How to calculate unit economics with Claude: are you earning on each customer or losing?
Unit economics in plain words: how to work out in one conversation with Claude whether you earn on a customer or lose money. Five numbers, a ready prompt, and an LTV to CAC check.
You are selling, money is coming in, and yet the account is empty at the end of the month. Usually the reason is not too few customers, it is that each sale earns you less than you think. Unit economics is the calculation that shows whether you make or lose money on a single unit: one customer, one order, one subscription. Claude (an AI assistant from the company Anthropic, a program that understands text and does the math on your numbers) helps you build that calculation in a single conversation, with no spreadsheet formulas and no finance hire.
What is unit economics in plain words?
A unit is the single piece of your business that you earn on. For a coffee shop it is one cup sold, for an online store one order, for a subscription service one paying customer. Unit economics answers one question: how much you earn or lose on that unit after every cost tied to it. If you are in the black on each customer, the business can scale: more customers, more profit. If you are in the red, then the more you sell the faster you sink, and no amount of advertising will save it.
Which numbers should you gather before talking to Claude?
The AI will be as accurate as your inputs are. Before the conversation, collect these for one typical customer or order:
- the price you sell at (average order value)
- the cost of the goods or of delivering the service
- acquisition cost: how much you spend on advertising to get one customer (this is called CAC, customer acquisition cost)
- how many times an average customer buys from you before they leave
- fixed costs you pay no matter what: rent, salaries, software subscriptions
Do not worry that the numbers are rough. Starting with an estimate and refining beats not counting at all.
How do you calculate unit economics with Claude in one conversation?
Open claude.ai and give the AI a role and your data in one message:
"You are a financial analyst. Help me calculate the unit economics of my business. Average order value is 2000 rubles, cost of goods is 800 rubles, I spend 15000 rubles a month on advertising and get 30 customers from it, and an average customer buys from me 3 times a year. Work out how much I earn per customer and explain every step in plain words."
Claude lays the calculation out step by step: what it costs to acquire one customer, what they bring in per purchase, and over their whole lifetime. You see not a bare number but where it comes from, and you can argue with each step.
How do you tell whether the unit economics add up?
The key comparison is this: how much a customer brings over their whole lifetime versus how much it cost to acquire them. The first is called LTV (lifetime value, the total money a customer leaves with you while they keep buying), the second is that same CAC. A healthy rule of thumb for a small business: a customer should bring in at least three times what it cost to acquire them. Ask Claude to work out that ratio and tell you straight:
"Calculate the LTV to CAC ratio from these numbers and tell me honestly whether my economics are healthy or I am losing money on every customer."
If the ratio is below one, you are paying out of your own pocket for each customer. If it is around two, the business is alive but fragile. If it is three or higher, you have room to grow.
How do you test "what if" scenarios?
The strength of a conversation is that Claude remembers all your numbers, so you can spin scenarios without recalculating by hand:
- "What happens to profit if I raise the price by 10 percent and sales drop by 15?"
- "How many customers a month do I need to break even, given rent of 50000?"
- "Advertising just doubled in cost. At what price am I still in the black?"
In five minutes you play out decisions that normally eat an evening in a spreadsheet, and you walk into them with numbers rather than by feel.
What should you never trust an AI with in the math?
Two honest limits.
First: Claude calculates from the data you give it. If the cost of goods is understated or you forgot the payment processor's fee, the answer will be neat and wrong. Garbage in, garbage out. So gather your real costs first, every one of them.
Second: for serious decisions (a loan, investment, a tax model) the AI is a first draft, not the final word. Check the key numbers yourself or with an accountant before you take them to a bank or an investor.
Where do you start today?
Take one of your products and calculate its unit economics in a conversation with Claude right now: gather the five numbers above, paste them into the prompt and ask for a step-by-step breakdown. Ten minutes later you will know whether you earn on each customer or subsidize them.
If you want to run the rest of your business tasks with an AI the same way, see our free materials in the guides and take the first free lesson about your digital twin UNIT. They show, on real examples, how Claude fits into an ordinary founder's working day.
AGINE Academy is an independent product, not affiliated with Anthropic. Claude is owned by Anthropic.
Questions
No. You only need five numbers for one customer: price, cost of goods, advertising spend, purchase frequency and fixed costs. Claude lays out the calculation step by step and explains each one in plain words.
CAC (customer acquisition cost) is how much you spend to get one buyer. LTV (lifetime value) is how much money a customer brings over the whole time they keep buying. The economics are healthy when LTV is at least three times CAC.
Claude calculates accurately from the data you give it, but it does not verify that the data is true. Gather your real costs, including fees and taxes, and double-check key figures with an accountant before any important decision.
To learn whether you earn on each customer or lose money. If you lose, growing sales only speeds up the losses, and that only shows up in the calculation, not in the revenue figure.