Day four of close. Fourteen countries. The consolidated register is open on your second monitor and it does not tie to the general ledger by €61,400, almost all of it Germany.
You know the German number is wrong because your controller in Frankfurt says net pay moved on 39 people and nobody can name the reason. So you email the provider’s service desk. The desk opens a ticket. The ticket routes to a regional coordinator, who forwards it to the in-country partner in Düsseldorf. The partner replies next morning, Düsseldorf time, asking you to clarify the question. It is now day six. The Beitragsnachweis is due on the fifth-last working day of the month, the contribution payment two working days after that.
Nobody in that chain works for you. Nobody in that chain works for each other either. They are separate companies with separate contracts, separate incentives and separate definitions of a finished payroll.
This is not a failure of the model. It is the model.
Fourteen templates, fourteen cut-offs, one consolidated month
The pitch for a payroll aggregator is one contract, one dashboard, one invoice. Behind the dashboard is a network of local providers, each already running payroll their own way before your name appeared on a subcontract.
Germany wants a spreadsheet with 41 columns and a specific header row. Poland wants a different spreadsheet, and treats a mid-month leaver as a separate submission rather than a line item. Brazil wants inputs six working days before pay date because the eSocial and DCTFWeb sequence gives them no slack: FGTS on the 7th of the following month, INSS on the 15th, income tax withholding on the 20th. France runs to the DSN, due on the 5th of the following month for employers with 50 or more staff and the 15th for smaller ones, so your French entity’s calendar changes when it crosses 50 heads. India files Form 24Q quarterly, on a fiscal year that is not your group’s.
Every one of those calendars is a real statutory constraint. The problem is that flattening them would mean owning the calculation, which is the one thing a coordination layer does not do.
Then there is the vocabulary. A “car allowance” in one country’s file is a taxable benefit-in-kind, in another a gross pay element, in a third a reimbursement that never touches the tax base. Deloitte’s 2025 payroll benchmarking survey, across 15 multinationals with 25,000 to 240,000 employees, found 47 percent maintain more than 1,000 active pay codes and 84 percent have more than 20 payroll system integrations. None of those codes were designed together.
Reconciliation is where this becomes a finance problem. Your consolidated cost-of-labour number is assembled from 14 output files that arrive over five or six days, in different formats, with different element taxonomies, some converted at a rate the provider chose rather than the one your treasury uses. Someone maps them. Usually in Excel. PayrollOrg’s 2025 Getting the World Paid survey of 585 respondents found only about one in five organisations reach the 80 percent process standardisation benchmark, and 38 percent do not track payroll performance at all.
That is why the group number is usually a few days behind. It is not slow reporting. Reporting cannot start until the last file lands, and the last file tends to be the same country every month.
The break happens at the third or fourth country
Two countries is a relationship. You learn the German partner’s rhythm, you learn who actually answers in Warsaw, you keep both calendars in your head. Three or four countries is a coordination problem, and coordination problems scale badly. Each added country brings a template, a cut-off, an approval convention and a person outside your time zone. The combinations grow faster than the headcount does.
Alight’s global payroll complexity research, published in February 2024 from a survey of close to 300 payroll professionals, put a number on the cliff. Among organisations operating in a single country, 24 percent had received a payroll-related fine. Among those operating in two to five countries, 67 percent had, the highest rate the study reported for any footprint band.
At two countries you are still doing payroll. At four you are doing vendor management, and the payroll happens somewhere you cannot see. Deloitte’s earlier large-sample benchmarking survey, of more than 750 organisations across 55 countries and sponsored by the APA and GPMI, asked how many managed payroll providers companies use per region, and found the counts highest outside North America. In the 2025 PayrollOrg survey, of the third who use a global payroll provider, 26 percent still run two to five additional providers alongside it, and 48 percent have no single global payroll system.
What is an in-country partner (ICP)?
An ICP is a local payroll bureau contracted by a global provider to run gross-to-net in one country. You are not their client; the global provider is. Their SLA, calculation engine and statutory research belong to their own business, and your name sits in a subcontract you have never read.
None of this is an argument against local expertise. Nobody runs a German payroll without people who know German payroll, in any delivery model, and the good ones are worth what they cost. The question is structural: who owns the calculation, who employs the specialist answering your question, and whose contract you are inside when a number is wrong.
When the German net is wrong, the round trip is the product
Take the €61,400. In a single-engine setup, someone with access to the calculation opens the run, compares the affected employees against the prior period, and finds the changed input or the changed rule. That is a same-day answer.
In an aggregator setup, the coordination layer is the answer. Your question has to be translated into the ICP’s terms, queued behind the ICP’s other clients, answered in the ICP’s working hours, translated back, then checked by a coordinator who cannot verify it because they do not have the engine either. Three hops out, three hops back. Two days is good. Four is normal. Over a German pay period with a filing due on the fifth-last working day, four days is the difference between a correction and a late filing.
The cost is not only the calendar. EY’s payroll error study of 508 US respondents at companies of 250 to 10,000 employees found an average payroll accuracy rate of 80.15 percent and an average cost of $291 per error, direct cost plus remediation labour. At a few hundred errors a month across a group, the arithmetic gets uncomfortable fast.
Here is the part that belongs in your renewal file. The SLA you signed is with the aggregator, and it is almost always a coordination SLA: acknowledge in 8 hours, respond in 24, resolve in 5 business days. The ICP has its own SLA, with the aggregator, and you have not seen it. When the two conflict, the difference tends to get absorbed by managing your expectations rather than the partner’s performance. Deloitte’s 2025 survey found half the participating multinationals have no defined SLA for payroll inquiry resolution at all, and a quarter target 72 hours or more.
One data model matters more than one login
Single sign-on across 14 providers is a convenience. It does not change where the calculation happens, which is the only thing that determines whether your consolidated month is reconcilable.
What changes the month is one data model: one employee record definition, one pay element definition, one period, one approval state, one audit trail. Everest Group’s 2025 Multi-country Payroll PEAK Matrix, which assessed 28 providers, frames the market’s real dividing line as “operational differentiation across self-covered frameworks, hybrid models, and aggregator solutions.” Self-covered means the provider calculates gross-to-net itself.
By that definition, HR Blizz is on the self-covered side. Its gross-to-net engine is native and single: the German net, the Polish net and the Brazilian net are all calculated in the same engine, statutory filings are generated inside the run, and AI checks run twice a period, once against the prior 12 periods before calculation and once against that country’s statutory rules after it. When the German net is wrong, the person answering has access to the calculation, and the consolidated report is a view of the run rather than a reconstruction of 14 files.
Be clear-eyed about the trade-off, because it is real. A native engine has to maintain the statutory rules for every country it covers, itself, forever. Every rate change, every threshold, every new e-filing schema is engineering work that has to land before the first run of the period. An aggregator can add a country by signing a partner in six weeks; a native engine cannot. Depth varies too: the twelfth country is usually better instrumented than the hundred-and-fortieth. Ask which countries are covered natively, and ask for a sample payslip and statutory return for the three countries you actually worry about, not the three the sales team offers.
Margin stacking, and why swapping Poland is harder than it looks
The commercial shape follows the delivery shape. Where a local bureau prices its work and a coordination layer sits above it, both have to come out of the same cost per payslip, and the second one is buying you handoffs rather than calculation. Ask how that split works on your contract. Nobody volunteers it.
Country switching is where the lock-in bites. In theory, if the Polish partner underperforms, you replace the Polish partner. In practice the aggregator selected them, the interface was built to their template, your historical Polish data sits in their system in their format, and your contract is with the aggregator, not with them. Replacing one country means renegotiating the layer above it. Replacing the aggregator means re-implementing every country at once, because the local relationships were never yours.
That is why unhappy multi-country contracts still renew. The switching cost is not the implementation fee. It is having to do all of it at once, while still closing a month.
Three things worth doing before the next renewal
Time the current month yourself. Not the reported cycle time, the real one: input cut-off to the moment your consolidated register ties to the GL, per country. One or two countries will be carrying the whole delay, and that is your negotiating position.
Then ask your provider, in writing, three questions. Which countries do you calculate gross-to-net yourself, and which are subcontracted? What is the SLA between you and the in-country partner in Germany, Poland and Brazil? And when a net is wrong, who has access to the calculation.
The answers tell you what you bought. If more than a couple of countries come back as subcontracted, you are running a vendor-management function with a dashboard attached, and your fourth country is where that starts costing real money. To see what the other model does to your own numbers, talk to the HR Blizz team about a parallel run in the three countries currently making your month late.
The filing dates and thresholds above are general information rather than legal or tax advice, and several of them move every year.
FAQ
Q: What is a global payroll aggregator?
A global payroll aggregator gives a client one contract, one dashboard and one invoice, but subcontracts the actual gross-to-net calculation in each country to local in-country partners (ICPs). Everest Group’s 2025 Multi-country Payroll PEAK Matrix distinguishes these aggregator solutions from “self-covered frameworks,” where the provider calculates payroll in its own engine.
Q: How many payroll providers does a typical multinational use?
PayrollOrg’s 2025 Getting the World Paid survey of 585 respondents found 48 percent have no single global payroll system, and that among those who do use a global provider, 26 percent still run two to five additional providers alongside it. Deloitte’s global payroll benchmarking survey of more than 750 organisations across 55 countries measured managed payroll providers per region rather than per company, and reported the highest counts outside North America.
Q: Why do multi-country payroll problems appear at three or four countries rather than at scale?
Coordination cost grows faster than headcount: each added country brings its own input template, cut-off, approval convention and time zone. Alight’s February 2024 payroll complexity research found that 67 percent of organisations operating in two to five countries had received a payroll-related fine, compared with 24 percent of single-country operations.
Q: What is the trade-off of a native single-engine payroll platform?
A native engine gives one data model, one audit trail and direct access to the calculation when a net figure is wrong, so queries resolve in hours rather than days. The cost is that the provider must maintain statutory rules for every covered country itself, ahead of each period, and coverage depth varies by country, so buyers should verify native coverage and see sample payslips and statutory returns for the specific countries that matter to them.