Operating Expense Forecasting: Stop Treating Every Cost the Same

The expense forecast often looks wonderfully consistent right up until you ask how the lines were forecast.

Rent: prior month.

Software: prior month.

Travel: prior month.

Marketing: prior month plus a little.

Four completely different costs. One forecasting method.

It is tidy. It is also how a lot of expense forecasts become historical reports with future dates.

Operating expenses do not all behave the same way, so I do not think they should all be forecast the same way.

Start by asking what makes the cost move

Before choosing a formula, I want to know what creates the expense.

Payroll moves with people, compensation and timing. Rent moves with leases. Payment processing may move with transaction volume. Cloud infrastructure may move with usage. Travel moves with activity and policy. Marketing programs can move because someone deliberately decides to spend the money.

Those are different forecasting problems.

A driver-based forecast does not mean every GL account needs a sophisticated operational driver. It means the important lines should have logic that resembles the business.

Some costs deserve a schedule

Known contractual expenses are usually the easiest.

Rent, insurance, major software contracts, debt-related fees and other scheduled commitments often have dates and amounts attached to them.

I would rather forecast those from the agreement than apply a percentage growth rate because last year happened to increase 6%.

This is basic, but it matters because known costs create a useful base. Finance can spend its judgment on the lines that are actually uncertain.

Some costs deserve an operating driver

Payroll is the obvious example, which is why the headcount forecast matters so much.

But there are others.

Transaction fees may follow payment volume. Customer-support tools may follow seats. Freight may follow shipments. Hosting may follow usage. Commissions may follow bookings or revenue depending on the plan.

The driver does not need to explain every dollar. It needs to explain enough of the movement to make the forecast useful.

And some costs are decisions

This category gets mistreated most often.

Marketing campaigns, outside consultants, conferences, recruiting programs and discretionary projects do not necessarily happen because revenue went up 8%.

They happen because somebody chooses to do them.

Forecasting these lines as a percentage of revenue can create a very smooth model and a very strange management conversation.

I prefer to know what has actually been committed, what is planned, and what can still move.

That distinction becomes especially useful in a downside scenario. Management can see which costs naturally move with the business and which require an explicit decision.

Run rate is useful until it isn’t

I use run rate. I just do not worship it.

For stable, low-materiality expenses, recent actuals can be a perfectly reasonable forecasting method.

The problem is when run rate becomes the default because it is easy.

December includes an annual renewal. March includes a conference. June has a legal settlement. A vacant role lowers software seats for two months. Last month is full of things that may not belong next month.

Before annualizing an expense, I want to know whether the recent period represents normal activity.

Sometimes the most valuable forecast adjustment is simply refusing to repeat an anomaly twelve times.

Do not make every department forecast every account

This is where expense planning can become a small administrative state.

Finance sends managers a template with 80 accounts. Managers stare at it. Someone copies last year’s numbers. Finance consolidates the fiction.

Business leaders should own the assumptions they can actually influence.

They usually know hiring plans, major vendors, projects, campaigns and changes in operating activity. They may not have a useful opinion about every accounting classification.

Finance can translate business assumptions into the financial structure without asking managers to become part-time accountants.

Watch timing, not just the annual total

A department can be exactly on its annual expense plan and still create a meaningful cash or earnings surprise by moving spend between quarters.

Timing matters.

A $600,000 implementation that moves from Q2 to Q4 changes the near-term outlook even if the annual budget remains untouched.

This is one reason I prefer a rolling forecast that updates the timing of real activity instead of forcing the budget calendar to remain true forever.

The expense review should sound like the business

If an expense forecast review consists entirely of account numbers and percentages, I usually think we are one level too low.

The useful conversation sounds more like this:

The hiring class moved six weeks. The product team added a contractor. The trade show was canceled. Cloud usage is running above the customer forecast. The insurance renewal came in higher than expected.

Those sentences explain why the financial statement is changing.

Then the model can do its job.

It can turn the business into numbers instead of turning last month’s numbers into next month’s numbers.

I group expenses by behavior before I group them by account

The chart of accounts is designed for accounting.

That does not mean it is the best forecasting taxonomy.

I often find it more useful to think about costs as fixed commitments, headcount-driven costs, volume-driven costs, discretionary decisions and one-time items.

Then I map those assumptions back into the financial statement.

This keeps the model from pretending that two expenses belong together simply because they sit next to each other in the GL.

Forecasting should reflect economic behavior first and reporting structure second.

Payroll deserves its own model

For many service and technology companies, payroll is the largest operating expense by a wide margin.

Forecasting it as “last month plus 3%” is one of those shortcuts that works until it very much does not.

I want employees, open roles, expected start dates, compensation, payroll taxes, benefits and material variable compensation reflected at an appropriate level.

I do not necessarily need every person named in the management model.

I do need hiring timing.

A role that starts three months late does not create the same expense as a role that starts January 2.

Headcount is where annual budgets routinely lie by accident.

Software has become its own little economy

Software expense used to be easier to ignore.

Now a growing company can have hundreds of applications, seat-based contracts, usage-based tools, annual renewals and overlapping systems nobody is quite ready to cancel.

I like separating major contractual software from variable or seat-driven software.

For material contracts, use the renewal schedule.

For seat-based tools, connect the forecast to headcount where that relationship is real.

And periodically ask whether the software still exists because the company needs it or because canceling subscriptions is apparently harder than launching them.

Corporate archaeology can be expensive.

Marketing is not one cost behavior

“Marketing expense” can include people, agencies, paid media, events, software, content, sponsorships and one-time campaigns.

Forecasting all of it as a percentage of revenue may create a reasonable annual number while getting every month wrong.

I want to know the actual program plan for material discretionary spend.

What is committed? What is approved but not committed? What can be delayed? What is expected to scale with leads or revenue?

That distinction becomes very useful when management asks for a downside case and everyone suddenly discovers the marketing line contains five different kinds of flexibility.

Travel is a good example of why history needs context

Last month’s travel can be an awful forecast for next month.

One month has a sales kickoff. Another has a customer conference. Another has almost nothing.

I prefer known events plus a reasonable baseline rather than blindly annualizing the most recent period.

This is not sophisticated modeling.

It is remembering that calendars exist.

Some of the best forecasting improvements are embarrassingly practical.

Accruals can make a good forecast look wrong

Operating expense analysis depends on clean actuals.

If a large vendor invoice is late and the accrual is missing, actual spend looks favorable. FP&A may then lower the run rate and accidentally forecast the same false savings into the future.

This is why I care about the Accounting-to-FP&A handoff.

Before declaring an expense trend, understand whether the accounting is complete.

A favorable variance caused by an invoice sitting in someone’s inbox is not efficiency.

It is mail.

Managers should forecast decisions, not GL codes

I have very little interest in asking a department leader to predict 63 expense accounts.

They will either guess, copy the budget or call Finance.

I would rather ask about the things they actually control.

Are you hiring? Which major vendors are changing? What projects are starting? What travel or events are planned? What commitments have you made? What can move if the business slows?

Finance can translate those answers into accounts.

That is part of the job.

We should not outsource our chart of accounts to Marketing because we sent them a template.

Committed, planned and discretionary are useful distinctions

For management decisions, I like knowing how much of the forecast is already committed.

A signed annual contract is different from a project someone hopes to start in Q4.

A required audit is different from an optional conference.

A filled role is different from an open requisition.

Those distinctions matter enormously in a downside scenario.

Management can see what naturally disappears, what requires a decision and what is difficult to change.

“Cut OpEx 10%” becomes a much more intelligent conversation when the cost base has been classified by flexibility.

Annual budget ownership creates strange behavior

Departments sometimes treat budget as permission rather than an estimate.

If $200,000 remains in December, suddenly $200,000 of ideas appear.

I understand the incentive. People worry that not spending this year means losing the budget next year.

But the forecast should not preserve spend simply because it was once approved.

If the business no longer expects to spend it, remove it from the outlook.

The budget remains available for variance and accountability.

The forecast has a different job: tell the truth about what we expect now.

I look for expense assumptions doing hidden strategic work

Sometimes an expense line contains a strategy decision nobody has named.

R&D grows because the company is accelerating product investment. Sales expense rises because the go-to-market model is changing. Professional fees increase because management is preparing for a transaction.

I want those changes described as business choices, not buried as “OpEx +7%.”

The more strategic the expense, the less satisfied I am with a mechanical forecast.

Numbers are useful partly because they force management choices into the open.

Expense variance analysis should improve the next forecast

If travel misses every quarter because events are modeled badly, fix the event logic.

If software is always higher because seats follow headcount, connect the driver.

If professional fees arrive unpredictably, create a project schedule for the material engagements.

I do not want variance commentary becoming a monthly explanation of why the same forecasting method failed again.

The miss is information.

Use it.

Forecasting gets better when the model changes in response to evidence rather than merely changing the number.

The goal is not a more complicated expense model

I can make an expense model complicated very quickly.

That is not the challenge.

The challenge is choosing where detail improves a decision.

Material payroll deserves detail. A large software renewal deserves a schedule. A strategic marketing campaign deserves explicit timing. Office coffee probably does not need a driver tree.

I want the simplest method that captures the behavior of the cost well enough for management to understand the outlook.

That is usually a mix of schedules, drivers, run rates and explicit decisions.

One method for every line is simpler.

It is also often wrong in exactly the places that matter.

One last filter: would management recognize the forecast?

I like asking business leaders whether the expense outlook resembles the operating plan they think they are running.

If Operations says a major project starts in September and Finance has the cost beginning in June, we have a problem even if the annual total matches.

If Marketing canceled an event and the forecast still contains it because nobody changed the run rate, the model is stale.

If the department leader cannot connect the financial outlook with the decisions they made, Finance has translated the business badly.

The forecast should not require managers to understand every account.

It should reflect the choices they recognize.

That is the standard I care about.

Not whether every expense line has a clever formula.

Whether the model describes the business that management is actually planning to operate.