Multi-entity, properly
Ask to see two entities with different PF codes running payroll in the same month, and a report across both. "We support multi-entity" sometimes means one database with a company column in it.
The difference between an HR system for two hundred people and one for five thousand is not size. It is entities, permissions, audit trails and the twelve other systems it has to talk to. This is what to look for, and what to be careful of.
Enterprise HR software is the system a large organisation runs its people records, payroll, attendance, hiring and performance on, across more than one legal entity and usually more than one location.
The word doing the work in that sentence is "entity". A company with 300 people in one registered business has a bigger version of the problem a 30-person company has. A company with 3,000 people across four registered entities, two states and a subsidiary has a different problem: the same employee record has to mean different things to payroll, to statutory filing and to the person who approves their leave.
That is why enterprise HRMS platforms cost more and take longer to put in. Most of what you are paying for is not features. It is the ability to say who can see what, to prove afterwards who changed what, and to hand clean data to the finance system without anybody exporting a spreadsheet on the 30th.
A tool built for 200 people is not a worse version of one built for 5,000. It is built for a company where one person can still hold the whole picture, and it is usually better at being that. These are the points where the two genuinely part.
Every vendor demo shows the same dashboard. These are the questions that separate them, and most of them do not come up unless you ask.
Ask to see two entities with different PF codes running payroll in the same month, and a report across both. "We support multi-entity" sometimes means one database with a company column in it.
Not three fixed roles. You will need a regional lead who sees one state, a payroll team who sees salaries and not performance, and a manager who sees their own reports and no one else.
Who changed a salary, when, and what it was before. You will need it for an internal audit long before you need it for a dispute, and it cannot be added afterwards.
Logging in through your own identity provider, and joiners and leavers flowing both ways. Without it, offboarding is a manual checklist somebody will eventually forget.
Ask for the docs, not a promise of an integration. An HRIS that cannot be read from is one you will be exporting out of for the next five years.
Professional tax and labour welfare fund differ by state and change. Ask who updates them and how fast, because the answer is either a product team or you.
Which country, under whose law, and what happens to it when the contract ends. Your legal and security teams will ask, so ask before they do.
How many of your people, for how long, and what they stop doing meanwhile. The licence is rarely the expensive part of an enterprise HR project.
Each entity keeps its own PF and ESI codes, its own payroll calendar and its own filings. Reporting reads across all of them without an export.
Build the role you actually have rather than picking the closest of three. Salary visible to payroll, performance to the manager, documents to neither.
Field-level history on the records that matter, kept from day one, so an audit is a search rather than a project.
SSO, directory sync, and an API with documentation. It feeds finance and the service desk rather than sitting beside them.
An AI helpdesk answers leave balances, payslips and policy. At five thousand people that is not a convenience, it is a headcount.
One entity in production while the next is still being configured. Nobody has to hold their breath for a single cutover weekend.
Employee records, salary structures, leave balances and reporting lines out of whatever holds them now. This is the stage that overruns, every time, and it overruns because the old data is messier than anybody remembered. Budget for it honestly and the rest of the plan holds.
Entities, roles and approval chains. Worth settling properly at this point, because changing a permission model after two thousand people are already using it is a different job with a different name.
The new system and the old one for the same month, compared line by line. It is a dull fortnight and it is the only thing that turns confidence into evidence before you switch anybody over.
Start with the smallest entity so the mistakes are cheap. What you learn there makes the second one faster and the fourth one boring, which is what you want a go-live to be.
The questions that come up in an enterprise HR evaluation.
It is the system a large organisation runs employee records, payroll, attendance, hiring and performance on, across more than one legal entity and usually more than one location. What separates it from a smaller HRMS is not the feature list. It is multi-entity payroll, role-based access, an audit trail on changes, single sign-on and an API that lets it feed the other systems a large company runs.
Headcount matters less than structure. The usual trigger is a second legal entity, because that is the point where one payroll run becomes two sets of filings. The others are operating across states with different professional tax rules, needing regional HR leads who should not see each other's data, or an audit that asks who changed a salary and when. A single-entity company of 800 people can be perfectly happy on a mid-market system.
Per employee per month, and the licence is usually the smaller half. Budget for the rollout as well: data cleaning, configuration, parallel payroll runs and your own people's time. A vendor who will not talk about implementation effort is quoting you half the price. Ask what a comparable customer actually spent in staff time, not what the project plan said.
Weeks to months, and the range is that wide because it depends almost entirely on the state of your existing data. Clean records in one system move quickly. Fifteen years of records across three systems and a lot of spreadsheets do not. Going live entity by entity rather than everything at once shortens the risk even when it does not shorten the calendar.
In practice the three are used interchangeably and vendors pick whichever sounds closest to what they sell. If you want the textbook split: an HRIS is the record system, an HRMS adds the operational work like payroll and attendance, and HCM adds the talent side such as performance, learning and succession. Do not choose on the label. Ask which of the three sets of things the product actually does.
It should, and this is the question worth testing rather than asking. In a demo, ask to see two entities with different PF codes running payroll in the same month, and then a headcount report across both. Some products model entities properly. Others have one database with a company field in it, which works until the filings differ.
Ask for the API documentation rather than a yes. A documented API and a sandbox mean your team can answer the question themselves. A named integration on a website means somebody built one once. The practical test is whether payroll output can reach your general ledger without a person exporting a file each month.
Role-based access down to the field, an audit trail on record changes, SSO against your identity provider, and encryption in transit and at rest. On residency, ask us where your data will sit and what happens to it at the end of the contract, and get the answer in the agreement rather than in an email. That is worth doing with every vendor you shortlist, including us.
The one with the odd payroll calendar or the permissions nobody can untangle. That is a more useful hour than a tour of the dashboard.