How Many People Work Here? It Depends on Why You’re Asking
What a headcount reconciliation taught me about Finance, People Analytics, and why making the numbers match isn’t always the point
I scheduled a meeting with Finance because I was building a headcount forecast and didn’t want to accidentally create a second version of the company’s future. I already had the People Analytics side of the equation: Workday actuals, historical hires and departures, workforce composition, and enough movement data to build scenarios around where headcount and representation could go next. Finance already had a headcount forecast of its own, so before I got too attached to mine, I wanted to understand theirs. What was feeding it? What did they know that I didn’t? Where should the two connect? I assumed the responsible ending to this story would involve reconciling everything until we had one clean number sitting in the middle that everyone could agree belonged to the company.
About halfway through the conversation, that stopped sounding like the goal.
Finance knew about people who didn’t exist in my data yet. Planned hires, future backfills, known departures, budget decisions and other workforce changes could already be represented in their forecast before those changes ever became employees or transactions in Workday. Some of that information was confidential for perfectly reasonable reasons. Their official reporting also followed financial close, while People Analytics could be looking at fresher workforce data from a different point in the month. Even the way we grouped the organization could differ. Finance might organize employees around budget ownership and executive functions while I was looking at divisions and structures that made more sense for workforce analysis.
Then there was my side. If someone wanted to know who actually worked at the company right now, who had joined or left, what the workforce looked like demographically, where turnover was concentrating, how tenure was distributed, or what was happening inside different employee populations, I had the stronger view.
We were both talking about headcount. We just weren’t always talking about the same headcount.
Headcount Gets Weird Fast
Headcount sounds like one of those metrics a functioning organization should be able to settle before lunch. How many people work here? Count them. Done.
Spend enough time with workforce data and you realize the question is almost offensively underspecified.
When? That alone can practically ruin your afternoon.
Are we talking about everyone employed this morning, the population as of the last day of the month, or the population in the last financially closed period? If someone terminates on the final day of June, do they count in June or July? Are people on leave included? Contractors? What about an accepted hire who starts next month? A role Finance knows will be filled but Recruiting has not opened yet? A departure leadership knows is happening three months from now? Is this actual headcount, approved headcount, budgeted headcount, forecasted headcount, or some combination somebody inherited years ago and now treats with the caution normally reserved for unexploded ordnance?
And whose organizational structure are we using? The same employee can sit in one division in Workday, roll into a different financial function for planning, and be discussed under yet another business grouping in an executive meeting. Nobody has necessarily done anything wrong. Sometimes the grouping changes because the job the number is doing has changed.
Say one report shows 13,418 employees and another shows 13,421. Our first instinct is usually to ask which one is wrong. Reasonable. But if one is an active Workday population as of Tuesday and the other is Finance’s closed-month number under a different month-end termination rule, we could spend hours hunting three phantom employees who are not missing at all.
The missing thing is context.
That changed what reconciliation started meaning to me. I used to think success meant finding the discrepancy, fixing it, and watching one number disappear. Now I want to know something first: were these numbers ever supposed to match?
Before We Fix the Difference
I want to be careful with the “both numbers can be right” idea because it can become a very convenient hiding place for bad data.
If Finance and People Analytics both tell me they are reporting active regular employees as of June 30 using the same agreed definition and one number is materially different from the other, that is not an interesting philosophical disagreement. Something is wrong.
A duplicate is still a duplicate. A bad join is still a bad join. A broken definition does not become nuanced because two teams happened to produce it independently. If we are answering the same question for the same population, under the same rules, at the same point in time, the numbers should reconcile.
What surprised me was how often teams begin the reconciliation before confirming any of that.
We see two values with the word headcount sitting beside them and assume they are competing answers to one question. Sometimes they are. Sometimes one is answering, “Who actually works here today?” while the other is answering, “What workforce are we financially expecting to carry by year-end?” Those numbers are connected, but there is no reason they should always be identical.
That’s also why I still believe in the idea behind a single source of truth. Anyone who has watched several teams maintain their own slightly different versions of the same business metric knows why the concept exists. We need agreed definitions. We need ownership. We need to know which system is authoritative for what. “Active employee” should not quietly acquire three meanings because three dashboards were built by three people at three different times.
What I’m less convinced by is the idea that one source of truth should produce one universal headcount number that follows us into every conversation.
There’s an old idea in data-quality research that I now appreciate: fitness for use. Wang and Strong argued that good data isn’t only accurate. It also has to be relevant, timely, understandable and appropriate for the task someone is trying to perform. That sounds almost obvious until you put two perfectly respectable numbers on a screen and ask which one the company should use.
A completely accurate Workday active headcount could still be the wrong number for a Board conversation about expected year-end workforce cost. A completely valid Finance forecast could be nearly useless for understanding how racial representation is changing inside a division today. Neither number becomes better by forcing it to answer the other question.
Sometimes the number is fine. You changed what you were asking it to do.
What Are You Actually Asking Me?
That may be the question I care about most now when someone asks for a workforce number.
Get Whitney Hosea’s stories in your inbox
Join Medium for free to get updates from this writer.
What are you actually trying to know?
Finance may be trying to understand the workforce the business is financially expecting to carry. People Analytics may be trying to understand who is actually here and what is happening to that population. Talent Acquisition may care about the workforce the company is trying to acquire. Workforce Planning may be thinking about what the business will need under a different operating plan. DEI may want to know what happens to representation over the next year if current hiring and attrition patterns continue.
We can call every output from those questions “headcount” and then act personally offended when the numbers refuse to cooperate.
Once the question is clear, the ownership gets clearer too. If you want the current employee population, I know where I’m going. If you want the official year-end financial forecast, I know who owns it. If you want to understand how that forecast could affect workforce composition under different hiring or attrition scenarios, now Finance and People Analytics have a reason to bring their numbers together.
That is a very different exercise from deciding which team wins.
It also makes all of the boring metadata around a metric suddenly much less boring. Source. Owner. Population. Effective date. Timing rule. Inclusion and exclusion logic. Actual versus forecast. Assumptions. Intended use.
You do not need a twenty-page governance manual every time someone says headcount. But the number should be able to travel through the organization without forgetting what it is.
That, to me, is the more interesting future state.
A number gets copied into a deck, sent into Slack, dropped into a spreadsheet, repeated in a meeting, and eventually reused by someone who has no idea where it came from. By then, “13,418 active employees as of June 30 under Workday rules” may have quietly become “our headcount,” full stop. The number survived the journey. Its meaning didn’t.
I would much rather build an environment where the context travels with it.
The Forecast I Almost Didn’t Need
My original instinct was still to see how closely I could recreate Finance’s forecast on the People Analytics side. I had actual headcount. I had historical flows. I could model attrition, hiring and different scenarios. Technically, I could produce something that looked convincing.
The problem was that Finance had information my model could not know.
A Workday-based forecast could not infer a confidential planned hire Finance already knew about. It could not reproduce a backfill assumption agreed with a budget owner. It could not know a future departure that had not yet become an HR transaction. I could make assumptions to compensate, but at some point I would be constructing estimates around things another function already knew more accurately.
Which raised a slightly uncomfortable possibility: I was perfectly capable of building something I did not need to build.
That ended up being one of the more useful conclusions from the whole project.
Finance should own the official financial headcount forecast because Finance owns the planning process, budget assumptions and executive information required to maintain it. That does not leave People Analytics with the leftovers. It lets us do something different.
If Finance says the workforce is expected to land at a certain size, I want to know what that might mean for the workforce itself. Which divisions are changing? What movement would have to happen to get there? What could different hiring patterns mean for representation? Where might attrition materially change the picture? How sensitive is the outcome to different scenarios? What is already happening in the current workforce that the topline forecast cannot explain?
That’s much more useful to me than building a competing forecast simply so People Analytics can say we have one too.
It also changed how I think about analytics ownership. Teams spend a lot of time establishing which dashboards, datasets and metrics are officially theirs, and for good reason. Clear ownership prevents duplicated work and endless debates over definitions. But ownership can become territorial before anyone notices.
Sometimes understanding the number well enough to know that someone else should own it is part of doing the work well.
What I Want Reconciliation to Mean Now
None of this made me care less about reconciliation. If anything, I care about it more because I’m clearer on what I’m trying to reconcile.
If Finance and People Analytics are describing the same population, for the same period, under the same definition, I want those numbers to tie. If they don’t, I want the variance explained and the actual problem fixed.
If the purpose is different, I want the difference preserved and labeled.
A Workday actual should look like a Workday actual. A financially closed headcount should tell me which period was closed. A Finance forecast should carry Finance’s assumptions and ownership. A People Analytics scenario should remain a scenario instead of quietly becoming “the forecast” because somebody copied it into an executive deck three meetings ago.
The goal isn’t several teams walking around with their own preferred realities. It’s knowing which numbers should agree, which ones should not, and why.
That distinction matters because the cost of getting it wrong isn’t just an ugly reconciliation file. Analysts burn hours chasing differences that are completely legitimate. Leaders lose confidence because two slides show different values without enough context to explain why. Teams recreate work that already exists. Forecasts built for one decision get reused for another. Eventually the organization stops debating what the number means and starts debating whether it trusts the people producing it.
I walked into that Finance conversation expecting a clean reconciliation to end with one number.
Now, before I try to fix a headcount variance, I want to know whether I’m looking at a broken number or two perfectly good answers to different questions.
Sometimes they need to match, but sometimes the more important work is understanding why they don’t.
