The craft
Reading a financial model
How to interrogate a startup financial model in an hour: find the assumptions that drive it, test them, and ignore everything past year two.
Realistic time One to two hours
An early-stage financial model is not a forecast. It is a statement of what the founders believe has to be true, expressed in numbers, and its value to you is entirely in the assumptions rather than in the outputs. The revenue line in year five is arithmetic performed on guesses; the guesses are the interesting part.
This makes reviewing one much faster than it appears. You are looking for three or four cells that drive everything else, checking whether they are plausible, and seeing what happens when you change them. An hour is usually enough.
The one output that matters directly is the cash line. Everything else is a projection; the runway is a fact about the present, and it determines how much negotiating room the company has and how urgent the round really is.
01
Find the drivers
Every model has a small number of assumptions that everything else follows from. In a subscription business it is usually new customers per month, price, and churn. In a marketplace it is transactions, take rate and repeat rate. In a consumer business it is acquisition volume, retention and revenue per user.
Find those cells and read them as claims about the world. "We will add forty new customers per month by month eighteen" is a claim; "revenue will be $2.4m in year two" is that claim multiplied out. Test the claim, not the product.
The most useful test is comparison with the present. A model showing a tenfold improvement in a conversion rate, or a sales cycle halving, or churn falling by two-thirds, is asserting that something will change fundamentally. Ask what specifically will cause it.
Check
- Identify the three or four assumptions that drive the model.
- Compare each with the current actual figure.
- Where an assumption improves, ask what specifically causes the improvement.
- Check whether headcount grows in line with the revenue it is meant to produce.
- Check whether cost of revenue scales properly with revenue.
02
Test the sensitivity
Change one driver at a time and watch the cash line. If halving the customer acquisition rate exhausts the cash six months early, that is the risk in the business and it is worth more attention than anything in the deck.
The specific test worth running: what happens if the next round is six months later than planned. Almost every model assumes the following round arrives on time, and a company that cannot survive a delay has a financing risk that dominates every operational one.
Ask the founders which assumption they are least confident about. Founders who have thought about their own model answer immediately and specifically. It is also a quick way to find out whether they built it themselves.
Check
- Halve the growth rate and see when cash runs out.
- Delay the next round by six months and see what happens.
- Increase acquisition cost by half and check the effect on margin.
- Ask which assumption the founders are least confident about.
- Ask who built the model.
03
The cash line
Cash is the only part of the model that is not a projection. Establish the current balance as of a recent date, the monthly net burn, and the resulting runway in months. Then compare that with the timeline of the round you are being asked to join.
A company raising with four months of runway is in a materially weaker negotiating position than one raising with twelve, and that shows up in terms, in urgency and in the quality of the investor it can attract. It is also a decision the founders made, and asking why the raise started when it did is informative.
Then ask what the round is meant to buy. A round sized to reach a specific milestone with a margin of safety is a plan. A round sized to eighteen months of the current burn is an assumption that things will work out.
Check
- Current cash balance, as of what date.
- Monthly net burn over the last three months, not the average for the year.
- Runway in months, and the timeline for the round.
- What milestone does this round reach, and with how much margin?
- What is the plan if the round is 30% smaller than target?
04
What to ignore
Anything past year two is decoration. Nobody can forecast a startup three years out and everybody knows it, which is why the exercise is conventional rather than informative. A model showing $80m of revenue in year five tells you the founders can use a spreadsheet.
Also ignore precision. A model producing revenue to the nearest pound is not more accurate than one rounded to the nearest ten thousand, and false precision occasionally indicates a founder who has confused the model with reality.
What is worth reading in the later years is the shape: does gross margin improve, does the cost base scale sublinearly with revenue, does the business become profitable at a plausible size. Those are claims about the business model rather than forecasts.
Check
- Ignore absolute figures beyond year two.
- Read later years for shape rather than for level.
- Check whether the business is ever profitable, and at what revenue.
- Check that gross margin improves for a stated reason.
Stop and think
Red flags
- A model the founders did not build and cannot explain.
- Key assumptions that improve dramatically with no stated cause.
- Headcount that does not grow with the revenue it is supposed to generate.
- Cost of revenue that stays flat while revenue grows tenfold.
- A cash line that assumes the next round arrives exactly on schedule.
- A raise starting with under four months of runway.
- Revenue projections presented with implausible precision.
The other side of the table
What founders are taught about this
Adora Cheung - How to Set KPIs and Goals
How founders are taught to choose the number they run the company against. When you ask which metric a founder cares most about, this is the reasoning you are testing for.
Published by Y Combinator. Channel verified from ycombinator.com, and YouTube's own oEmbed response names that channel as this video's author. Pressing play loads content from YouTube.
Take these into the room
Questions to ask the founders
- Which assumption in here are you least confident about?
- What is your actual conversion rate today, versus the one in the model?
- What happens to this if the next round is six months late?
- Who built this model?
- What milestone does this round get you to, and with what margin?
- What is your plan if you raise 30% less than target?
- When does this business become profitable, and at what revenue?
From the decoder