The Problem: A $95 Invoice for $0.185 of Work

Last month we took apart a competitor price tracking and weekly report tool that one team was using. It billed $95 every month. We looked at what it actually did: read 40 product pages four times a day, write changed prices into a table, and once a week ask a language model for a short summary and email it. That's it.

We rebuilt the same job one-to-one on our own infrastructure and measured every cent for a month. Real consumption, including servers, storage, bandwidth and model calls: $0.185. The ratio between the invoice and the cost is roughly 500 to 1.

This isn't one bad example. Most of the tools we looked at told a similar story:

• cancelPer-seat pricing: A separate fee for every person who opens the report, while for the server the third reader is free.
• cancelThe tier wall: Going from 50 records to 51 doubles the price; the cost doesn't move even 2%.
• cancelAnnual commitment: Use the tool for three months and quit, and you've still paid for twelve up front.

“A subscription price doesn't measure what the software costs to run. It measures what the vendor can sell it for.”

What a Subscription Price Is Really Pricing

To be fair, there are real expenses inside that $95. A sales team, an ad budget, upkeep of features nobody uses, a support queue, a growth rate promised to investors. None of it has anything to do with reading your 40 products, but all of it is on your invoice.

There's also a quiet line item: unused capacity. Subscription models are built on the assumption that the average customer will barely use the tool. The customer who logs in twice a month subsidizes the one who logs in every day. So if you use it lightly, you're actually paying for someone else's usage.

We're not saying this structure is malicious; it's how you keep a large sales organization running. But we don't have one, and we don't want to build one. So we have nothing else to put on the invoice.

Unit Cost: Where Does $0.185 Come From?

Here's the breakdown for the month we measured:

Two things stand out. First, two thirds of the cost is the model call, and even that is twelve cents a month. Second, the page reading, which is the bulk of the work, stays inside the free tier. What drives the real cost isn't how often you read but how smartly: instead of rendering the whole page in a headless browser, we fetch only the DOM node that holds the price. A page averages 125 KB; a browser session easily runs to 5 MB.

Our Model: Ask, Use, Pay Only the Cost

At Incapon the price list is short and public:

• check_circleNew app, $10: The job you describe in one sentence on the request board is built from scratch and installed in your dashboard.
• check_circleImprovement, $5: A new column, a new channel, a new rule on an existing tool.
• check_circleBug fixes, ₺0: For life. Charging money to fix something that broke makes no sense to us.
• check_circleRunning cost: Server and model consumption at the provider's list price, with no margin added on top.

Under this model, the price tracker above costs $10 + $0.185 in the first month and just $0.185 after that. The $15 of free credit you start with covers it many times over; we don't ask for a credit card, and the credit never expires.

In the dashboard, each tool shows that month's consumption line by line: how many requests, how many tokens, how many pages. When we first designed that screen we had a worry: would people be put off by seeing the cost? The opposite happened. A user who sees that comparing a signature against past signatures costs $0.04, or that transcribing an hour of video costs $0.35, uses the tool more freely. Uncertainty deters people more than a high price does.

Honest Limits

This model has a price for us too. We don't have predictable income the way subscription revenue is predictable; tools that get used less in a given month earn us less. We don't build a sales team, so we grow slowly. And because we keep every tool focused on one job, we're not a fit for anyone looking for a platform that "does everything".

It also takes technical discipline. The moment you commit to showing the cost, you can't hide an inefficient architecture. A browser spun up for no reason or a model prompt longer than it needs to be shows up in the dashboard as a line item right away. That pushes us to write smaller and cleaner every day. Frankly, we don't see that as a side effect; we see it as the main benefit of the model.

Before you renew your next subscription, try a small exercise: write down in three sentences what the tool actually does. Those three sentences could become a request on the board, and we'll measure its cost together.