The Real Cost of a Custom Web App in 2026

"How much does a custom web app cost?" is the first question every founder asks and the one most agencies answer least honestly. The truthful answer is a range — realistically $5,000 to $40,000+ in 2026 for the kinds of products most startups and SMEs commission — but the range itself is not useful. What is useful is understanding which decisions move you along it, because after scoping and pricing a great many of these projects at Fajarix, we can tell you the cost is set by perhaps eight decisions, most of which get made casually in the first meeting.

The Three Tiers, Honestly Described

$5,000-$12,000: A Focused Tool That Does One Thing

One user role, one core workflow, 5-10 screens, standard authentication from an off-the-shelf provider, one or two third-party integrations (Stripe, a CRM, an email service), and an operations backend that is a hosted admin tool rather than custom UI. Think: a client portal for a services business, an internal approval workflow, a booking tool, a calculator-plus-checkout product. This tier assumes you accept sensible defaults — component-library UI styled to your brand rather than a bespoke design system, and a proven stack (for us, Next.js, Postgres, and a managed host) with no exotic infrastructure.

$12,000-$25,000: A Real SaaS MVP

This is the most common tier we build. Two user roles (say, customers and staff), 10-20 screens, subscription billing with plan logic, role-based permissions, custom business logic that is genuinely yours (matching, scoring, scheduling, quoting), file handling, transactional email flows, and a modest custom admin view. Custom design on the screens that matter, systematised components everywhere else. Most B2B SaaS products people actually launch live here, whatever their pitch decks imply.

$25,000-$40,000+: A Multi-Sided Platform or Regulated Product

Three or more user roles, marketplace mechanics or multi-tenant architecture, real-time features (live chat, collaborative editing, live dashboards), complex integrations with legacy or poorly-documented systems, compliance work (audit logs, data residency, HIPAA/GDPR tooling), heavy reporting, or serious data migration from an existing system. The plus sign is doing real work in that heading: a two-sided marketplace with payments escrow and dispute flows, or a healthcare product with compliance requirements, moves past $40k without anything being gold-plated.

The Eight Decisions That Actually Move the Number

1. Number of user roles. The single biggest multiplier, and the one founders underestimate most. Each role brings its own flows, permissions, empty states, and edge cases. Going from one role to two rarely adds 30% — it usually adds 60-80%.
2. Authentication and permissions complexity. "Users log in" is cheap. "Organisations have teams, teams have roles, roles have granular permissions, and enterprise customers need SSO" is a project inside your project. SSO/SAML alone routinely adds $3,000-6,000 of properly-tested work.
3. Integrations — specifically, the quality of the other side. Stripe is a day or two because its API and docs are excellent. A regional logistics provider with a SOAP API, no sandbox, and a PDF for documentation can consume two weeks. We price integrations by the quality of the counterpart API, not by the number of them, and we ask to see the API docs before quoting.
4. Real-time requirements. The moment anything must update live — chat, presence, collaborative state, live order tracking — you add websocket infrastructure, reconnection handling, and a class of bugs that only appear under concurrency. Budget 15-25% extra the moment the word "live" enters the spec, and interrogate whether polling every 30 seconds would honestly be fine (it usually would).
5. Design maturity on arrival. Arriving with validated Figma designs versus a verbal description is a swing of thousands of dollars. Not because design is padding — because undesigned products get designed during development, through rework, which is the most expensive design process ever invented.
6. Reporting and dashboards. "A simple dashboard" never is. Aggregations, filters, date ranges, exports, and charts across large datasets involve real backend work and query optimisation. A genuinely useful analytics section is $3,000-8,000 on its own.
7. Data migration. If your users and history currently live in spreadsheets or a legacy system, migrating them cleanly — with validation, deduplication, and a rollback plan — is a workstream, not a task. It is also the item most often left out of competing quotes, which is worth remembering when one quote seems suspiciously low.
8. Compliance and audit requirements. Audit trails, data retention policies, right-to-deletion tooling, and access logging are entirely buildable, but they touch every feature and are dramatically cheaper to include from the start than to retrofit.

Where Teams Overspend

We spend a surprising amount of first-meeting time talking clients out of spending. The recurring offenders:

• Custom admin panels before product-market fit. A hosted admin tool or direct database UI covers the first year for most products. We have watched teams sink 25-30% of an MVP budget into internal screens their two staff members use twice a week.
• Rebuilding commodity features. Custom auth, custom file storage, hand-rolled billing logic, in-house analytics. Every one of these has a best-in-class service costing tens of dollars a month. Custom-building them buys you thousands of dollars of liability and no differentiation. Our rule: build what makes you different, buy everything else.
• Premature scale engineering. Microservices, Kubernetes, multi-region databases — for a product with zero users. A well-built monolith on managed infrastructure serves your first hundred thousand users for the vast majority of products. Architecture astronautics at MVP stage is the most expensive form of procrastination we know.
• A bespoke design system for an unvalidated product. Full custom design language, animation systems, dark mode, design tokens — before anyone has confirmed users want the product. Spend design money on the two or three screens where your value lands.
• Native mobile apps alongside the web app, at launch. Two more codebases, app-store review cycles, and release overhead — for reach a responsive web app already had. Unless the product is camera- or notification-dependent, mobile apps are a v2 decision.

The Costs Nobody Puts in the Proposal

A quote is not the cost of owning software. Budget for these from day one:

• Maintenance: 15-20% of the build cost, annually. Dependency updates, security patches, framework upgrades, small fixes, and the steady stream of "can we just change..." This is not a vendor upsell; unmaintained apps rot measurably within a year, and the ones that skip two years of updates eventually pay for a partial rewrite instead.
• Third-party services: $150-800/month for a typical SaaS. Hosting, database, email, auth, monitoring, file storage, and whatever APIs you integrate. Individually small, collectively real, and they scale with usage.
• Your own time. A custom build needs a decision-maker reviewing progress weekly and answering questions within a day. Projects where the client disappears for three weeks mid-build cost more — not as a penalty, but because guessed requirements get rebuilt.

A Worked Example

To make the tiers concrete, here is a composite of a project shape we see constantly. A logistics company wants a customer portal: clients log in, create shipment requests, upload documents, track status, and receive email updates; staff review requests and update statuses. As first described — "a portal for our clients" — the founder expected the bottom tier. Walking through the eight decisions told a different story: two user roles (immediately out of tier one), a status workflow with notifications at each stage, document uploads with validation, and an integration with their existing transport management system whose API documentation turned out to be an email thread from 2019. Realistic scope: $18,000-24,000 over eight to ten weeks.

Then we cut. Staff kept working out of a hosted admin view instead of a custom dashboard (saving roughly $4,000), notifications shrank from six trigger points to the two customers actually care about — request received and shipment delivered — and the legacy integration became a phase-two item, replaced at launch by a twice-daily CSV import that one staff member runs in ninety seconds. Revised build: about $13,500, shipped in six weeks, and the client got real usage data before committing to the expensive integration. Three months later the usage data killed one planned phase-two feature entirely and promoted another nobody had thought of. That sequence — scope honestly, cut deliberately, let production data spend the next dollar — is worth more than any rate negotiation.

How to Keep the Number Down Without Getting a Worse Product

Three moves consistently save money without degrading the outcome: cut scope, not quality — a smaller product built properly beats a full product built badly, because the badly-built one gets rebuilt within eighteen months and you pay twice; phase the delivery — ship the core workflow in phase one, let real usage reorder the phase-two backlog (it always reorders it); and arrive with decisions made — a written scope, chosen examples of products you like, and a named decision-maker will do more for your budget than any negotiation over rates.

And one honest warning in the other direction: if a quote for a multi-role platform with integrations and billing comes in at $4,000, the number is not a bargain — it is a description of what you will receive. The expensive version of cheap software is the second team you hire to rebuild it.

If you want a real number instead of a range, our web development team will scope your project properly — decisions, tiers, and trade-offs on the table. Book a free consultation and bring whatever you have, even if it is a napkin sketch.