Why the monthly SaaS fee is the wrong number and when building your own gets cheaper

What you really pay is not on the monthly invoice

Open your company’s credit card statement and count the software subscriptions. CRM, project planning, accounting, newsletters, reporting, ticketing, HR, document storage, e-signatures, password management. In many companies with 50 people nobody knows the total, because it is spread across ten or more invoices and every single one looks harmless.

The question that rarely gets asked: What does the hardware cost that this software runs on?

A dedicated server at Hetzner in Germany costs a low three-figure euro amount per month at list price. A comparable virtual server at AWS sits at roughly one US dollar per hour at list price, so several hundred dollars per month. Both are a long way from what you pay for a CRM with 50 users.

That is not fraud. The price of a subscription has little to do with server cost. You pay for the vendor’s development, support, sales and margin. The price follows what the software is worth to you, not what it costs the vendor. Which is exactly why you should check whether that value holds for you.

Why renting used to be the right answer

Software subscriptions took over between 2005 and 2015 because they solved a real problem. Custom software was expensive to build, expensive to run and expensive to keep current. The vendors’ argument was correct: one platform for a thousand customers is cheaper than a thousand individual installations.

That argument rested on three assumptions:

1. Building is expensive. A custom system needed a team for months. The investment only paid off after years.
2. Running is expensive. Servers needed their own people to apply updates, check backups and be reachable at night.
3. Keeping current is expensive. Security updates, new features and adjustments to regulations ate a share of the original build cost every year.

All three assumptions have shifted since 2022, but they have not disappeared. AI coding tools speed up the standard parts of a system considerably: forms, database access, interfaces to other programs, tests. How much depends on the project. The hard parts stay hard. Business logic that only exists in your people’s heads. Data migration from the old system. Permissions and roles. Operation and maintenance have become easier with rented servers and standard tooling, but they still need someone who takes care of them.

Subscription prices have not fallen in the same period. Salesforce raised list prices for its Enterprise and Unlimited editions by an average of six percent as of August 2025. Other vendors have made similar moves. The vendors’ costs go down, your prices go up. That gap is what this article is about.

The three-year calculation

Most comparisons fail on method. A monthly fee is put against a one-time build sum. That is like comparing the monthly rent of an apartment with its purchase price while leaving out how long you will live there.

Instead, calculate both sides over three years. Three years is long enough for a custom build to show its value and short enough for the assumptions to still hold.

Item	Subscription	Custom build
Licenses	price per user times users times 36 months	none
Price increases	carry forward the increases of the last two years	none
Growth	every new employee needs a license	no additional cost
Build	none	one-time, including data migration
Servers	included in the price	rent per month times 36
Maintenance	included in the price	hours per year times hourly rate times three
Integration with other systems	often only in the higher tier	part of the build
Exit	export, migration, parallel operation	code and data are yours

Two items are almost always forgotten on the subscription side. First, integration: If your CRM needs to talk to your inventory system, you frequently need the next tier up or another subscription in between. Second, the exit: Getting your data out of a system, including history and attachments, is rarely free and never fast.

On the custom build side, maintenance is almost always underestimated. A system nobody maintains is a security risk after two years. Budget a fixed amount per year, even when nothing is broken.

As a rough orientation from our projects: Building pays off when the build sum is below two to three years of subscription fees and the tool is central to your business. For peripheral tools that three people use occasionally, it almost never pays off.

The costs that appear on no invoice

The monthly invoice is only the surface. Below it are costs that are built into the vendors’ pricing models.

Per-user licenses. Your company grows from 50 to 80 people. Your software costs grow with it, although the load on the servers barely changes. The price per user is a levy on your growth. Why this model is under pressure right now is covered in The End of Per-Seat Pricing.

Limited interfaces. On the standard plan you may only send a limited number of requests per minute to the interface. For unlimited requests you need the enterprise tier. Connecting to your other systems becomes a surcharge.

Export as a hurdle. A complete export with relationships between records, attachments and history is often only possible on the most expensive tier or through a paid interface. The Data Act improves your position when switching, but it does not make the move free.

Proprietary data formats. Every tool stores your data in its own schema. Switching to a competitor costs months of alignment and adjustment. The vendors know this and factor it into their price increases.

Where the data lives. With many vendors your customer data sits on servers outside the EU or with a US company operating an EU data center. What that means legally depends on the individual case. With your own servers in Germany the question does not arise. More on that in Sovereign Cloud Is an Architecture Claim.

Features you do not need. You use a small share of the features and pay for all of them. Every quarter new ones arrive and justify the next price round.

What a system of your own brings, apart from money

The cost argument is the weakest argument for building, because it is the easiest to argue away. Three others matter more.

Your data stays with you. Customer data, process data, history. All of it sits on your servers. No vendor changes the privacy terms, no acquisition changes the storage location. And you may do with your data whatever you want, such as run your own analyses or train your own AI models. Many subscriptions restrict that contractually.

The software follows your process. With a subscription you adapt your process to the software. Your competitor gets the same feature on the same day. A system of your own reflects how you actually work. How to do that without depending on the developer is covered in Automation Without Handcuffs.

Code stays. A subscription disappears as an expense at the end of the month. Your own code stays and gets better with every change. Whether you can capitalize self-built software on your balance sheet is a question for your accountant. It is possible, but tied to conditions.

When a subscription is still the right choice

Subscriptions are not wrong across the board. In four cases they remain the better choice.

Network effects. Slack works because your partners use Slack too. For tools whose value depends on the network outside your company, building is pointless.

Changing regulations. Accounting software that is automatically adjusted to new tax rules has a real maintenance advantage. You only match that yourself with permanent effort.

Short-term use. You need a tool for three months during a migration. Building does not pay. Renting does. But short means short, not “we will cancel at some point”.

Nobody takes care of it. If there is nobody in your company who maintains a system of your own or has it maintained, the subscription is the safer choice. An unmaintained system is more expensive than any subscription.

For the core tools on the other hand, meaning CRM, ticketing, reporting, document storage and internal workflows, the comparison is almost always worth doing. It depends on your starting point, but the math favors the subscription less often than it did three years ago.

What you can check this week

1. List all subscriptions. Credit card statement, incoming invoices, purchasing. Record price, number of licenses and notice period.
2. Count active users. Most tools show who has logged in during the last 30 days. Cancel licenses without a login right away.
3. Calculate the three-year total. For every subscription, using the price increases of the last two years. Sort by total.
4. Pick the three largest items. For each one, ask the question from the table above: What would a system of your own cost, including maintenance?
5. Test the exit. Try a complete export of your data from the most expensive tool. Whatever is missing tells you how tightly you are bound.

If you need help with the estimate in step four, Custom Development describes how we approach such projects. For a first assessment without any commitment, a short message is enough.