What does your own platform cost? Development, operations and what AI changes

Key takeaways

• The development cost of a platform is the effort in hours multiplied by the hourly rate. According to freelancermap, IT freelancers in software and web development quote a median of €90 per hour.
• The effort is driven mainly by user roles, workflows, payment flows and integrations. Starting with a lean MVP keeps the first investment small and lets you learn from real users early.
• Operations come with manageable base costs for servers and email delivery. On top of that there are fees that grow with your revenue, such as 1.5% + €0.25 per card payment with Stripe.
• AI speeds up development measurably. In controlled studies with GitHub Copilot and at Google, developers needed 21% to 56% less time depending on the task, with the biggest gains on new projects.

As of October 2026

You have an idea for a platform: customers book appointments at several studios, pay online, and each studio automatically receives its share. Before you ask for quotes, an obvious question comes up: What does something like this cost, and what should you expect after launch? This guide shows what the costs are made of, which running expenses you can plan for and what AI changes about the numbers today.

How a platform differs from a website

A website presents content. A platform connects people and processes: it has user accounts with different roles, stores data, handles payments and gives operators a back office. A marketplace brings two sides together, for example customers and providers, and money often flows between them.

In our projects, most of the work goes into exactly this logic: permissions, workflows, integrations and billing. The visible pages matter for the first impression, but they usually account for the smaller part of the effort.

What drives development costs

At its core, the calculation is simple: effort in hours times hourly rate. The Freelancer-Kompass 2026 by freelancermap gives a reference point for the rate. It puts the median for IT freelancers in software and web development at €90 per hour, based on a survey of freelancers in Germany, Austria and Switzerland. A full working week of one person, 40 hours, therefore comes to about €3,600. Every week a project finishes sooner saves a noticeable share of the budget.

The effort is driven mainly by these building blocks:

• Roles and permissions: guests, providers, staff and administrators each see and do different things.
• Workflows: booking, cancelling, approving, billing. Every step brings its own special cases.
• Payments: a simple purchase is quick to connect. Distributing money automatically to several partners takes more work.
• Integrations: maps, email, calendars, accounting or even devices on site.
• Languages and devices: multiple languages and native apps for iOS and Android extend the scope.
• Design: your own design that fits your brand.

Start small: the MVP

Founder and author Eric Ries describes the minimum viable product, or MVP, as the version of a new product that lets a team learn the most about its customers with the least effort. For a platform, this means the core workflow works completely, and everything else follows in later stages.

For a booking platform, that could look like this: find a location, book an appointment, pay online, receive a confirmation by email. Reports, additional languages or the connection of devices can be planned as separate stages and prioritized based on the first experiences. This keeps the initial investment manageable, and every further stage builds on real usage.

What operations cost

After launch, running costs come in. Some are fixed, others grow with your revenue. The following figures are taken from the providers' pricing pages, as of October 2026:

• Server: a cloud server with two CPU cores and 4 GB of RAM costs €19.49 net per month at Hetzner, plus €0.50 for the IPv4 address (CPX22 model). In our experience, one such server is often enough to launch a platform.
• Card payments: Stripe charges 1.5% + €0.25 per payment for standard cards from the European Economic Area. For a €90 booking, that is €1.60. A SEPA Direct Debit costs €0.35.
• Payouts to partners: with Stripe Connect, a platform distributes money to providers. If the platform lets Stripe handle pricing directly with the partner accounts, the partners pay the usual Stripe fees and, according to Stripe, the platform pays no additional Connect fees. If the platform sets the pricing itself, each partner account that is active in a month costs €2, and each payout costs 0.25% + €0.10.
• Maps: Google Maps Platform includes 10,000 free calls per month for standard features such as dynamic maps. After that, 1,000 map loads cost 7 US dollars.
• Email: the delivery service Resend sends up to 3,000 emails per month for free, with a maximum of 100 per day. 50,000 emails cost 20 US dollars per month.

On top of that comes maintenance and further development: updates, new features your users ask for and adjustments to new versions of the services you use. Planning a fixed budget for this from the start lets the platform grow with your business.

What AI changes about the numbers

AI assistants are now part of everyday software development. In the 2025 Stack Overflow Developer Survey, 84% of respondents said they use AI tools or plan to, and 51% of professional developers work with them daily. Several studies have measured how much time this saves:

• In an experiment by Microsoft Research, GitHub and MIT, developers using the GitHub Copilot assistant built a small web server in 71 instead of 161 minutes on average. That is 56% less time.
• At Google, software engineers solved a realistic everyday enterprise task with AI support in 96 instead of 114 minutes. The researchers estimate the time saving at about 21%.
• In three field experiments at Microsoft, Accenture and a large electronics manufacturer with a total of 4,867 developers, AI users completed 26% more tasks.

AI has the strongest effect on new projects and clearly defined tasks. For experienced professionals working in very large codebases that have grown over years, the research organization METR even measured longer completion times in early 2025. METR now describes these results as outdated itself, and data from 2026 points to a time saving. For a new platform, this is good news: it is built from scratch, exactly where AI helps the most.

Speed alone does not make a good platform. A study by Carnegie Mellon University shows that in open-source projects, code complexity increased alongside speed after an AI editor was introduced. Experience in architecture, automated tests and careful review therefore remain decisive. AI handles the routine work, people make the decisions. On our booking platform for Wasserplasma Experience, for example, more than 1,400 automated tests safeguard the money flow.

How to prepare a reliable estimate

The clearer your project is described, the more accurate the estimate. Start with one sentence on the goal: who uses the platform, and what problem does it solve? Then describe the roles, meaning who signs up, who manages and who earns money, and outline the core workflow in a few steps, from the first visit to the payment.

It also helps to separate what must be there at launch from what can come later. Name any existing systems that should be connected, as well as your budget range and a target date. With this information, the effort per building block can be estimated, and it becomes clear what fits into a first stage.

In practice

Two platforms we have recently built show the range. For Wasserplasma Experience, we created a booking platform: guests book and pay online, the device at the studio unlocks for the booked slot, and every partner studio is settled monthly and paid out via Stripe Connect. byclaw.io is our own product, an AI shopping agent that searches for products, compares prices and makes recommendations via WhatsApp, directly in Claude and ChatGPT, or on the web. One platform connects booking, payment and devices on site, the other connects AI with messaging and product catalogs.

Frequently asked questions

Roughly how much does your own platform cost?
That depends mainly on the scope: roles, workflows, payment flows and integrations determine the effort, the hourly rate determines the price per hour. A lean MVP costs a fraction of a fully built platform. You can get a free estimate for your project with our AI project estimator.

How long does it take to develop an MVP?
That also depends on the scope. Our standard: a project that would traditionally be planned for six months, we deliver in eight weeks. Proven building blocks and AI for routine work make this possible.

What are the minimum running costs?
You should plan for a server, a domain and email delivery. For smaller platforms, in our experience, these base costs are in the double-digit euro range per month. Payment and map fees only arise once your platform is in use, and they grow with your revenue.

Your own platform or off-the-shelf software?
Off-the-shelf software works well for processes that are the same at many companies. Your own platform pays off when your workflow is what makes your offer special, when money flows between several parties, or when you want to connect partners, devices and data your own way. Then the platform belongs to you and grows with your business.

Conclusion

The costs of a platform follow a clear logic: effort times hourly rate for development, plus manageable base costs and revenue-based fees in operation. A lean MVP keeps the first investment small and delivers real experience early. AI speeds up development measurably, especially on new projects. With experienced developers who keep architecture and quality in view, the result is a platform that launches faster and grows on a solid foundation.

What would your platform cost?

Describe your idea in our AI project estimator. You will receive a free estimate by email, and we will get back to you within 24 hours.

Learn more about our services for platforms and marketplaces and SaaS products. Or simply write to us at our email address.

Sources

1. freelancermap: Freelancer-Kompass 2026, study page, accessed October 12, 2026. https://www.freelancermap.de/freelancer-kompass
2. freelancermap: blog post on IT freelancer hourly rates, July 2, 2026, and press release of July 23, 2026. https://www.freelancermap.de/blog/stundensatz-it-freelancer/, https://www.freelancermap.de/media/press_release/2026-07-23-IT-Freelancer.pdf
3. Eric Ries: "Minimum Viable Product: a guide", Startup Lessons Learned, August 3, 2009. https://www.startuplessonslearned.com/2009/08/minimum-viable-product-guide.html
4. Hetzner: cloud servers, Regular Performance product page, prices accessed October 12, 2026. https://www.hetzner.com/de/cloud/regular-performance/
5. Stripe: "Preise und Gebühren" (pricing, Germany), accessed October 12, 2026. https://stripe.com/de/pricing
6. Stripe: "Preisinformationen | Stripe Connect" (Connect pricing, Germany), accessed October 12, 2026. https://stripe.com/de/connect/pricing
7. Google for Developers: "Google Maps Platform core services pricing list", last updated October 8, 2026. https://developers.google.com/maps/billing-and-pricing/pricing
8. Resend: "Pricing", accessed October 12, 2026. https://resend.com/pricing
9. Stack Overflow: "2025 Developer Survey: AI", July 29, 2025. https://survey.stackoverflow.co/2025/ai
10. Sida Peng, Eirini Kalliamvakou, Peter Cihon, Mert Demirer: "The Impact of AI on Developer Productivity: Evidence from GitHub Copilot", arXiv, February 13, 2023. https://arxiv.org/abs/2302.06590
11. Elise Paradis et al. (Google): "How much does AI impact development speed? An enterprise-based randomized controlled trial", arXiv, November 11, 2024. https://arxiv.org/abs/2410.12944
12. Kevin Zheyuan Cui, Mert Demirer, Sonia Jaffe, Leon Musolff, Sida Peng, Tobias Salz: "The Effects of Generative AI on High-Skilled Work: Evidence from Three Field Experiments with Software Developers", Management Science, February 27, 2026. https://doi.org/10.1287/mnsc.2025.00535
13. METR: "Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity", July 10, 2025. https://metr.org/blog/2025-07-10-early-2025-ai-experienced-os-dev-study/
14. METR: "We are Changing our Developer Productivity Experiment Design", February 24, 2026. https://metr.org/blog/2026-02-24-uplift-update/
15. Hao He, Courtney Miller, Shyam Agarwal, Christian Kästner, Bogdan Vasilescu (Carnegie Mellon University): "Speed at the Cost of Quality: How Cursor AI Increases Short-Term Velocity and Long-Term Complexity in Open-Source Projects", MSR 2026. https://arxiv.org/abs/2511.04427