Where good ventures stall.
This is where many good ventures stall. The idea is rarely the difficulty. The difficulty is translation: turning a vision and a business model into numbers that a stranger can test and still believe. The principal of this firm has mentored more than 100 founders, and the pattern is consistent. The founder who struggles to raise usually has a sound business and a model that does not show it.
This article explains what an investor looks for in a model, and the seven red flags that keep an investor from finding it. Each one comes with a test you can run yourself and a fix.
What an investor is reading for.
No investor believes a five-year forecast. Everyone at the table knows the numbers will turn out differently. The model is read for something else, which is evidence of how you think.
Three things are on trial. The first is your approach: do you understand what drives revenue and cost in this business? The second is your logic: does each number follow from the ones before it? The third is your thoroughness: did you check your own work before you asked someone to put money behind it? A pitch deck can claim all three. The model is the one document in which an investor can test them.
That is why a weak model costs more than a weak slide. An investor who finds an error does not conclude that the spreadsheet has an error. The investor concludes that this is how the company will be run.
Errors are also far more common than founders expect. Raymond Panko of the University of Hawaii compiled field audits of working spreadsheets, and 49 of the 54 contained significant errors. His research puts the error rate at one percent or more of formula cells, even in spreadsheets built with care. A fundraising model can hold thousands of formula cells. It is usually built late, against a deadline, and more and more often with an AI tool. The person who built it is the person least able to see what is wrong with it, because the builder reads what the model was meant to say.
How the model gets read.
A partner seldom opens the model first. An analyst does, with one instruction: find out whether these numbers can be relied on.
The analyst checks whether the statements agree with one another. The analyst traces the largest revenue number back to the inputs that produce it, changes one assumption to see what moves, and scans each row for a cell that does not match its neighbors. Only after that does anyone weigh the business case.
So the red flags below are found before your vision gets a hearing. They fall into the three groups already named: approach, logic, and thoroughness.
Approach: does the model show how the business works?
i. Revenue is a number, not a mechanism.
You know how the business makes money: this many customers, at this price, staying this long. In the model, the revenue row simply grows by a percentage each month. The mechanism is in your head and not in the file.
What the investor reads. The founder has a target, not yet a business model.
The test. Find the cells for the number of customers, the price, and retention. Change one. If revenue does not move, the model is not built from them.
The fix. Build revenue from its drivers, one row for each, so that the model tells the same story you tell in the room.
ii. The assumptions are scattered.
Every model rests on a handful of choices: price, conversion, hiring dates, churn. In many founder models, those choices are typed inside formulas across a dozen worksheets. An investor wants to change the price and watch the result. An investor who cannot find the price cannot test your thinking.
What the investor reads. The founder has not separated what was decided from what follows from it.
The test. Select your price cell and use Trace Dependents on the Formulas tab. If the arrows do not reach the whole revenue build, part of the model runs on numbers you do not control.
The fix. Keep one assumptions sheet. Label each input and state the reason or the source beside it.
iii. The growth story contradicts itself.
Revenue triples while the sales team stays the same size. Gross margin improves every year with no stated reason. Receivables stay flat as sales grow. Each of these is the vision running ahead of the plan. A forecast can be ambitious. It cannot disagree with itself.
What the investor reads. The founder has not worked out what the growth will cost.
The test. For each year, compute revenue per salesperson, marketing spend per new customer, and days of receivables. A ratio that improves sharply without a stated reason is the question you will be asked in the meeting.
The fix. Tie growth to hiring, spending, and working capital, so that the model shows what the growth costs. That is also where the answer to "how much are you raising, and why?" comes from.
Logic: does each number follow from the last?
iv. Typed numbers sit inside formula rows.
A projection row should hold one formula, copied across every period. A typed number in the middle of the row means that someone overrode the logic, usually for a good reason such as a signed contract or a one-time cost. From that cell onward, the model no longer responds to its own assumptions.
What the investor reads. Which other numbers were adjusted by hand?
The test. In Excel, open Find and Select, choose Go To Special, and select Constants. Anything highlighted inside a formula row is an override.
The fix. Keep the formula. Move the exception to the assumptions sheet, with a label that says what it is.
v. The formula changes partway across a row.
This one is quieter. Every cell holds a formula, but the formula in year three is not the formula in year one. A reference points at the wrong row, or a range stops one column short. The numbers still look plausible, and that is what makes the flaw dangerous.
What the investor reads. If there is one of these, how many more are there?
The test. Select a projection row, open Go To Special, and choose Row Differences. A sound row shows none.
The fix. Compare every row, cell by cell. The work is tedious, and it is the reason a written protocol matters.
Thoroughness: did you check your own work?
vi. The three statements do not agree.
Assets must equal liabilities plus equity in every period. Closing cash on the cash flow statement must equal cash on the balance sheet. Net income must flow into retained earnings. When they do not, or when a line named "other" or "adjustment" quietly absorbs the difference, the model is three tables that share a file.
What the investor reads. The cash forecast, which decides how much you need to raise, cannot be trusted.
The test. In the final year, subtract liabilities plus equity from assets. Then subtract cash on one statement from cash on the other. Both answers must be zero. A difference of one dollar is a failure.
The fix. Remove the balancing line, trace the difference to its source, and add check rows that show zero in every period.
vii. Loose ends are left in the file.
These are error values, links to files the investor does not have, circular references, and hidden sheets with live calculations. None of them is a matter of judgment, and that is why they cost so much.
What the investor reads. Nobody read this from start to finish before it was sent.
The test. Search the workbook for #REF!. Open Edit Links on the Data tab. Look for a circular reference warning in the status bar. Unhide every sheet and row.
The fix. Make one clean pass before the file leaves your hands.
What the seven have in common.
Not one of them says anything about the quality of your business. Strong companies carry all seven. The trouble is that an investor cannot tell a strong business with a weak model from a weak business, and has many other models waiting.
Removing the flags does not make your forecast more optimistic. It makes the forecast legible. The conversation can then return to where you are strongest, which is the vision.
Say it in one page.
Send the model with a one-page note. State the version and the date, where actual results end and the forecast begins, the five or six assumptions that drive the result, and what the model does not cover.
The note is a test of its own. A founder who can name the five assumptions that matter has done the translation. A founder who cannot has found the next piece of work.
Where to start.
The tests above take an afternoon, and they will find the visible problems. They will not find every problem, for the reason given earlier: you read what you meant to write.
If the mechanics are the concern, that is the work of the Financial Model Red Flag Review. It is an independent, cell-level review of one Excel model against a written protocol of 259 checks, delivered in five business days from complete intake. On one inherited 30-worksheet model, the protocol found 141 formula errors across six worksheets, a $2.7 million balance sheet imbalance, and a $520,000 cash flow overstatement. The founder receives every finding with its location, its severity, and why it matters to an investor.
The review is not an audit, attestation, or assurance engagement. It gives no accounting, tax, legal, or valuation opinion, it does not validate the assumptions or the valuation, and it carries no promise of funding.
If the difficulty lies earlier, in turning the vision into a business model that can be written down at all, that is the work of the firm's Fundraising and Capital practice.