TL;DR
- The real risk: Global payroll doesn't fail at the calculation. It fails at the calendar, where cut-offs, filing dates and statutory year ends no longer line up across countries.
- When to wait: If you're under three countries, under fifty headcount outside the home market, and the head of finance is still on a first-name basis with each local provider, the current arrangement will hold.
- What has to be true: If you go further, the new setup has to give one named human, in writing, ownership of the cross-country calendar. Without that, the work is a coordination tax nobody has been given a budget for.
- How the options split: Local providers per country, a single global vendor, an aggregator, an EOR, or one in-house system. Each solves a different problem and creates a different one.
- Decision rule: Pick the model that fails least badly in the country where you're weakest, not the one that looks tidiest on a slide in the home market.
- Outcome to expect: A good setup reduces surprises, not workload. If a vendor is promising less work, ask what they have actually moved.
A Friday in the Operations Director's Calendar
The operations director opens a shared folder on the last Friday of a quarter and sees seven files, each named for a country, each with a different colour tab. The UK tab is green because the local accountant has already filed and the payment reference is sitting in the inbox. The German tab is amber because the provider sent a query about a new mid-month joiner, and the answer won't come back until the German office opens on Monday. The US tab is red because a state-level filing that runs on a different cycle to federal has slipped a day and the operations director only just noticed. The fifth country, the one being added next quarter, isn't in the folder yet. Someone is going to drop a spreadsheet into it and hope.
This is the moment where global payroll stops feeling like an extension of what was working and starts feeling like a second job. Most writing on the subject begins with a definition of multi-country payroll and ends with a recommendation of a particular kind of provider. The reader doesn't need either. They need to understand why a setup that has held together across four countries quietly stops holding together at five, and what the work is that's about to land on someone's desk without anyone having asked for it.
The real issue isn't how to calculate net pay in a new jurisdiction. Most providers can do that, and the reader already knows it. The real issue is who owns the calendar, the cross-country sequence of cut-offs, filings, year ends, retroactive adjustments, public holidays and the one country whose tax authority publishes its forms only in its own language and on its own schedule. The fifth country doesn't break the maths. It breaks the calendar.
Best tools for Global HR
When You Genuinely Do Not Need to Act Yet
Most pieces on this subject treat the reader as if they have already decided something is broken. It's worth saying out loud that it often isn't. The trap is acting because an article made the current setup sound fragile, when in fact it's doing its job for the company it was built for. Four honest stages run from "you're fine" to "you're the edge case the article is written for", and the first two of them don't require a project.
Stage one, the current setup is genuinely fine. The company is in three countries or fewer, headcount outside the home market is under fifty, the head of finance knows each local provider by name, and the year end closes inside a quiet two-week window in January with no surprises. Files go out on time, queries are answered by people who pick up the phone, and there's no real pressure to consolidate. Nothing here's broken. Adding a provider comparison to next year's agenda is the right response, not a procurement exercise. A quarter spent reading the market is the right spend. A quarter spent migrating is the wrong one.
Stage two, friction is showing up but it's contained. The company has crossed into four countries, or it has crossed the fifty-headcount mark, and certain people in finance and people operations are now spending a measurable share of their week chasing information across time zones. The work is contained, though. The number of late filings in the past two years is zero, the cost of the friction is known, and the head of finance can still name each provider. This is the stage where a benchmarking conversation makes sense, and where a deliberate decision to do nothing for another year is also defensible. What isn't defensible is a quiet decision to keep going without anyone naming the friction in a budget meeting.
Stage three, real risk has arrived. The company is at five countries or more, or it has at least one country with above fifty headcount, or the local provider in one of those countries has changed hands twice in three years. The work has stopped being contained, and the risk of a late filing, a missed year end, or a miscalculated retroactive adjustment is no longer theoretical. At this stage, continuing as-is is a decision in its own right, and it's a decision somebody should be making on the record. Whether the response is consolidation, aggregation, or an EOR for a specific subset, the response should be a project, not a procurement habit.
Stage four, the edge case the rest of the article is written for. The reader is at five or more countries, headcount outside the home market is well over fifty, the in-house team is small, the head of finance is no longer on a first-name basis with anyone at any of the providers, and there has been at least one near-miss on a filing in the last twelve months. The reader suspects the current arrangement won't survive another year, and the question is no longer whether to change, but how. The rest of this piece is for them.
Five Questions This Reader Asks Themselves at 11pm
Why does this feel harder than it should, given that we already do it in four countries? Because the work that breaks isn't the work that scales. The fourth country added a manageable increment of effort. The fifth country adds a coordination problem that's not proportional to headcount, because the work is in the cut-offs and the filings lining up, and those don't line up by themselves. The right answer to this question is to accept that the curve isn't linear, and that the right response is structural rather than incremental.
Are we really exposed to permanent establishment risk from a remote employee, or is that something a vendor is using to scare us? The exposure is real and the shape of it's set out above in the OECD's November 2025 update to the Model Tax Convention. The time threshold is less than half of total working time, and the commercial-reason test excludes employee preference, retention and internal cost reasons. A reader can think through their own situation against those two prongs. The vendor's job is to describe the shape of the exposure, not to adjudicate it, and neither is this article's. What the reader should walk away with is that the question isn't whether to worry, it's whether to take local advice on their specific facts.
If the provider we are talking to really covers thirty countries, why do they keep asking us to find a local partner for the fifth one? Because payroll coverage in thirty countries is often a network of local partners stitched together under one contract, and the contract isn't the same as the capability. The honest version of the answer is that the provider is telling the reader what their own operating model is. The job of the conversation is to find out whether the provider is the employer of record in those countries, or whether the provider is reselling someone else's payroll, and to be clear about which it's in writing.
What happens if the local provider in the new country misses a filing, and who actually owns that? Whoever the contract says owns it, and the contract is usually less clear than the reader would like. This is the question that pays the most to answer before signing, and it's the question the reader should write into the RFP if they're running one. The honest answer from a provider is a name and a process. The dishonest answer is "we do payroll, so it's covered", which usually means the local partner owns it, and the partner is in a different time zone.
Are we about to choose a model because the finance director likes the slide, or because the work actually fits? The model that looks tidiest on a slide in the home market is the one that often fails least tidily in the country where the reader is weakest. The honest test is to draw the slide with the hardest country's calendar on it, not the easiest. If the model still looks tidy in that drawing, it's the right one. If it doesn't, the reader is choosing a slide, and the cost of that choice is going to land in a quarter when someone is on a call at midnight.
Three Categories the Approaches Split Into
The first category is country-by-country ownership, where each country is its own contract and its own relationship. This is the model most readers at the four-country stage are already running. It's right when the company has the time and the headcount to run it, and when the local providers in each market are people the company would choose again today. It fails when the in-house team is too small to hold the relationships, when the work is becoming a full-time job nobody has been given, and when the calendar across countries stops being something one person can hold in their head. The shape of the failure is that nothing goes wrong on a single country basis, and a great deal goes wrong on a cross-country basis, because no one is in charge of the cross-country view.
The second category is unified ownership, where one provider or one platform owns the payroll in every country, either as the actual employer of record or as the contracting party with local partners in each market. This is the model that looks best on a slide, and it's right when the company is large enough to need a single source of truth, when the readers in finance and people operations are both spending more than a day a week on cross-country coordination, and when the cost of running country-by-country has stopped being a cost the company can absorb in headcount. It fails when the provider is reselling in the harder countries, when the single contract isn't matched by a single operating model, and when the reader is sold "thirty countries" and finds out that in the country they care about most, the actual payroll is run by a partner they have never spoken to.
The third category is hybrid ownership, where the reader runs their core markets directly and uses an employer of record, an aggregator or a regional provider for the rest. This is the model that fits the most companies at the five-country stage, and it's the model that takes the most care to design. It's right when the company has one or two markets where headcount is growing fast or where the local employment law is a poor fit for the home market's model, and when the company doesn't want to incorporate in those markets. It fails when the split between the in-house piece and the outsourced piece isn't drawn in a way the company can describe on one page, when two systems of record end up carrying two versions of the truth, and when the company outsources the country that needs the most attention because the work is hardest there.
Five Diagnostic Questions the Reader Can Self-Assess Against
How many countries are we actually running payroll in, and how many of those countries does the head of finance know by name? The answer to "do we need to act" is a function of the gap between those two numbers. If the head of finance can name every provider, the calendar is in human hands. If not, the calendar is in nobody's hands, and the question isn't whether to act but how.
In the last twenty-four months, have we had a late filing, a missed filing, or a near-miss? The reader can answer this by asking the head of finance and the head of people operations separately. If both can answer without checking, the work is held in their heads. If one of them has to check, the work is held in a folder. If neither can answer, the work isn't held anywhere, and that's the answer.
If a new country joined tomorrow, who would own the cross-country calendar, and what would they do in the first week? The reader should be able to draw an answer with a name on it. If the answer is "we would figure it out", the work isn't owned, and the work is about to grow.
What is the cost of a single late filing in our largest market outside the home country, in penalties, in management time and in the relationship with the local tax authority? The shape of the cost is known to the reader even if the number isn't. The honest answer is that the cost is always larger than the number on the invoice, and the difference is the reason this article exists.
Which country's calendar is hardest, and does the model under consideration make that country easier or harder? The model that makes the hardest country easier is the right one. The model that makes the easiest country easier, while leaving the hardest one as it's, is the wrong one, and the reader should be told that in the meeting where the model is being chosen.
Five Operating Models, Reviewed
Local provider per country
This is the model the reader is most likely running already, and it's the right one for a specific kind of company. The company is small enough outside the home market that each relationship can be held in one person's head, the local providers are people the company has chosen deliberately and would choose again, and the work of cross-country coordination is contained because there's not enough of it to need a project. In this model, each provider runs the payroll in their country, files with the local authorities, and handles the year end locally. The reader holds the cross-country view.
Where this model genuinely falls short is at the moment the reader is at right now, which is the point at which the cross-country view stops fitting in one head. The failure isn't that the providers are bad. The failure is that nobody owns the calendar across them, and the work of holding that calendar together becomes a full-time job that has not been written into anyone's job description. The risk is that the reader runs a good payroll in five countries, files on time in every one of them, and still finds out at a year end that the work was not in fact coordinated, because coordination was never anyone's named responsibility.
A single global payroll vendor
This is the model that's easiest to explain to a board, and it's the right one when the reader needs a single contract, a single source of truth, and a single point of contact for the cross-country work. In this model, one provider contracts with the reader across all countries, either as the actual employer of record in each market or as the prime contractor with local partners underneath. The reader's finance team and people operations team deal with one relationship instead of five.
Where this model falls short is in the gap between the contract and the operating model. A single contract across thirty countries isn't the same thing as a single operating model across thirty countries, and the reader's job in the sales process is to find out which it's. The honest version of the answer is that some countries are run by the provider's own entity, and some are run by a partner the reader won't meet until something goes wrong. The risk is that the reader buys a single contract and inherits a network, and the network is only as good as the partner in the country the reader cares about most.
An aggregator over local providers
This is the model that sits between the other two, and it's the right one when the reader is happy with their local providers but wants a single pane of glass and a single cross-country calendar. The aggregator doesn't run the payroll. The aggregator connects to each local provider, normalises the data, and gives the reader a single view of the work across countries. The reader's local providers keep doing what they're doing. The reader gets a single dashboard and a single reporting layer on top.
Where this model falls short is that the aggregator can't make a late filing on time. If the local provider in one country misses a filing, the aggregator's dashboard will reflect the late filing, but it won't prevent the late filing. The aggregator is a view, not an action, and the reader's job is to be clear about which one they're buying. The risk is that the reader buys a dashboard and discovers, at the moment it matters, that the underlying providers are still the ones doing the work, and the underlying providers are still the ones the reader needs to call.
Payroll inside an employer of record
This is the model that's right when the reader is adding a country where the company doesn't want to incorporate, where headcount is small or where the local employment law is a poor fit for the company's home model. The EOR becomes the legal employer in the new country, runs the local payroll, files the local returns, and handles the local year end. The reader's company directs the work but doesn't employ the worker. This model also substantially reduces but doesn't eliminate permanent establishment exposure, because activity beyond employing staff, such as concluding contracts locally, maintaining a fixed office or holding inventory, can still create one, and the EOR doesn't answer corporate tax.
Where this model falls short is that it's a per-country decision, not a global one, and the reader's job is to draw the line between the markets where the EOR is the right answer and the markets where the reader is better off running their own payroll. The risk is that the EOR becomes the default for every new country, and the reader ends up with a portfolio of EOR relationships that look like one model on a slide and feel like five different relationships in practice.
Running it in house on one system
This is the model that's right when the company is large enough to absorb the cost of building the capability in house, and when the readers in finance and people operations want to own the system of record themselves. The reader's own team runs the payroll in every country, on one system, with their own process. The model removes the dependency on providers and gives the reader full control of the data.
Where this model falls short is that running payroll in house in five countries isn't the same as running payroll in house in one country, and the cost is in headcount, in local expertise and in the calendar across markets. The risk is that the reader builds a system that's excellent in the home market and thin in the markets the home market doesn't know, and the work of filling that gap is what the reader thought they were outsourcing. This model only works for companies large enough to hire local expertise in every market, and most readers at the five-country stage aren't.
The Decision Table
| Situation | Scale | Setup | Primary Pain | Recommended Starting Point |
|---|---|---|---|---|
| Under three countries, in-house team knows every provider | Under fifty headcount outside home market | Country-by-country contracts | None worth fixing this year | Do nothing, benchmark next year |
| Four countries, friction showing up in finance team time | Between fifty and one hundred headcount outside home market | Country-by-country, no aggregator | Coordination tax nobody owns | Run an aggregator evaluation, keep providers |
| Five countries, one country growing fast on its own | Headcount spread unevenly, one market over one hundred | Country-by-country plus one EOR | Integration of the EOR with the rest | Keep core markets, formalise the EOR line |
| Five plus countries, multiple markets with over fifty headcount | Total outside home market over two hundred | Multiple local providers, contracts renewed at different times | Cross-country calendar, late filings | Run a structured evaluation of a global vendor, with a written test in the hardest country |
| Six plus countries, at least one hard market (own entity, local statutory complexity) | Above five hundred outside home market, finance team of more than five | Mixed in-house and outsourced, no single source of truth | Two systems of record, year ends in different quarters | Move to one global vendor or one in-house platform, with a parallel run before cut over |
| One country with a single employee, no entity | One headcount, no incorporation in that market | None | Incorporation cost versus EOR cost | EOR for the single employee, revisit at the threshold for incorporation |
| Two adjacent markets, similar employment law, headcount under thirty each | Total under sixty in the pair | Two local providers, no shared calendar | Friction in the two markets, not in the others | Aggregator on top of the two existing providers, no contract changes yet |
| Most of the workforce outside home market, finance team is the home market team | Majority of headcount outside home country | One global vendor under contract, several countries run by partners | The contract is unified, the operating model is not | Renegotiate the contract to name the partner in each market, or move the country to direct ownership |
The Payroll Calendar Problem, Mapped
The point of this table is to make the calendar visible. Most readers know their own country's calendar in their head. The fifth country is the one whose calendar isn't in their head, and the work of holding all the calendars together is the work that has broken. The table below is a structure, not a country-specific plan, and the reader should fill in the country-specific dates from the local provider.
| Step before the run | Who normally owns it | What goes wrong across countries |
|---|---|---|
| Confirm the headcount and the joiners, leavers and changes for the cycle | Local provider, with input from in-house people operations | In one country the cut-off is two business days before pay date, in another it is five, and a leaver entered on day three in the slower country is paid by mistake |
| Calculate gross to net, including any mid-cycle adjustments | Local provider | Adjustments in one country are applied immediately, in another they wait for the next cycle, and the in-house team has to remember which is which |
| Produce the bank file and obtain internal approval | In-house finance, with sign-off from the country lead | Approval windows sit in different time zones, and a sign-off that arrives during a public holiday in one country arrives too late for the file |
| Send the bank file to the local bank and confirm funding | Local provider, in-house treasury | Cut-offs at local banks differ by hours, and a file sent on a Friday afternoon in the home market is a file that arrives on Monday morning in the hardest country |
| File the statutory return with the local tax or social security authority | Local provider, with sign-off from in-house finance | Filing windows in some countries are monthly, in others quarterly, and the in-house team only finds out a filing has been missed when the local authority sends a letter |
| Close the period and reconcile the payments | Local provider and in-house finance | Year ends in some countries align with the calendar year, in others they align with the tax year, and the in-house team has to hold two year ends in their head at once |
| Handle the next cycle's retro adjustments, for example a backdated pay rise or a corrected tax code | Local provider, with input from in-house people operations and finance | The retro lands in different cycles in different countries, and the in-house team has to remember which country's retro has been applied and which has not |
Questions to Put to a Provider Before You Sign
The point of this table is to draw the line between a real multi-country capability and a reseller relationship dressed up as one. The honest provider will answer these questions on the record. The dishonest one will deflect to a slide. The reader's job is to write the answers into the contract, not into a slide deck.
| Question | What the honest answer looks like | What the dishonest answer looks like |
|---|---|---|
| In country X, are you the legal employer of record, or are you contracting with a local partner? | A clear yes or no, with the partner named if the answer is no | "We have coverage in country X", with no further detail |
| Who owns the local filing in country X, and what is the process if a filing is missed? | A named role, a named contact, a written process and a service-level commitment | "We take responsibility for payroll", which means the partner owns the filing and the partner is in a different time zone |
| Can you show me the local payroll calendar for country X, with the cut-offs and the filing windows marked? | A document, ideally signed by the local partner, with the dates the reader can verify | "Our platform handles the calendar for you", with no document |
| How do you handle a parallel run, and what does it cost? | A named process, a named timeline, a written scope and a clear cost | "We can support a parallel run", with no commitment to the cost or the timeline |
| If I leave, what happens to the data, the local filings, and the employee records? | An exit clause with a transition period, a data export format and a named owner of the transition | "We will support the transition", with no exit clause |
| Who pays the local statutory penalties if a filing is late because of a partner, not because of us? | A named cap, a named process, and a clear line between the partner's fault and the reader's | "We are committed to on-time filing", with no commitment on who pays |
| What is the cost of a mid-year change, for example adding headcount in country Y or moving a worker from one country to another? | A written rate card for the change, a named timeline, and a clear owner | A change request process, with no rate card |
| Who do I call at 11pm on a Friday in the home market if country Z has a payroll emergency? | A named contact in country Z, with a phone number and a service-level commitment | A global helpdesk, with no named contact in country Z |
The Cost of Getting This Wrong
The cost of getting this wrong is rarely the number on the invoice. A late filing in one country carries a penalty that's small enough to write off. The cost that matters is the second-order cost, and it's the cost the reader isn't pricing when the model is being chosen. A late filing in the country where the reader is weakest is a hit to the relationship with the local tax authority, and that relationship is the one the reader will need at the next year end, the next audit, and the next cross-border review. A missed retroactive adjustment in one country is a conversation with the affected employee, and that conversation is a hit to the company's standing with the workforce in that market. A year end that runs late in one country is a year end that runs late in every country that depends on it, and the cost of the domino isn't the first late filing, it's the fifth.
So the cost of getting this wrong is paid in management time, in employee trust, in the relationship with local authorities, and in the credibility of the in-house team that owns the work. The number on the invoice is the smallest part of the cost, and the reader who is choosing on price alone is choosing on the smallest part of the cost. The question the reader should be asking, in the room where the model is being chosen, is which model's failure mode they would rather live with, and which model's failure mode they can absorb without losing the relationship that matters.
When You Are Ready to Go Further
If the reader is at the stage where the current arrangement won't survive another year, and where the work of holding the cross-country calendar is a full-time job nobody has been given, the next step is a structured comparison of the options above, with a written test in the country where the reader is weakest. HROpsLab's independent comparison work is built for that moment. We don't sell software, we don't run payroll, and we don't provide legal or tax advice. What we do is publish the kind of side-by-side evaluation the reader can't build themselves, because the reader is too close to the decision to evaluate the options fairly.
We also publish case studies of companies that have made the change, written with the operators who did the work. The case studies aren't vendor pitches, and the companies aren't named in the marketing sense. They're the closest thing the publication has to a record of what the work actually looks like, and they're written for the reader who is about to start.
Frequently Asked Questions
Can one vendor really cover every country we operate in?
The honest answer is that some vendors operate their own entity in many countries and contract with a local partner in the rest. The reader's job is to find out which model applies in the country they care about most, and to get the answer in writing, with the partner named if the answer is that a partner is involved. A contract isn't the same as a capability, and the difference is what the reader should be testing in the sales process.
What does an aggregator actually do?
An aggregator sits on top of the reader's existing local providers, connects to each one, and gives the reader a single view of the payroll work across countries. The aggregator doesn't run the payroll and doesn't own the local filing. The aggregator is a view, not an action, and the reader's job is to be clear about which one they're buying, and to make sure the underlying providers are still the ones the reader needs to call when something goes wrong.
How should we handle a country where we have one employee?
The honest answer is that an employer of record is usually the right starting point, because the cost of incorporating in a new market for one employee is rarely justified. The EOR becomes the legal employer, runs the local payroll, and handles the local filing. The reader's company directs the work but doesn't employ the worker. The reader should revisit the arrangement when headcount in that market reaches the threshold for incorporation, and the EOR will usually help the reader think through that threshold.
Who owns the error if the local provider calculates something wrong?
The honest answer is whoever the contract says owns it, and the contract is usually less clear than the reader would like. The reader's job is to write the answer into the contract before signing, with a named cap on the provider's liability and a named process for handling the error after the fact. The error is the easy part. The conversation with the affected employee is the hard part, and the contract should make clear who has it.
What do we do if the local provider misses a filing?
The honest answer is that the reader should be told as soon as the local provider knows, and the local provider should own the conversation with the local authority. The reader's job is to make sure the contract requires the local provider to tell the reader in writing, within a named window, when a filing has been missed. The reader should also make sure the contract names who pays any statutory penalty, and the reader should test the process with a scenario before signing.
How do we run a parallel period, and how long should it take?
The honest answer is that a parallel run is a period in which the new provider runs the payroll alongside the existing one, so the reader can compare the two and resolve the differences before the cut over. The length of the parallel run is a function of the country and the complexity of the payroll, and the reader should write the length and the cost into the contract. The reader should also make sure the parallel run covers a year end if the country is one whose year end is different from the home country's, because a year end is where most of the differences show up.
When should we bring payroll in house?
The honest answer is when the company is large enough to hire local expertise in every market it operates in, and when the readers in finance and people operations want to own the system of record themselves. The cost is in headcount and in local expertise, and the benefit is full control of the data. The reader should bring payroll in house only when the cost of the in-house model is less than the cost of the outsourced model plus the cost of the work the outsourced model doesn't do, and the reader should write that calculation out before committing.
HROpsLab: independent, vendor-neutral research for people operations and payroll leaders who would rather choose well than choose fast.
