Finance and HR report different headcount numbers because they are reading different organizational structures, and those structures were built to be different. The supervisory hierarchy answers who reports to whom. The cost center hierarchy answers whose budget pays for someone. The legal entity hierarchy answers which company employs them. A single person can occupy a different position in all three.
Neither function is making an error, and in most organizations neither one is even arguing. They are measuring the same workforce through different instruments, and the fragmentation is severe enough that they often cannot see they are looking at one thing. There is no agreed external definition to arbitrate between them, so the difference surfaces late, in a board pack, as a discrepancy nobody can explain.
Why do finance and HR report different headcount numbers?
This is a design property of the systems, not a data quality failure, and the clearest evidence sits in the product documentation of a widely deployed enterprise HCM.
Workday's own datasheet for organization management describes "a few organizational structures you can create." Three of them are directly relevant here, and the datasheet lists them as separate things:
- Supervisory, to "design and manage your reporting relationships"
- Company hierarchy, to "manage your organization across reporting legal entities and create roll-up structures"
- Cost center and cost center hierarchy, to "use financial cost center information to manage the organization"
The list continues with geographic hierarchies, custom organizations, and matrix organizations, the last described as representing "collaborative organizations, dotted-line relationships, and helix management styles."
Three structures, one person
An HCM system is not confused about this. It is doing exactly what it was configured to do. The structures are independent by design because the questions they answer are genuinely different, and collapsing them would break something real. A shared services analyst may report to a functional lead, sit in a legal entity in another country, and be charged to the budget of the business unit she supports. Every one of those facts is true. Counting her produces a different answer depending on which structure you count through.
The matrix case is where it compounds
Matrix organizations make the divergence structural rather than occasional. When an organization deliberately represents dotted-line relationships, a person has more than one legitimate reporting position at the same time. The supervisory count and the cost center count are then guaranteed to differ for that population, and the size of the difference is a function of how matrixed the company is, which nobody tracks either.
One caution about this evidence. The Workday datasheet is product literature describing configurability. It establishes that these are distinct structures. It does not claim they produce conflicting numbers, and it should not be quoted as if it did. The conflict is an inference from the architecture, and a reasonable one, but the architecture is the citable part.
Is there an agreed definition of headcount?
No. The absence is documented at the level of securities regulation, which is as close to an external standard as this gets.
In a recommendation approved on September 21, 2023, the SEC's Investor Advisory Committee stated that "the SEC's current headcount requirement has led to inconsistent disclosures." The passage that follows is specific about how, relaying an observation the Commission itself made in a 2016 Concept Release that headcount data disclosed by companies "may vary substantially." For example:
some companies report the number of full-time and part-time employees overall, while others report headcount by business unit or division. Still others report numbers only for their domestic workforce, and others report information that includes international employees.
The recommendation also records that inconsistencies in reporting contingent labor persisted despite 2008 staff guidance, and that disclosures made after the 2020 Regulation S-K amendments "continue to reflect these same problems."
Read that against the internal problem. If public companies cannot produce a consistent definition of headcount for the SEC, with securities counsel involved and liability attached, then two departments inside one company have no external standard to appeal to. They are not failing to apply a rule. There is no rule.
Which is why this rarely presents as a dispute. A dispute would at least mean both sides had noticed. What usually happens is quieter: each function produces a number that is correct within its own system, nobody compares them line by line, and the gap only surfaces when someone puts both on the same slide.
The regulator tried, then stopped
This is the part most coverage misses, and it changes the outlook.
Item 101(c)(2)(ii) of Regulation S-K, the human capital disclosure requirement adopted in the 2020 modernization rule, remains in force unchanged. The amendment history for 17 CFR 229.101 still ends at 85 FR 63759, dated October 8, 2020.
The follow-on effort is dead. Under RIN 3235-AM88, the Division of Corporation Finance was, in its own words, "considering recommending that the Commission propose rule amendments to enhance registrant disclosures regarding human capital management." That item was formally withdrawn from the Unified Agenda on April 21, 2025. Note what it was: a division-level consideration of whether to recommend a proposal. It never reached a proposed rule, and it has now been abandoned.
The principles-based 2020 rule is the whole of the requirement. Academic work points the same direction: Bourveau, Chowdhury, Le and Rouen, examining US human capital disclosures from 2017 to 2024, conclude that "disclosure remains highly heterogeneous across firms and topics, with limited standardization despite the regulatory mandate," and that "principles-based regulation alone may be insufficient to achieve comparability."
Nobody is coming to define this. Organizations that want a defensible workforce number will define it themselves or not have one.
How large is the gap? Nobody has measured it
We went looking for research quantifying headcount disagreement between finance and HR inside a single organization. Practitioner sources, academic databases, analyst coverage. We did not find a single study with a disclosed sample and a stated methodology.
What we found instead was a set of confident, specific-sounding figures circulating across vendor marketing with no traceable origin. One widely repeated claim holds that roughly 60% of finance and HR leaders report friction from inconsistent metric definitions. We could not trace it to any study, and we are not repeating it as evidence.
That absence is the finding, and it is worth sitting with. A structural problem that every operator recognizes has no measurement behind it. For a company that argues organizations should stop reporting sentiment as performance, the consistent position is to name the gap and decline to fill it with a number we cannot source.
What each system can and cannot answer
| System | The question it answers well | What it cannot resolve |
|---|---|---|
| HCM supervisory hierarchy | Who reports to whom, today | Whose budget carries the person |
| ERP cost center hierarchy | What the workforce costs, by budget owner | Who actually manages the work |
| Legal entity hierarchy | Which company employs someone, for compliance | Either of the above |
| Payroll | Who was paid, in what period | Contingent labor outside payroll |
| Contingent workforce systems | Non-employee labor, where one exists | Whether that labor is in anyone's headcount |
| Observed work signals | What the person actually produced, across tools | Employment status and cost, which come from the systems above |
No row is sufficient. The reconciliation is not a matter of picking the authoritative system, because each one is authoritative for its own question and none is authoritative for the others.
What a shared number actually requires
The instinct is to declare one system the source of truth. That fails, because the disagreement is not about accuracy. It is about definition.
Four things are required instead:
- A written definition, produced before the number. What counts, what does not, and how contingent labor is treated. A definition reconstructed after two teams disagree is a negotiation, not a standard.
- An explicit mapping between hierarchies. Supervisory, legal entity and cost center, so a person who sits in different places in each resolves to one record rather than three.
- Observation of the work itself, read from the systems where work happens, so the reconciled number can be checked against something other than another self-reported system.
- A disclosed confidence level. A reconciled figure built on partial data is an estimate, and an estimate that says so survives scrutiny better than one that does not.
Be clear about which of those four any measurement layer can supply, including ours. The first two are organizational decisions. Nobody can write your headcount definition for you, and no product can map your hierarchies to each other without you deciding what the mapping should be. Those are governance, not software.
The third is where an observation layer contributes. Levos reads behavioral signals from the tools where work happens, through customer-authorized connections, and applies what it calls controlled cohort analysis with confidence scoring: comparing adopting teams to non-adopting teams while controlling for tenure, role and tool stack, without claiming full attribution.
What that gives you on this problem is narrower than a reconciliation and more useful than another opinion. It is an independent read on what a team actually produced, which does not depend on how any hierarchy was configured. When the supervisory count and the cost center count disagree about who belongs to a team, the observed work is a third input that neither finance nor HR authored.
What it is not. Levos connects to HRIS, communications, productivity, engineering and CRM systems. It does not connect to the ERP or to payroll, so it cannot reconcile a cost center ledger against an HCM record for you. Anyone claiming a product does that end to end should be asked which systems it reads. The honest scope is: the definition and the mapping are yours, and the observed work is the check.
Frequently asked questions
Why do finance and HR report different headcount numbers? They read different hierarchies. Workday's own documentation lists supervisory, company and cost center hierarchies as separate structures. Someone can hold a different position in each, so both counts can be correct and different simultaneously.
Is there an agreed definition of headcount? No. The SEC's Investor Advisory Committee found in September 2023 that the Commission's headcount requirement "has led to inconsistent disclosures," with companies varying on part-time, domestic-only and contingent labor.
Is the SEC going to fix it? No. Item 101(c)(2)(ii) is unchanged since October 2020, and the rulemaking that would have tightened it was withdrawn on April 21, 2025.
How big is the gap? Unmeasured. We found no study with a disclosed sample. The figures circulating in vendor marketing have no traceable origin and should not be relied on.
Which system should be the source of truth for headcount? None of them alone. Each is authoritative for its own question. Reconciliation requires a written definition, an explicit mapping between hierarchies, and a stated confidence level.
The next step
The CFO and the CHRO are not arguing about a number. In most companies they are not arguing at all. They are each reporting correctly from a system built to answer a different question, with no external standard to reconcile them and no regulator coming to supply one, and the fragmentation is complete enough that the mismatch is invisible until something forces the two views onto one page.
That is a solvable problem, but not by winning an argument nobody is having. It is solved by writing the definition down, mapping the hierarchies to each other, and checking the result against observed work rather than against another system that also depends on how it was configured. The related question of how finance should frame workforce investment is covered in how CFOs should actually measure AI investment.
Request a Demo to see what an observed read on team output looks like against your own stack. Levos is accepting design partner applications from US organizations of 150 employees or more with an active AI rollout.
