How to Review a Financial Model Before You Trust It
I don’t start reviewing a financial model with the formulas.
This occasionally bothers people.
They’ve spent a lot of time building those formulas.
There may be nested IF statements. INDEX MATCH. XLOOKUP. Maybe something involving OFFSET that nobody wants to discuss.
I’m sure it’s all very impressive.
I want to know why revenue grows 12%.
Because I’ve learned that a financial model can be technically flawless and still tell you something completely ridiculous.
Excel will calculate a bad assumption with tremendous confidence.
It considers this none of its business.
So when someone hands me a financial model and asks whether I trust it, I don’t begin by hunting for broken formulas.
I start with the story the model is telling me.
Then I try to figure out whether I believe it.
What Does It Mean to Review a Financial Model?
A financial model review is an evaluation of the model’s assumptions, logic, calculations, structure, outputs and usability.
But I think there’s another part that matters just as much:
Does the model make economic sense?
A formula can be correct.
The number can tie.
The workbook can balance.
And the model can still be wrong.
That’s why I separate two questions:
Does the model work?
and
Should I believe it?
They’re related.
They’re not the same.
The first question is largely technical.
The second requires judgment.
And the second one is usually where I spend more of my time.
Step 1: Don’t Touch Anything Yet
My first move when opening someone else’s model is usually not to start clicking through formulas.
I look around.
What is this model trying to do?
Is it a budget?
Forecast?
Long-range plan?
Valuation?
Operating model?
Cash forecast?
Scenario model?
Board model?
Something somebody started four CFOs ago and everyone is now afraid to delete?
I want to understand the purpose before evaluating the construction.
A model should have a job.
If nobody can explain what decision or process the model supports, I’m already suspicious.
Not every spreadsheet needs to become a permanent corporate institution.
Some deserve a dignified retirement.
Step 2: Find the Assumptions
This is where my trust issues really begin.
Show me the assumptions.
Revenue growth.
Pricing.
Volume.
Headcount.
Compensation.
Gross margin.
Utilization.
Conversion.
Inflation.
Capital expenditures.
Working capital.
Whatever actually drives the model.
I want to know:
Where did this assumption come from?
Who owns it?
When was it last updated?
What evidence supports it?
Does it reflect what’s happening in the business now?
And my personal favorite:
Does anyone actually believe this?
That last question has produced some wonderful silences.
Sometimes an assumption has been sitting in a model so long that nobody remembers where it came from.
It was 8% last year.
It’s 8% this year.
Presumably it will be 8% until the sun burns out.
At that point, the assumption isn’t really an assumption anymore.
It’s company folklore.
Step 3: Look at the Model Without the Formatting
I’ve written before about my trust issues with beautiful financial models.
I stand by them.
Formatting is useful.
A well-designed model is easier to understand and operate.
But formatting also creates authority.
Blue inputs.
Green links.
Perfect borders.
Beautiful charts.
Everything aligned.
Suddenly 12% looks extremely employed.
So mentally strip all of that away.
Imagine the model as plain numbers.
Would the assumptions still look convincing?
Would the logic still make sense?
Would you still believe the outputs?
I’ve seen mediocre thinking become considerably more persuasive after someone gave it a nice font.
This is not unique to Finance, but we are very good at it.
Step 4: Understand What Actually Drives the Model
Next I want to find the handful of assumptions doing most of the work.
Every model has them.
Maybe it’s revenue growth.
Maybe pricing.
Maybe customer retention.
Maybe utilization.
Maybe hiring.
Maybe gross margin.
Maybe a terminal value quietly carrying half the valuation while everyone admires the chart on the summary tab.
Find the variables that materially change the outcome.
This is especially important in FP&A models because good forecasting should connect financial results to the underlying business drivers.
If revenue grows 15%, I want to understand what produces that growth.
More customers?
Higher prices?
Better retention?
More capacity?
A new product?
Divine intervention?
“Revenue grows 15%” is an output pretending to be an assumption.
I want the mechanism underneath it.
Step 5: Trace a Few Important Numbers All the Way Through
Now I start getting into the mechanics.
Pick several important outputs and trace them backward.
If EBITDA is $18 million, where does it come from?
If revenue is $100 million, what creates it?
If ending cash is $7 million, how did we get there?
If headcount reaches 240, what assumptions drive those hires?
Follow the number from output to source.
This is where you start finding things.
Hardcoded values.
Broken links.
Old assumptions.
References to hidden tabs.
Numbers pulled from mysterious external files.
Formulas copied one column too far.
The random +20000 sitting at the end of a formula.
I have encountered enough of those little additions to know that they rarely come with a heartwarming backstory.
Usually someone needed the model to work.
And apparently $20,000 helped.
Step 6: Look for Hardcodes in Formulas
Hardcodes aren’t automatically wrong.
Sometimes they’re appropriate.
I just want to know they’re there.
A formula like:
=B17*C24
is telling me something different from:
=B17*C24+20000
That extra $20,000 and I need to become acquainted.
What is it?
Why is it there?
Who added it?
Is it recurring?
Should it be an assumption somewhere else?
Does anyone remember?
A good model makes important assumptions visible.
When assumptions are buried inside formulas, they become harder to review, update and challenge.
That’s how numbers survive long after their reason for existing has disappeared.
Step 7: Check the Time Logic
Models are frequently correct in total and wrong in timing.
That matters.
When does revenue begin?
When do new hires start?
When do price increases take effect?
When are bonuses paid?
When does CapEx hit?
When does depreciation begin?
When does a customer churn?
When do collections happen?
I pay particular attention to timing around major changes.
A model may correctly assume 20 new employees and still overstate expense because all 20 are accidentally hired in January.
Or understate it because apparently everyone joins on December 31.
Hiring plans occasionally become very optimistic about recruiting speed when the model needs EBITDA to cooperate.
Timing deserves its own review.
Step 8: Check Whether the Financial Statements Talk to Each Other
If the model includes an income statement, balance sheet and cash flow statement, they should behave like they know one another.
Net income should flow appropriately.
Working-capital movements should affect cash.
CapEx should affect cash and fixed assets.
Debt should affect cash, the balance sheet and interest.
Retained earnings should roll.
The balance sheet should balance.
This sounds obvious.
Yet there is a special kind of moment when someone says:
“The balance sheet is off by a little.”
How little?
$4.7 million.
Excellent.
Let’s stay here for a while.
A model that doesn’t integrate correctly can produce outputs that look reasonable while the underlying economics are broken.
That’s dangerous because reasonable-looking numbers don’t always trigger questions.
Step 9: Compare the Model With History
This is one of my favorite checks because history has no obligation to support our optimism.
Compare forecast assumptions with actual performance.
Revenue growth.
Margins.
Hiring.
Churn.
Working capital.
Conversion rates.
Utilization.
CapEx.
Whatever matters.
If gross margin has been between 42% and 45% for three years and the model jumps to 53%, I want to know what changed.
Maybe there is an excellent reason.
New pricing.
Better mix.
Automation.
Supplier renegotiation.
Scale.
Fine.
Show me.
But if the answer is:
“That’s what we need to hit the plan.”
We have discovered something.
Not necessarily about gross margin.
Step 10: Compare It With What the Business Is Saying
A financial model should not live in a Finance-only universe.
Talk to Sales.
Operations.
HR.
Marketing.
Customer Success.
Whoever owns the underlying activity.
If the model assumes 30% growth while Sales is quietly worried about pipeline, I’d like those two pieces of information to meet.
If Finance assumes 40 hires but HR thinks 25 is realistic, we should probably introduce them.
If the model assumes improving retention while Customer Success is dealing with three major at-risk accounts, that’s worth knowing.
This is one reason I think good FP&A people spend so much time outside Finance.
The spreadsheet contains numbers.
The business contains the reasons.
You need both.
Step 11: Stress the Model
Now make something go wrong.
This is usually my favorite part.
Lower revenue.
Delay hiring.
Increase churn.
Reduce gross margin.
Push collections.
Increase costs.
Lose a large customer.
Move a launch.
Then watch what happens.
Does the model behave logically?
Do the financial statements respond correctly?
Does cash move?
Does headcount move?
Do margins react?
Does anything break?
Sometimes a model works beautifully right up until reality deviates from the base case.
Unfortunately, reality has a long history of doing that.
A model should be able to survive disagreement with its creator.
Step 12: Find the Assumptions the Model Is Most Sensitive To
Not every assumption deserves equal attention.
If changing an assumption by 1% barely moves the outcome, I’m probably not spending half the meeting debating it.
If changing another assumption by 1% wipes out $3 million of EBITDA, I’d like to linger.
Sensitivity analysis helps identify where uncertainty actually matters.
This is useful for another reason.
It tells management what to watch.
Maybe the entire plan is unusually sensitive to retention.
Or pricing.
Or utilization.
Or hiring timing.
Now we’ve learned something more valuable than whether the spreadsheet works.
We’ve learned where the business plan is fragile.
That’s the part I care about.
Step 13: Try to Break It
Once I understand the model, I become slightly annoying.
What happens if I enter zero?
What happens if growth turns negative?
What happens if a start date changes?
What happens if I add a department?
What happens if I extend the forecast?
What happens if I remove a product?
What happens if I change the scenario?
I’m not trying to torment the person who built it.
Usually.
I’m trying to figure out whether the model is robust enough for other people to use.
Because somebody eventually will.
And that person will do something the model builder never imagined.
I have children.
I understand this principle extremely well.
Step 14: Look at the Outputs Last
Only after I understand the mechanics and assumptions do I spend much time with the executive outputs.
Charts.
KPIs.
Scenario summaries.
Valuation.
Cash.
EBITDA.
Whatever leadership is supposed to use.
Now I can ask:
Does the output answer the question the model was built to answer?
Can an executive understand it?
Does it highlight the important drivers?
Does it show uncertainty?
Does it make decisions easier?
Or did we spend 60 hours constructing an extraordinarily sophisticated machine that produces a chart everyone could have made in 15 minutes?
This happens more than Finance would like to admit.
My Financial Model Review Checklist
If someone handed me a financial model tomorrow, this is roughly the sequence I’d use.
Purpose
What is this model supposed to do?
What decision does it support?
Assumptions
Where did they come from?
Who owns them?
Are they current?
Do we believe them?
Business drivers
What actually drives the results?
Are those drivers modeled explicitly?
Formulas
Are there errors, hardcodes, broken links or inconsistent formulas?
Timing
Do revenue, expenses, hiring, collections and investments occur when they realistically should?
Integration
Do the income statement, balance sheet and cash flow statement connect properly?
History
How do the assumptions compare with actual performance?
Business reality
Do the operating teams agree with what Finance has modeled?
Sensitivity
Which assumptions matter most?
Scenarios
Does the model behave correctly when things change?
Usability
Can somebody other than the creator operate it safely?
Outputs
Does the model actually help someone make a decision?
That last question matters more to me than almost everything else.
The Biggest Red Flags I Look For
There are certain things that make me slow down immediately.
A model with no clear assumptions section.
Hardcodes buried inside formulas.
External links nobody can explain.
An assumption that hasn’t changed in years.
Revenue growth with no operational driver.
A balance sheet that “mostly” balances.
A huge improvement in margins with no clear explanation.
Cash that somehow remains healthy in every scenario.
A forecast that always lands exactly on the target.
A model only one person understands.
And one phrase:
“Don’t touch that tab.”
I will absolutely be looking at that tab.
A Model Can Be Wrong Without Having an Error
This is probably the distinction I care about most.
People sometimes treat model review like an audit for broken formulas.
That’s necessary.
But it’s not enough.
Imagine the model assumes:
Revenue grows 20%.
Gross margin improves four points.
Hiring happens exactly on schedule.
Churn falls.
Customers pay faster.
Costs stay controlled.
Every formula could be perfect.
The model could balance to the penny.
And I could still have absolutely no interest in betting the company on it.
The danger isn’t always an Excel error.
Sometimes the spreadsheet is doing precisely what we asked it to do.
We just asked it to model nonsense.
When Do I Finally Trust the Model?
Probably never completely.
This is Finance.
We don’t need to get carried away.
But I trust it considerably more when I can understand the business logic without needing the person who built it sitting beside me.
I want assumptions I can find.
Drivers I can explain.
Formulas I can trace.
Outputs I can reproduce.
Scenarios that behave logically.
And a model that changes when reality changes.
Most importantly, I want to know where the model is uncertain.
A financial model shouldn’t create the illusion that we know what’s going to happen.
We don’t.
If we did, I would have found a much more lucrative application for Excel by now.
The model’s job is to make our assumptions explicit, connect them to financial outcomes and help us understand the consequences if we’re wrong.
That’s why I start with the assumptions.
Formulas tell me whether the spreadsheet works.
Assumptions tell me whether the thinking does.
And given the choice, I’ll take ugly Excel with good thinking over beautiful Excel with bad assumptions every time.
I can fix ugly.
