What does one customer actually cost you to serve?
Your usage bill arrives as one number a month, keyed to an account. Your revenue is keyed to customers. Neither one carries a field that identifies the other, so nothing joins them, and the gross margin you quote is an estimate. Answer five questions, enter your numbers, and this page does the join.
Built by ClarIQ Advisory, a founder-run firm that does the monthly numbers for growing businesses: the close, the cash forecast, and a plain-English read of what they mean. This is one of our free tools and it asks nothing of you.
Step 1
Five questions. No spreadsheet.
Check the ones you could answer right now, out loud, without opening anything.
You can answer 0 of 5. That is the normal starting point, and question five is the one that changes a number on your reporting.
Step 2
Five numbers you already know
Last full month. Round figures are fine; this is a size check, not a close.
Your figures never leave this browser. Nothing you type is sent anywhere or stored, there is no account, and the calculation runs with the network off. Like any site, this one counts page views, and that is the only request the page makes.
Your read
Fill in the three fields to see it
Monthly revenue, your model usage bill, and paying customers. The other two sharpen the picture.
Worth a second opinion on your own figures? Book a free 30-minute call with ClarIQ Advisory and we will tell you on the call whether they are worth a full read. Nothing to bring.
Why you could not answer those five
Not for lack of discipline. The link genuinely is not there. The usage bill is aggregated by month and keyed to an account or a key, and it does not know who your customers are. Revenue is keyed to your customers and knows nothing about usage. Neither system carries a field that identifies the other, so the join has to be built before the question can even be asked, and nobody builds it while there is a product to ship.
Then there is the posting. A bookkeeper who has never priced an AI product sees a monthly bill from a technology vendor and books it where every other technology bill goes: a software subscription, in operating expenses. Gross margin is revenue minus cost of goods sold, so anything below that line is invisible to it. Model spend is a cost of delivering the product. Booked as operating expense, it is missing from the margin on every report you produce, including the one that goes into a data room. That is not a rounding issue. It is the whole line item.
The moment it usually surfaces
Someone asks for the gross margin build-up and what it does at ten times the volume. There is a number in the deck, and there is no build-up behind it, because the number came from an assumption made when the product did one model call per task. Since then the product got better: longer inputs, more steps, retries, an agent loop that makes forty calls where one used to do. Cost per customer climbed quietly and the price did not move. The margin in the deck is describing a product that no longer exists.
What the full read does that this page cannot
This page works on averages you typed in. The real distribution is per customer and it is never flat, so the average is the thing hiding the problem. The full read prices your actual usage records against your actual customer list: cost to serve per customer, gross margin restated with model spend booked where it belongs, the list of customers that lose money, margin by feature, what a free user costs, and the pricing moves in priority order. It ends in your books, so next month's statements are right without anyone redoing this by hand.
Asked often
Why does it matter whether model spend sits in cost of goods or operating expenses?
Gross margin is revenue minus cost of goods sold. Anything posted below that line is invisible to it. Model spend is a cost of delivering the product, so when it is booked as a software subscription in operating expenses, gross margin is overstated by exactly that amount on every report the company produces, including the ones that go to a board or into a data room. It takes ten seconds to check: open your chart of accounts and see which side of the gross profit line the usage bill lands on.
Why can nobody answer these questions already?
Because the join does not exist. The usage bill is aggregated by month and keyed to an account or a key, and it does not know who your customers are. Revenue is keyed to your customers and knows nothing about usage. Neither system carries a field that identifies the other, so the link has to be built before the question can even be asked. That is a data problem, not a discipline problem.
How accurate is this estimate?
It is arithmetic on averages you typed in, so it is exactly as good as those averages. The real distribution is per customer and it is almost never flat, which is the reason the average is misleading in the first place. Treat the output as the size of the question rather than the answer to it. If the gap looks material, the next step is measuring the real spread instead of assuming one.
Does anything I type get sent anywhere?
No. Every figure you enter is used in your browser and nowhere else. Nothing is uploaded, nothing is stored, there is no account, and closing the tab is all it takes to be rid of them. Disconnect from the network and the check still runs, which is the easiest way to prove it. To be exact about what you would see in a network tab: the page counts a view the way every page on this site does, and that request carries no part of what you typed.
Get the real distribution, not the average
Bring one month of usage records and your customer list. We do the join, restate the margin, and hand you the pricing moves in priority order. Start with a free 30-minute call and we will tell you on the call whether your numbers are worth reading.
Book a free 30-minute call See the monthly servicePublished August 7, 2026. This is a first read on figures you entered, not an audit, a valuation, or advice on a specific transaction.