes in from a biometric device as a CSV with a different employee-code format than your sheet uses, so somebody VLOOKUPs it every month and occasionally VLOOKUPs it wrong. Leave balances live in a second file that HR updates when they remember. A mid-month joiner needs a pro-rata calculation that the formula was not written for, so someone hardcodes a number into a formula cell and that cell stays hardcoded for the next eleven months.
Then PF. The wage ceiling moved from ₹15,000 to ₹25,000 on 17 September 2026, the first change since 2014, so every capped-contribution formula written before that date is now wrong. The employee who crosses the ceiling mid-year because of a promotion. The ESI threshold at ₹21,000 gross where an employee exits coverage only at the end of the contribution period, not the month they cross it. Each of these is a rule that software encodes once and a spreadsheet encodes in whoever’s head built it.
Professional tax is where multi-state kills you outright. Karnataka, Maharashtra, West Bengal and Tamil Nadu all run different slabs and different filing calendars. Three states in one Excel file means three parallel logics in one set of formulas, maintained by one person, verified by nobody.
The rework cost nobody budgets
Payroll errors are not one-time costs. They compound in a specific way. The first mistake gets a shrug. The second one makes employees start checking their payslips line by line, and now every month you field five queries that each take twenty minutes to answer because the answer lives inside a formula.
Statutory corrections carry money attached. Late or short PF remittance attracts interest under Section 7Q plus damages under 14B. TDS short-deduction shows up as a Form 24Q mismatch that you discover in the quarterly return, not in the month you made the error, so you are now fixing three months at once. None of this is dramatic. It is just steady.
Put a number on the time instead. If payroll eats two full days a month for one person at a ₹9 lakh salary, that is roughly ₹72,000 a year of that person’s cost, before any error. Software at 30 employees runs ₹14,400 to ₹32,400 a year.
Where software genuinely does not help
Vendors will not tell you this part. HR software does not fix a bad salary structure, and it does not fix undecided policy. If your leave policy is “ask the founder”, a leave module will not save you, it will just make the ambiguity visible in a dashboard.
Migration also is not free. Expect two to three weeks of real work: cleaning master data, deciding what your policies actually are, running one parallel cycle where you process payroll in both systems and reconcile the difference. Skip the parallel run and you will find out in month two that the opening leave balances were wrong. The easiest window for Indian companies is April to June, right after financial year closure, when your books are already reconciled.
And if you have 40 employees but all of them are on identical fixed salaries in one city with no attendance-linked pay, you can probably push Excel further than this article suggests. Complexity, not count.
The 2026 wrinkle
The four labour codes came into force in November 2025, and the piece that matters for payroll is the 50% wage rule: basic pay has to be at least half of total wages. For companies that had spent years pushing basic down and allowances up to keep PF and gratuity liability low, that reversed overnight. PF, gratuity and leave encashment liabilities all rise when basic rises.
Alongside that, the Income Tax Act 2025 took effect on 1 April 2026, which changes TDS mechanics and Form 24Q handling.
Both of these landed on every Indian employer at once, and both of them mean your payroll logic from FY 2024-25 is stale. If that logic lives in a spreadsheet, somebody has to go rewrite the formulas and hope they caught every case. This is the single strongest argument for software in 2026: the rules moved, and software vendors update the rules for you.
Before you restructure anything, run your actual salary bands through the gross to CTC calculator to see what a 50% basic does to your CTC-to-net gap, and check the PF impact in the EPF calculator.
Picking the tier, not just the tool
Once you have decided to move, the next fork is payroll-only versus a full HRMS. Payroll-only computes salaries, generates payslips and files your statutory returns. A full HRMS adds leave, attendance, onboarding and employee self-service, and costs roughly two to three times as much.
Companies coming off Excel at 30 people almost always want payroll-only first. The reconciliation pain that justifies a full suite shows up later, usually somewhere between 75 and 150 employees, when one HR person can no longer eyeball attendance. The cluster guides here work through that: 50 employees, 100 employees, 200 employees.
greytHR runs payroll, payslips and full HR (attendance, leave, onboarding) plus PF, ESI, PT and TDS compliance for Indian teams, from small companies to 1,000+ employees.
What to keep in Excel even after you switch
Nobody tells you this either. Keep Excel for modelling. Offer negotiations, appraisal budget scenarios, what-if on a 12% versus 15% hike pool, headcount planning. Spreadsheets are excellent at questions with no fixed answer.
What you stop doing in Excel is the part with a legally correct answer and a filing deadline. Payroll computation, statutory contributions, payslip generation. Those have one right answer and a penalty for getting it wrong, and that is exactly the work you want a system to own.
If you are still sizing the decision, model a few real offers in the take-home salary calculator and see how much of the CTC-to-net gap your current sheet is actually getting right.