Web application development cost in 2026: five price bands, plain numbers.

Five price bands. Sub-100 page marketing site: 8,000-25,000 USD over 4-6 weeks. SaaS marketing + product UI: 25,000-60,000 USD over 8-12 weeks. Mid-market product surface: 60,000-120,000 USD over 12-18 weeks. Enterprise platform: 120,000-250,000 USD over 18-26 weeks. Programmatic SEO site at scale: 25,000-90,000 USD over 8-16 weeks. Cost is driven primarily by feature complexity, not page count or framework choice.

A defensible quote covers discovery (paid, 1-2 weeks, 1,500-5,000 USD), design, content modeling, build, testing, deployment, post-launch monitoring (60-90 days), and a written technical spec that documents everything. Anything quoted as "we build, you launch" without those line items is missing 30-50% of the actual work and the price will climb mid-project.

Three bands. Senior independent consultants: 50-150 USD/hour. Mid-market agency rates: 90-200 USD/hour. Tier-one consultancies (Accenture, Deloitte Digital, etc.): 200-400 USD/hour. Offshore firms publish rates 15-40 USD/hour; the production-failure rate matches the rate. The right pick depends on project scope, not just the hourly number, fixed-price beats hourly above 15,000 USD of total scope.

Three structural reasons. Agencies pad for unknowns when they skip discovery. Agencies that include compliance, performance hardening, and post-launch monitoring as defaults cost more than agencies that bill those separately. And agencies with a senior lead engineer charge differently than agencies that staff a junior team behind a senior salesperson. Get three quotes that each include a written spec, and compare specs not prices.

Four phases on any meaningful engagement. Discovery: 1-2 weeks. Build + iteration: 4 weeks (marketing site), 8-12 weeks (SaaS), 12-18 weeks (mid-market product), 18-26 weeks (enterprise). Testing + UAT: 1-2 weeks. Post-launch monitoring: 60-90 days. Anyone promising a 4-week timeline for a real SaaS application is skipping discovery or skipping testing, both are how projects fail at launch.

You can save money on the headline rate. You will pay it back on rework, missed requirements, time-zone-induced project delays, and the senior engineer you eventually hire to clean up the codebase. Offshore works well for execution against tight specs written by an experienced product lead in your time zone. It works badly for ambiguous greenfield work where the spec evolves during the build. Most failing offshore engagements I have seen had no senior in-time-zone product lead.

Primary stack: Next.js 15 + React 19 + Server Components + Supabase Postgres + Vercel. Astro for content-led sites and marketing surfaces. Tailwind + shadcn/ui for the UI layer. Stripe for billing, Anthropic Claude API for AI features, Resend for transactional email. The stack is opinionated because it ships faster and operates cheaper than the alternatives at this size. For larger teams that want Java or .NET, I will refer you to specialists.

Fixed-price on any scope above 15,000 USD, after a paid 1-2 week discovery that produces a written spec. Hourly only on retainers where the work is genuinely open-ended (3,000-6,000 USD/month minimums). Anyone who fixed-prices a six-figure project from a 30-minute call has padded the number to absorb unknowns. Discovery is the price-discovery mechanism; the fixed price follows.

Pick a no-code tool (Bubble, Webflow, Glide) that fits the use case, accept the lock-in, and ship in 2-4 weeks for 2,000-8,000 USD. The catch: if the product takes off, you will rebuild from scratch in 6-18 months. Worth doing for MVPs you want to validate before investing in custom code. Not worth doing if the spec is already proven or if you need to scale past about 10,000 monthly active users.

Five artifacts. A written technical spec listing every screen, every data model, every integration. A wireframe of every distinct UI surface (8-15 typically). A content-model translation table if there is existing content to migrate. A plugin-equivalency matrix if there is an existing CMS to replace. A fixed-price quote against the spec, valid for 30 days. You own all five artifacts. If you do not proceed with the build, you can hand them to another engineering team.