People are paying. Are you making money?

Suppose that small pilot finds people who want the product and are willing to pay. That gives us evidence of demand, but it does not yet tell us whether the money covers the work. The per-request costs we counted earlier now meet a price, a pattern of use and the fixed bills for running the service.

We will use deliberately simple numbers, in whatever currency you use, and leave out taxes and payment processing fees so the calculation stays visible. In a real budget, include both; local tax treatment needs advice appropriate to where you operate.

Say you charge 10 a month. Each customer costs you 3 a month to serve, their share of model calls, storage and email. On top of that you pay 200 a month in fixed costs that do not change with the number of customers, such as your domain, your base hosting plan and the tools you subscribe to.

Each customer leaves you 10 minus 3, which is 7. That 7 is the contribution margin, the part of each payment left over to cover fixed costs. To cover 200 you need 200 divided by 7, about 28.6 customers. Customers come in whole numbers, so you break even at 29. At 28 you are still slightly short.

The 3 deserves a closer look, because it is an average. Not every request your product handles costs you a model call. Opening the app, starting a timer, reading a saved result: these can be ordinary operations without a model call, although their hosting and storage still belong in your costs. A repeated question can sometimes be answered from a cache. Only the requests that actually reach a model API carry that cost. Suppose the typical customer triggers 300 model calls a month at 0.01 each, which gives 3. A customer who triggers 3,000 calls costs 30, three times what they pay. So look at the spread, not only the average, and decide what heavy use leads to: a limit, a higher plan, or a cheaper path for the expensive feature.

Next, keep two numbers apart that people often merge: cash received and revenue. If a customer pays 100 upfront for a year, you have 100 in the bank today, but after delivering the first month you still owe eleven months of service. In this simplified example, the service revenue is spread across the year. Spending the whole 100 as if it were this month's income is how a product with money in the bank ends up unable to serve the people who paid.

Count churn in customers. Say you started the month with 40 paying customers, 4 cancelled, and 6 new ones arrived. Churn is 4 out of 40, 10 percent for the month, and you end at 42. Now you can see that most of your new customers went into replacing the ones who left, which the single figure of 42 would have hidden.

A useful test for any number on your dashboard is whether it would change a decision if it went up or down. Signups or page views can help diagnose discovery and onboarding. They cannot, on their own, tell you whether the product pays its bills. Paying customers, churn, contribution margin and cash on hand answer different questions, so keep the question beside the number.

Together these answer a blunt question: at the current pace, does the product reach the point where it pays for itself before the money runs out? Some people call the two answers default alive and default dead.

Take the numbers above. You have 20 customers, so you earn 140 a month in contribution against 200 in fixed costs, a loss of 60 a month. You have 1,200 in the bank.

If nothing changes, 1,200 divided by 60 gives you 20 months. Now suppose you gain 2 customers a month after churn. The monthly loss shrinks from 60 to 46, 32, 18 and 4. In the sixth month you have 30 customers and earn 10 more than you spend. The total lost along the way is 160, well within 1,200. Under these assumptions the product is default alive.

Change one assumption. If cancellations rise until net growth is zero, you are back to losing 60 a month with no end in sight. Nothing dramatic happened. A few more cancellations a month turned one story into the other.

Here is an exercise. Write down your price, your variable cost per customer and your fixed monthly costs, even as rough guesses. Work out your contribution margin and your break-even customer count. Then take last month and count by hand how many customers started, how many left, and how much cash actually arrived.