Headcount Forecasting: Why Approved Hires Keep Breaking the Plan
The hiring plan says 14 open roles.
The forecast says 14 open roles.
Everyone nods, because at least the numbers agree.
Then someone asks when those people are actually starting.
That is usually where the headcount forecast stops being a headcount forecast and reveals itself as a list of approved jobs with salaries attached.
I see this more often than I would expect. A company can have a fairly sophisticated revenue model, a detailed operating plan and a hiring sheet where every open role somehow begins on the first day of a month. Recruiting, apparently, has achieved industrial precision.
It has not.
Headcount forecasting is mostly a timing problem disguised as a payroll calculation.
An approved role is not an employee
There are several dates hiding inside every planned hire.
The date management approves the role. The date recruiting opens it. The date an offer is accepted. The date the person starts. And the date that person is productive enough for whatever operating assumption justified the hire in the first place.
Those dates are not interchangeable.
If Sales gets approval for six account executives in January, Finance cannot quietly translate that into six January salaries and call the model conservative. Maybe two start in February, two in March, one in April and one role gets reconsidered after Q1 bookings come in light.
That difference changes payroll. It changes commissions. It changes recruiting fees, benefits and equipment. More importantly, it changes the operating capacity the business thought it was buying.
This is why I like driver-based forecasting for headcount. The useful driver is rarely “number of approved positions.” The useful drivers are the things that move people from an org chart into the P&L.
The cleanest headcount model starts with people, not departments
A departmental total is useful for reporting. It is not enough for forecasting.
I want to be able to see the people already employed and the positions that do not exist yet as two different populations.
For existing employees, the model should know enough to calculate the run rate: base compensation, payroll taxes, benefits or a benefits assumption, bonus or variable compensation where relevant, and any known changes in timing.
For open roles, I care about a different set of questions:
- What role is being hired?
- Who owns the hire?
- What compensation assumption are we using?
- What start date is in the forecast?
- How confident are we in that date?
- What business assumption depends on this person arriving?
That last question is where the model gets interesting.
A new engineer may be tied to a product milestone. A customer-success hire may be tied to customer volume. A salesperson may be tied to capacity in the revenue forecast.
If the headcount plan changes but those assumptions stay frozen, Finance has updated payroll and left the business plan untouched.
Vacancy is a forecast assumption
Suppose the annual plan includes 20 new hires at an average fully loaded cost of $120,000.
On paper, that looks like $2.4 million of annualized cost.
But the company is not going to hire 20 people at 12:01 a.m. on January 1.
If the average start date slips by 60 days, the current-year expense can move by hundreds of thousands of dollars without anybody “cutting” headcount. The organization chart still shows the same destination. The cash path to get there changed.
This is one reason a favorable payroll variance deserves a little suspicion.
Finance sees payroll under budget and it is tempting to put a green dot next to the line.
Maybe it is good news.
Or maybe the company is three months behind on hiring the people required to deliver the plan.
The accounting result is favorable. The operating result may not be.
That distinction is exactly why variance analysis should not stop at explaining the dollar difference.
I would rather forecast hiring friction than pretend it does not exist
Not every role deserves the same start-date assumption.
A position that is already at final interviews is different from a role that was approved yesterday. A common role with an active candidate pipeline is different from a specialized executive search. Ten planned hires in one function may be harder to absorb than one.
You do not need a recruiting science project inside the financial model. You do need some acknowledgment that hiring takes time.
One practical approach is to assign open roles a forecast status.
Committed: offer accepted or start date known.
Active: recruiting is underway and the expected start date has reasonable support.
Planned: approved in the operating plan but not yet actively recruiting.
Conditional: hiring depends on a trigger such as revenue, funding, customer volume or another operating milestone.
The labels are less important than the discipline. Finance should be able to explain why a start date is in the forecast.
“That is what the budget said” is technically an explanation. It is just not a very useful one.
Headcount should move when the business moves
The annual budget has a bad habit of turning hiring assumptions into historical artifacts.
A manager requested three analysts during planning. The roles were approved. Six months later, demand has changed, priorities have moved and one of those jobs is still sitting in the forecast because deleting it feels suspiciously like reopening the budget.
This is where a rolling forecast earns its keep.
The question each cycle is not whether the company is still “allowed” to hire the role.
The question is whether the current outlook still supports the timing and need.
That can work in both directions.
A stronger sales pipeline may justify pulling a customer-success hire forward. A delayed product launch may push an implementation hire back. A cash constraint may turn a planned role into a conditional one. An unexpected resignation may create a vacancy that needs to be separated from growth hiring.
The headcount model should be allowed to notice reality.
Separate replacement hiring from growth hiring
This is a small modeling choice that makes management conversations much better.
If a company plans to grow from 100 employees to 115, it is easy to assume there are 15 hires in the plan.
Then 12 people leave during the year.
Now recruiting may need to make 27 hires just to end at 115.
Those are very different workloads, and they tell management different things.
Replacement hiring preserves existing capacity. Growth hiring adds capacity. Backfills may have different compensation than the people who left. Some positions may not be replaced at all.
Net headcount alone hides most of this.
I like seeing beginning headcount, planned additions, expected attrition or known departures, replacements, deliberate eliminations and ending headcount separately. It makes the bridge understandable without turning the model into an HR database.
The model needs an owner outside Finance
Finance can calculate the cost. Finance should challenge timing. Finance can reconcile actual employees to payroll and keep the forecast honest.
But Finance should not be inventing hiring dates alone.
Managers know whether the role is still needed. Recruiting knows whether a candidate pipeline exists. HR knows compensation, start dates and departures. Finance knows what those changes do to the plan.
The headcount forecast gets better when those facts meet in one place.
It gets worse when Finance sends a spreadsheet around once a quarter asking everyone to “please update your rows by Friday.”
That process usually produces a remarkable number of Friday afternoon start dates.
What I want the CFO to be able to see
A good headcount forecast does not need to be beautiful.
It needs to answer a few questions without a forensic investigation.
How many people do we have now? How many are we actually expecting to add? Which hires moved since the last forecast? What did that do to payroll and cash? Which operating assumptions changed because the hiring timing changed?
And one more:
Are we under budget because we became more efficient, or because we failed to hire the people the plan assumed would be here?
Those two stories can produce the same favorable variance.
Only one of them deserves the green dot.








