The economics don't lie.
What unit economics actually reveal about a business, and why most founders avoid building the model until it is too late.
There is a version of every business that exists only in the founder's head. It is well margined, scalable, and inevitable. The customers are real. The market is large. The timing is right.
Then someone builds the model.
Most founders treat financial modelling as a fundraising exercise. Something you do when an investor asks for it, formatted to look serious, projected to Year 5, and built backwards from a number that seemed ambitious but fundable. This is not a model. It is a narrative with a spreadsheet attached.
A real model does something different. It forces you to be specific about things you would rather leave vague.
Unit economics are not complicated. They answer a single question: does this business work at the level of one transaction?
Revenue per customer. Cost to acquire that customer. Cost to serve them. What is left over. How many times they come back. Whether the margin holds as volume grows or quietly collapses under it.
That is it. But answering those questions with real numbers, not estimates, not industry benchmarks, not once we have scale assumptions, is where most business thinking breaks down. Because every assumption in the model is a belief you hold about how the business works. Building the model forces you to name those beliefs explicitly, where they can be examined, pressure tested, and if necessary, killed.
Most founders would rather not do that. Not because they are dishonest. Because the model might say no.
A clean unit economics model tells you three things most founders do not know about their own business.
The first is where the money actually comes from. Not the product or service line you think is the core, but the transaction that is carrying the business. Often these diverge. The offering a founder leads with is not always the one the market is paying for. The model shows you this immediately; the P&L usually hides it.
The second is which assumption you are most attached to. Every model has one number that, if you change it, everything else falls apart. Acquisition cost. Average order value. Retention rate. Gross margin. When you find that number and ask why you are so confident in it, the answer is usually: because the business does not work if it is wrong.
That is not confidence. That is hope, formatted as a spreadsheet.
The third is what scale actually does to the margins. A business can look viable at small volume and structurally broken at large volume. Or it can look unworkable now and only make sense at a size you have not earned yet. Both are important to know. Both are invisible until you run the numbers at different volumes and watch what happens.
The model is uncomfortable because it is honest in a way that investor pitches, team meetings, and founder narratives are not required to be.
You can tell a room that you are building something important and receive genuine agreement. You cannot tell a spreadsheet the same thing and expect it to cooperate.
There is also a timing problem. Early stage, the data does not exist. So founders approximate, and the approximations feel arbitrary, so they set the model aside and focus on things that feel more real: product, customers, hiring. By the time the data exists, the model has been absent long enough that building it now feels like an audit. And audits are only commissioned when something is already wrong.
This is exactly backwards. The model is most valuable when the data is thin, because that is when assumptions are doing the most work. Naming the assumptions early, even roughly, creates a set of specific things to test as the business develops. Without it, you spend the first two years learning things the model would have told you in a week.
A model that breaks is not a failure. It is the entire point.
When we evaluate a venture, we build the economics until the thesis either holds or it does not. Most do not. That is not a sign that we chose bad ideas. It is a sign that the discipline is working. The goal is to find the break in the spreadsheet, not in the market. Fixing a thesis costs nothing. Unwinding a business costs everything.
The questions a breaking model asks are always more useful than the answers a comfortable model provides.
Why does customer acquisition cost this much? Why does retention not hold past this point? Why does margin compress at exactly this volume? Each of those is a real question about how the business works, surfaced before you have spent the capital to find out the hard way.
Build the model before you think you have enough data to build it. Use your assumptions explicitly: label them, own them, share them with someone who will challenge them. Run it at three different volume levels and watch where it behaves unexpectedly. Find the one number the whole thing depends on and ask yourself what you actually know about it.
Then update it. Not once a quarter. Continuously, as the business generates real data to replace the assumptions. The model should get more accurate over time, not more optimistic.
A business whose economics you understand is a business you can run. A business whose economics you are still figuring out is a business that is running you.
The numbers are not the enemy of the vision. They are the only honest version of it.
Empirica builds new ventures from zero and rebuilds existing businesses from the inside. The work always starts with the economics.