What Investors Actually Check in Your Financial Model

Founders often treat the financial model as a formality: a tab to hand over once the real conversations are done. Investors treat it as a window into how you think. They are not grading the formatting. They are checking whether the numbers reason, and whether the person who built them understands what they mean.

They pay attention, too. DocSend’s analysis of investor behaviour found the team and financials slides draw the most attention of any in a deck, because that is where the real bet is made. A model that reasons is not a formality; it is one of the few places an investor decides whether to keep reading.

That is the reframe worth holding onto. The model is not a document you submit; it is a conversation you have. Every assumption in it is something an investor can question, and how you answer tells them more than the number itself. Here is what an experienced investor actually clicks into, in the order they do it.

They check that the three statements connect

A credible model is not three separate stories. Profit flows into the cash flow, cash flows onto the balance sheet, and the balance sheet balances. When an investor changes one assumption and the model breaks, or worse, silently does not, confidence drains quickly.

The quickest test they run: change revenue in one month and watch whether the cash balance, the receivables, and retained earnings all move in step. If only the profit and loss reacts, it is a projection with a balance sheet taped on, not a model.

Why does this matter so much? Because the connections are where the real insight lives. The link between the profit and loss and the cash flow is exactly where the profit-is-not-cash problem shows up: a model that connects them will show you burning cash in a month you booked a profit, which is the single most useful thing a model can tell a founder. A model that does not connect them hides that, from the investor and from you.

If your model is a single profit and loss tab with a revenue line growing at a fixed percentage, that is a forecast, not a model.

They separate assumptions from calculations

Before the content, investors read the structure, and good structure is itself a signal. A model built well keeps its assumptions in one place, clearly labelled and coloured, separate from the cells that calculate off them. Growth rate, price, headcount, churn, payment terms: these are inputs a reader can find and change. Everything else flows from them.

The opposite, numbers hard-typed into the middle of formulas, tells an investor two things: the founder cannot easily test scenarios, and nobody can trust that a figure is not a leftover from an old version. Clean inputs are not cosmetic. They are what make the model a tool rather than a snapshot.

They pressure-test the growth assumption

Every model has one line doing most of the work, usually revenue growth. Investors will click into that cell and ask a simple question: where does this come from?

“We will grow 20 percent a month” is a hope. “We add 12 customers a month at an average contract value of X, and here is why 12 is reasonable given our pipeline and two salespeople” is a plan.

Drivers beat percentages. Build revenue bottom-up from the things you actually control: leads, conversion rate, average order value, churn. Then the growth rate is an output of your assumptions, not an input you typed in. The bonus is that a driver-based model is also a management tool: when reality diverges from plan, you can see which driver missed, and fix that, rather than staring at a revenue line that is simply lower than hoped.

The fastest way to lose the room: a hockey-stick that appears in month four with no change in the drivers underneath it. If the inputs do not change, the output should not either.

They look for the moment you run out of cash

This is the least glamorous line and the most important one. Investors trace the cash balance down the timeline to find the point where it dips lowest. That point tells them how much you need to raise, and how much room you have if things run slow.

If your model never shows a dip, they will not believe it. Real businesses have a trough, usually a few months after a hiring or inventory push. Showing yours, and showing you have planned for it, reads as maturity.

Label that low point. Then show that the raise you are asking for clears it with a sensible buffer, ideally enough runway to hit the next milestone plus a few months. The ask is not a round number you liked; it is the size of the trough plus the cushion, and a model that derives it that way answers the “why this amount?” question before it is asked.

They sanity-check against your actuals

The fastest way to undermine a forecast is to ignore history. If your last six months averaged 40 percent gross margin and your projection jumps to 65 percent in month one with no explanation, the whole model becomes suspect.

Anchor the forecast to your actuals, then justify every improvement. A margin that climbs from 40 to 55 percent over eighteen months because you are renegotiating a supplier contract is credible. The same jump overnight, with no reason attached, is not. The same applies to every line: the model should start where the business actually is, then move for reasons you can name.

Show scenarios, not a single future

No forecast is right, and investors know it. What they want is a founder who understands the range. A simple base, upside and downside, driven by changing two or three key assumptions, shows that you have thought about what could go wrong and what it would cost. The downside case is often the most reassuring thing in the whole model, because it shows the business survives a bad year, and it tells the investor you will not be surprised by one.

You do not need a dozen scenarios. Three, cleanly driven off your assumptions tab, are enough.

Do not over-model

A caution in the other direction: precision is not credibility. A five-year monthly model with revenue to the naira implies a confidence no early business has, and experienced investors read false precision as inexperience. Monthly detail for the first 12 to 18 months and annual thereafter is plenty. The goal is a model that reasons, not one that pretends to predict.

A quick self-check before you send it

• Change one input. Do all three statements respond correctly?
• Are your assumptions in one labelled place, separate from the calculations?
• Can you explain the revenue line in terms of units, not just a percentage?
• Where is the lowest cash point, and does the raise clear it with a buffer?
• Does month one line up with last month’s actuals?
• Is there a downside case, and does the business survive it?

None of this is really about the spreadsheet. A model that connects, reasons from drivers, shows its trough, respects history and admits a range tells an investor something no deck can: that the founder understands their own business. That is what they are checking for.