Custom Web App Cost: An Honest 2026 Breakdown
2026-08-24 · 7 min read
Ask three agencies what a custom web application costs and you will get three numbers, all vague, all large. The international norm for a "real" custom build sits around $60,000 and nine months. In Portugal, the standard answer starts at €15,000 and climbs from there.
Those numbers are not lies. They reflect how most agencies work: a project manager, a designer, two or three developers, a QA person, and weekly meetings to coordinate them all. You pay for every hour of that coordination.
But the numbers are not laws of physics either. This article breaks down what actually drives the cost of a custom web app, so you can estimate your own project before anyone sends you a proposal.
What actually drives the price
Strip away the jargon and four factors determine most of the cost of any web application.
1. Screens
Every distinct screen — a dashboard, a settings page, an order form, a report view — has to be designed, built, and tested. A rough rule: a focused internal tool has 5–10 screens; a customer-facing product has 15–30; a marketplace or platform has 40+.
Screens are the most honest unit of scope. If a proposal doesn't count them, ask why.
2. User roles and permissions
One type of user is simple. The moment you add roles — admin, manager, customer, guest — every screen potentially behaves differently for each role, and the testing surface multiplies. Two roles roughly doubles the permission logic; four roles can quadruple it.
3. Integrations
Each external system your app talks to — accounting software, a CRM, a shipping provider, an email platform — is a mini-project. A well-documented modern API (Stripe, most SaaS tools) might take a day or two. A legacy system with poor documentation can take weeks. Count your integrations before you count anything else.
4. Payments
Taking money changes everything. Payments bring refunds, disputes, invoicing, tax rules, failed-card handling, and compliance. A payment-enabled app is reliably 30–50% more work than the same app without payments. If you can launch without payments and add them in version two, you save real money.
Rough numbers for 2026
Here is how these factors translate into price bands, comparing the traditional agency model with what a single senior engineer using modern AI tooling can deliver:
| Project type | Typical agency (intl.) | Typical agency (PT) | Feitura |
|---|---|---|---|
| Internal tool, 5–10 screens, 1–2 roles | $25,000–50,000 | €15,000–25,000 | quoted per project |
| Customer portal, 10–20 screens, payments | $50,000–90,000 | €25,000–50,000 | scoped and quoted per project |
| E-commerce with custom logic | $40,000–80,000 | €3,500–10,000 (template) / €20,000+ (custom) | quoted per project |
| Marketing site with a booking or quote engine | $10,000–25,000 | €3,000–6,000 | quoted per project |
The Feitura column is deliberately number-free: we quote every project individually, and the quote is a real figure, not a moving range. Timelines follow the same pattern: agency projects run 3–9 months; ours are dramatically faster — we'll tell you exactly how fast after a two-minute brief.
How "dramatically faster" works — honestly
Skepticism is fair here, so let's be precise about where the difference comes from.
What changed: one senior engineer (10+ years) working with AI-accelerated tooling now produces what previously required a small team. The AI handles the repetitive 70% — boilerplate, standard CRUD screens, test scaffolding, documentation. The engineer handles the 30% that actually matters: architecture, security, edge cases, and the judgment calls the AI gets wrong.
What didn't change: the code is reviewed, tested, and owned by a human who has been shipping production software for over a decade. AI without senior review produces plausible-looking software that fails in production. Senior review without AI produces good software slowly. The combination is the point.
What you lose: nothing structural — but you should know the trade-offs. A one-engineer pipeline has less redundancy than a team; that's mitigated by clean, documented, standard-stack code that any competent developer can take over. You also won't get a 40-page requirements document. You'll get working software you can react to instead.
When you don't need a custom app
An honest breakdown includes the cases where the answer is "don't build one."
A spreadsheet is enough when fewer than five people use the data, the process is linear, and nobody needs real-time access from outside the office. Google Sheets with some formulas beats a €10,000 app you didn't need.
No-code tools (Airtable, Notion, Zapier) are enough when your workflow matches what the tool already does, you're under their usage limits, and monthly fees of €50–200 are acceptable indefinitely. The trap: costs scale with users, and the day you need something the tool can't do, you're rebuilding from zero.
You need custom software when the process is your business advantage, you're hitting no-code limits, multiple roles need different views of the same data, or per-seat SaaS pricing has crossed what a one-time build would cost.
A useful test: if you're paying more than €300/month in tool subscriptions to duct-tape a workflow together, a custom build likely pays for itself within 12–24 months — and you own it.
FAQ
Why do agencies quote €15,000+ for something you deliver for far less? Headcount. A five-person team spends a large share of every project on internal coordination. We removed the coordination, not the seniority.
What about maintenance? Budget for it. Market rates run €30–300/month depending on complexity. Our Care plan covers updates, monitoring, and small changes. Any quote that ignores maintenance is incomplete.
Can I start small? You should. The cheapest project is the one scoped to version one only. Launch the core workflow, learn from real usage, then extend. This is also why fixed large quotes are risky: they price features you may never need.
Who owns the code? You do. Full source, standard stack, no lock-in. If we disappear tomorrow, any developer can continue.
Estimate your own project
Count your screens, roles, integrations, and whether you take payments. Those four numbers put you within range before you talk to anyone. If you want a concrete figure for your specific project, describe it at /estimate and you'll get a real number, not a "let's schedule a discovery call."
Comments (0)
No comments yet — start the conversation.
Sign in to join the conversation. Sign in · Create an account