Budget Variance Analysis: How FP&A Finds What Actually Matters
There is a particular kind of monthly Finance meeting where every variance gets treated like it has committed a crime.
Travel is $8,000 over budget.
Explain it.
Software is $11,000 under.
Explain it.
Office supplies are $3,700 over.
Apparently we need closure.
Meanwhile, revenue is $2.4 million below plan and somehow we’ve allocated roughly the same amount of human curiosity to all four.
This is one of the stranger habits in FP&A.
We build variance reports containing dozens, sometimes hundreds, of differences between what we expected and what actually happened.
Then we start at the top.
And explain them.
One.
By.
One.
By the time we reach anything management could actually do something about, everyone has been staring at the same slide for 38 minutes and someone is visibly reconsidering a career in Finance.
I understand the instinct.
Variance analysis is important.
But not every variance deserves an investigation.
And not every explanation deserves to be in the meeting.
If you’re doing budget variance analysis well, the goal isn’t to explain everything that moved.
The goal is to figure out what changed, why it matters, whether it will happen again, and what management should do about it.
Those are very different things.
What Is Budget Variance Analysis?
Let’s do the definition first.
Budget variance analysis compares actual financial results with what the company budgeted or expected.
At its simplest:
Variance = Actual − Budget
You can also calculate the variance as a percentage.
If revenue was budgeted at $10 million and actual revenue was $9 million, you have a $1 million variance to investigate.
If an expense was budgeted at $500,000 and actual spending was $450,000, you’ve spent $50,000 less than planned.
The mechanics aren’t particularly difficult.
Current guidance from CFI and IBM describes variance analysis similarly: compare actual results against a budget or forecast, identify meaningful differences, understand the causes, and use that information to improve decisions and resource allocation. Corporate Finance Institute
Excel can calculate the difference in approximately the amount of time it took you to read that sentence.
That’s not the hard part.
The hard part is deciding which differences mean something.
Favorable and Unfavorable Variances
You’ll usually see variances classified as either favorable or unfavorable.
Revenue above budget?
Usually favorable.
Expenses below budget?
Usually favorable.
Revenue below budget?
Usually unfavorable.
Expenses above budget?
Usually unfavorable.
Simple.
Except reality enjoys ruining simple things.
Suppose payroll is $300,000 below budget.
Green.
Wonderful.
Except the reason is that Operations couldn’t hire the 12 people it needed.
Now production is behind.
Is that still favorable?
Technically, yes.
Operationally?
We should probably keep reading.
Or suppose revenue beats plan by $2 million.
Excellent.
Except Sales got there through aggressive discounting and gross margin is deteriorating.
Still celebrating?
This is why I’ve become increasingly suspicious of green cells.
Current variance-analysis guidance makes the same distinction: a favorable variance can reflect delayed spending, understaffing, discounting, timing or another condition that isn’t necessarily positive for the business. Finamodel
Green means different from plan in the financially favorable direction.
It does not mean:
Nothing to see here. Everyone go home.
I wish Finance were that easy.
Start With Materiality
If I were reviewing your variance report, this is one of the first things I’d ask:
What actually matters?
Not:
What moved?
Everything moves.
Your business is not a museum exhibit.
Things happen.
Maybe revenue is $200,000 below plan.
Payroll is $150,000 under.
Marketing is $40,000 over.
Office supplies are $6,000 over.
Travel is $3,000 under.
You could explain all five.
Or you could decide which ones deserve management’s attention.
CFI specifically recommends considering both absolute dollar and percentage variances rather than treating every difference equally, and it suggests establishing materiality thresholds based on business significance. Corporate Finance Institute
I like thresholds.
But I don’t like using them blindly.
Maybe your rule is:
Investigate anything greater than $100,000 or 10%.
Fine.
Now suppose a $60,000 variance signals that customer churn has started increasing.
Below threshold.
Still interested.
Materiality should help Finance focus.
It shouldn’t replace judgment.
Dollar Variance and Percentage Variance Tell Different Stories
This catches people.
Imagine two expenses.
Expense A:
Budget: $10 million
Actual: $10.5 million
Variance: $500,000, or 5%.
Expense B:
Budget: $100,000
Actual: $150,000
Variance: $50,000, or 50%.
Which matters more?
I don’t know.
That’s the point.
The first has a much larger financial impact.
The second has a much larger percentage miss.
Maybe Expense B reveals something structurally wrong.
Maybe Expense A is ordinary timing.
Or maybe half a million dollars is half a million dollars and somebody should explain it.
You need both views.
Dollar impact tells me how much it matters financially.
Percentage tells me how far reality moved from the assumption.
Then I want context.
Finance keeps trying to find one column that will tell us what to think.
Unfortunately, thinking remains stubbornly manual.
A Variance Isn’t an Explanation
This one drives me a little crazy.
You’ve probably seen commentary like:
Marketing expense unfavorable by $420K due to higher marketing spend.
Thank you.
I was worried we’d never solve it.
That’s not variance analysis.
That’s the variance wearing a sentence.
A useful explanation tells me what actually changed.
Maybe:
Marketing expense was $420K above plan primarily because the product launch campaign moved forward from Q4 into September. Full-year spending is still expected to remain within forecast.
Now I know something.
The spending is timing.
It’s tied to a known activity.
And the full-year outlook hasn’t changed.
Or:
Marketing expense was $420K above plan because paid acquisition costs increased 18% while lead volume remained flat. We are reducing spend in two channels while Marketing reviews conversion performance.
Different problem.
Same $420,000.
Very different management response.
Current budget-vs-actual guidance similarly recommends moving beyond the numerical gap to identify operational causes such as price, volume, mix and timing. Wall Street Playbook
That’s what I want from FP&A.
Don’t repeat the number.
Tell me what happened.
I Usually Want to Put Variances Into Buckets
Once you’ve found something material, I’d classify it.
You don’t need my exact categories.
But you need some way to separate fundamentally different explanations.
I usually think about things like:
Timing
The expense or revenue moved between periods.
It hasn’t necessarily changed the full-year outlook.
Volume
You sold, produced, hired, shipped or consumed more or less than expected.
Price or rate
The quantity may be right, but the price wasn’t.
Mix
The composition changed.
Maybe you sold the right total volume but more lower-margin products.
One-time
Something happened that shouldn’t repeat.
Structural
Something changed in the economics of the business.
Execution
The assumption may have been reasonable, but the business didn’t execute against it.
Bad assumption
Nothing necessarily went wrong operationally.
We were simply wrong.
I have a strange affection for that last one.
Finance should be allowed to say:
We were wrong.
It’s remarkably freeing.
You don’t need to construct a 14-slide defense of the original forecast.
Sometimes the assumption was bad.
Learn something.
Update it.
Move on.
Timing Variances Deserve Their Own Little Warning Label
Timing is one of the most common variance explanations.
And sometimes it’s completely legitimate.
A $500,000 expense was expected in September.
It lands in October.
Fine.
September is favorable.
October will be unfavorable.
Nothing structurally changed.
But I get suspicious when everything becomes timing.
Revenue miss?
Timing.
Hiring miss?
Timing.
Capex?
Timing.
Customer collections?
Timing.
At some point I start wondering whether timing is the Finance equivalent of when my kids tell me they’ll clean their room later.
Possible.
Historically unreliable.
If something is timing, tell me when it reverses.
That’s the part that makes the explanation useful.
If you can’t tell me that, I’m not sure we have a timing variance.
We may just have a variance we’re hoping goes away.
Price, Volume and Mix: Where Revenue Variance Gets Interesting
Suppose revenue beats budget by $1 million.
Fantastic.
Why?
Maybe you sold more units.
That’s volume.
Maybe customers paid more.
That’s price.
Maybe customers bought a different combination of products.
That’s mix.
Those are not interchangeable.
Price-volume-mix analysis is commonly used to break revenue variances into their underlying economic components rather than stopping at the total difference. Current FP&A examples specifically recommend decomposing material revenue variances this way. Corporate Finance Institute
Imagine revenue is up 8%.
But unit volume fell 4%.
Price increases created the growth.
Interesting.
Now I want to know whether that was intentional.
What happened to customer retention?
What happened to margin?
Can the pricing hold?
Or perhaps volume surged but only because customers shifted toward a lower-margin product.
Revenue looks wonderful.
Margin would like a word.
This is why I don’t like congratulating numbers until I’ve met their parents.
The Forecast Should Learn Something
Here’s where variance analysis becomes FP&A instead of monthly archaeology.
Suppose you discover that customer churn is consistently higher than the assumption in your forecast.
What happens next?
Do we:
A. Explain the variance again next month.
Or:
B. Change the forecast assumption.
Please choose B.
I know.
Trick question.
But Finance teams sometimes become extremely good at explaining the same miss repeatedly without allowing the explanation to change the forecast.
If the variance taught us something about the business, the forecast should reflect it.
That’s the loop:
Expectation → Actual result → Variance → Explanation → Updated expectation
If the last step never happens, we’re just writing historical commentary.
Perfectly respectable.
Not FP&A’s highest calling.
Favorable Variances Deserve Curiosity Too
I’ve written about this before because I genuinely find the behavior fascinating.
Bad result?
The room becomes CSI: Finance.
What happened?
Who owns this?
When did it start?
Can we pull the data?
Can we call Sales?
Can someone check whether this includes Europe?
Good result?
Nice work.
Next slide.
Apparently curiosity has a P&L threshold.
But let’s say gross margin beats forecast by 400 basis points.
I want to know why.
Maybe pricing is stronger.
Maybe product mix improved.
Maybe procurement negotiated something meaningful.
Maybe productivity increased.
Maybe a cost hasn’t hit yet.
Maybe the forecast was simply wrong.
Maybe something genuinely improved in the business and we should understand it well enough to repeat it.
That’s worth knowing.
A favorable variance is still reality disagreeing with our expectation.
Sometimes good news is information.
Not just a reason to move to the next slide.
Recurring Variances Are Trying to Tell You Something
One miss can be noise.
Three similar misses start looking like a pattern.
Let’s say freight expense is over forecast every month.
January.
February.
March.
April.
At some point, I don’t want another freight variance explanation.
I want to know why the forecast still doesn’t know about freight.
This is where variance analysis becomes useful for improving the planning process itself.
Maybe:
the rate assumption is wrong,
the volume driver is wrong,
the mix changed,
the contract changed,
or the model doesn’t reflect how the expense actually behaves.
Repeated variances are feedback.
Your forecast is telling you where it doesn’t understand the business.
I would listen.
It’s one of the few times the spreadsheet is trying to communicate with you without a #REF!.
Separate Controllable From Uncontrollable
This is another distinction management usually cares about.
Suppose gross margin falls because commodity prices suddenly spike.
Different conversation from gross margin falling because production scrap doubled.
Both matter.
But one may be external.
The other may be operational.
I don’t like the idea that “uncontrollable” means “nothing to do.”
Management may still need to respond.
Pricing.
Sourcing.
Inventory.
Product mix.
Hedging.
Cost reduction.
Something.
But separating what happened to the business from what happened inside the business helps leadership understand the response.
FP&A isn’t there to assign moral character to the variance.
We’re trying to understand it.
Please Talk to the Business
I cannot emphasize this enough.
You cannot do great variance analysis entirely from the general ledger.
You can identify the difference.
You can slice it.
You can calculate it.
You can make the conditional formatting extremely aggressive.
Eventually you need a human.
Why did contractor expense increase?
Ask the department.
Why did conversion fall?
Ask Sales.
Why did overtime spike?
Ask Operations.
Why is hiring below plan?
Ask HR.
Why did customers stop buying Product B?
I don’t know.
Please find someone who does.
This is where strong FP&A people become part analyst, part business partner, part detective.
I’ve said before that Finance people probably need little PI badges.
I remain available to design them.
But Don’t Just Accept the First Explanation
This is the other side of business partnering.
Department leader:
Travel was over because we traveled more.
Finance:
Great, thanks.
No.
You don’t need to interrogate people under a bright light.
This isn’t Law & Order: FP&A.
But you should understand the explanation.
Why was travel higher?
Customer visits?
Conference?
New sales territories?
Unplanned executive travel?
Will it continue?
Is it included in the forecast?
Was there a return from it?
The first answer often explains what happened.
Your job is to understand why it happened and whether it matters going forward.
That’s where curiosity earns its salary.
The Monthly Review Should Not Be a Reading Exercise
If your monthly Finance meeting consists of everyone reading the variance report together, I have wonderful news.
Email exists.
The meeting should be for the things that need discussion.
What changed materially?
What surprised us?
What’s becoming a trend?
What assumptions are now questionable?
What decisions are needed?
Where does management need to intervene?
Which favorable result might be repeatable?
What changed in the forecast?
Those are meeting questions.
Whether telecommunications expense was $4,800 over budget probably isn’t.
Unless something very interesting happened with the phones.
What Should Good Variance Commentary Include?
I don’t need a novel.
Actually, please don’t give me a novel.
If your variance commentary requires three paragraphs, I start wondering whether we know what caused the variance.
For most important variances, I’d want something close to:
What moved?
Why?
Timing, one-time or recurring?
Does it change the forecast?
Is action required?
That’s enough surprisingly often.
For example:
Revenue was $1.1M below forecast due primarily to lower conversion in the enterprise segment. Pipeline volume remains on plan, but close rates fell from 24% to 19%. We reduced the Q4 revenue forecast by $2.3M while Sales reviews stalled opportunities.
Now I know:
the size,
the driver,
the operating metric,
the forecast impact,
and what is happening next.
That’s useful.
Compare that with:
Revenue unfavorable due to lower sales.
I may frame that one.
AI Is Going to Write Variance Commentary
Of course it is.
AI is extremely well suited to parts of variance analysis.
It can identify anomalies.
Compare actuals with budget.
Analyze trends.
Find patterns.
Generate first-draft explanations.
Potentially connect financial and operational data faster than an analyst manually could.
Wonderful.
Please take the first draft.
I have no sentimental attachment to writing:
SG&A unfavorable by $327K.
But this creates another problem.
If AI can generate commentary for every variance, companies may suddenly possess vast industrial capacity for explanations nobody needed.
That’s not progress.
That’s automation discovering Finance’s existing habits.
The value will increasingly come from deciding:
Which variance matters?
Which explanation is believable?
What changed structurally?
What should management do?
What assumption should change?
That’s judgment.
And if AI eventually gets excellent at that too, I’ll have more time for coffee.
I’m prepared to adapt.
A Simple Test for Every Variance
Next time you’re reviewing budget vs. actual results, pick a variance and ask:
Is it material?
If no, maybe stop.
Do I understand what caused it?
If no, investigate.
Is it timing, one-time or structural?
That changes the interpretation.
Will it happen again?
Now we’re getting somewhere.
Does it change the forecast?
If yes, update it.
Can management do anything about it?
If yes, discuss the action.
If management does nothing, does anyone care?
This one eliminates more commentary than you might expect.
And I mean that as a compliment.
What I’d Want to See in a CFO Review
Not 83 lines with comments.
I’d rather see the handful of things that explain what changed in the business.
Maybe:
Revenue is $2.4 million below plan because enterprise conversion fell.
Gross margin is 250 basis points above plan because pricing held while material costs declined.
Payroll is $600,000 favorable because 11 planned hires haven’t started.
Working capital is worse because customer collections slowed.
Marketing is $300,000 unfavorable due to a campaign pulled forward from next quarter.
Now we have a conversation.
Which ones are temporary?
Which ones are structural?
What changes our full-year outlook?
Where does management need to act?
That’s a Finance meeting.
The rest can live in the appendix.
The appendix has never complained.
Go Look at Your Variance Report
Seriously.
Open the last one.
How many lines have explanations?
Twenty?
Forty?
Eighty?
Now ask yourself:
How many of those explanations changed a decision?
That’s the number I’m interested in.
Then look for recurring explanations.
“Timing.”
“Higher than expected.”
“Lower volume.”
“Open positions.”
“Delayed project.”
If you’ve written the same explanation three months in a row, stop writing it.
Fix the forecast.
Or fix the business.
Possibly both.
Variance analysis shouldn’t become a monthly ritual where Finance explains why reality once again refused to follow the spreadsheet.
Reality has made its position clear.
Our job is to learn something from it.
The Point Isn’t to Explain the Past Perfectly
This is probably where I land on all of this.
Yes, Finance should understand what happened.
Yes, variances matter.
Yes, good commentary matters.
But I’m much less interested in producing a perfect explanation of last month than I am in figuring out what last month tells us about the next one.
What assumption was wrong?
What changed?
What is becoming a pattern?
What should we expect now?
What should management do differently?
That’s where variance analysis earns its place in FP&A.
Otherwise we’re just annotating history.
And Accounting already has the actuals covered.
I’d rather use them to make the next decision a little less surprising.
I am naturally pessimistic, so “a little less surprising” is about as optimistic as I’m willing to get.
Have a Variance That Won’t Leave You Alone?
If you’ve got one of those Finance problems that keeps showing up every month wearing a slightly different number, send me a note.
Maybe it’s a forecast that never quite explains reality.
Maybe your monthly review has become a 70-line reading exercise.
Maybe you have a favorable variance that looks suspiciously favorable.
Or maybe you just want another Finance person to look at something and say:
“Okay. That’s weird.”
Those are some of my favorite conversations.
I’m always happy to talk through the problem, answer a question, or compare notes about what you’re seeing in the business.
No 47-slide deck required.
