( Observatory · OBS-01 )

What custom software costs in India in 2026

Real ranges in rupees, where the money goes, and the line item most quotes hide — the years after launch.

· 10 JULY 2026 · 7 MIN

The short answer: in India in 2026, a business website costs roughly₹15,000–80,000 from a capable freelancer and ₹80,000–3,00,000from a studio, with premium custom builds — bespoke design, motion, engineering discipline — running ₹2,50,000 and up. Custom software is its own ladder:₹1,50,000–6,00,000 for a focused tool, ₹6–20 lakh for an SME-grade ERP or platform, and past ₹20 lakh as scale and compliance pile on. Then — the part most quotes leave out — running it well costs two to three times the build price again over the software's life.

Those ranges are wide because "custom software" covers everything from a five-page site with a booking form to a healthcare platform with role-based access and audit trails. This piece breaks down what moves a project up or down the range, where the money actually goes, and how to spend less without buying regret.

The ranges, honestly

Who builds itTypical rangeWhat you're paying for
Freelancer website₹15k – ₹80kOne person's hours. Great for bounded scopes; risky for products that must evolve.
Studio website₹80k – ₹3LStrategy, design, and engineering from people who have shipped before — with continuity after launch. Premium custom builds sit at ₹2.5L and above.
Custom software & ERPs₹1.5L – ₹20LA focused internal tool at the low end; an SME ERP with roles, audit trails, and integrations toward the top.
Agency team₹20L+A staffed team: PM, designers, engineers, QA. You also pay for the overhead that coordinates them.
In-house hire₹15L – ₹40L / yrA senior engineer's fully loaded annual cost — the right call when software is your core business.

Every one of those numbers assumes competent execution. A ₹30,000 build that has to be rebuilt at ₹3,00,000 eighteen months later was never a ₹30,000 build.

Where the money goes

On a well-run project, the budget splits into recognisable pieces:

  • Discovery and strategy (10–15%) — deciding what not to build. The cheapest place to kill a bad idea.
  • Design (15–25%) — flows, interface, and the system behind them, so screen twelve looks like it belongs with screen one.
  • Engineering (40–55%) — the build itself: front end, back end, integrations, and the unglamorous plumbing that decides whether it survives real users.
  • Testing and hardening (10–15%) — edge cases, devices, accessibility, load. Skipping this doesn't remove the cost; it moves it after launch and multiplies it.
  • Launch and handover (5–10%) — deployment, monitoring, documentation, DNS, analytics — the difference between "live" and "operational".

The line item most quotes hide

IEEE lifecycle studies put maintenance at 60–80% of a system's total cost — the build is the smaller share. Across our own maintenance contracts, three years of running a system costs two to three times its original build. Whoever quotes you launch-day pricing without mentioning this is quoting a third of the real number.

Software decays without anyone touching it: dependencies age, browsers change, traffic grows, and the business the software serves keeps moving. Year two has a cost whether or not anyone budgeted for it. The only choice is whether it's planned — a support engagement with a named engineer — or unplanned, at emergency rates, with whoever is available.

What moves a project up the range

  • Integrations — every external system (payments, ERPs, GST and government APIs) adds real engineering and real failure modes.
  • Compliance and audit — healthcare, finance, and education carry data-handling requirements that shape architecture, not just paperwork.
  • Real-time behaviour — live dashboards, chat, tracking. Anything that must update without a refresh costs more than anything that doesn't.
  • Legacy data — migrating years of inconsistent records is often more work than the new system itself.
  • Scale requirements — 100 users and 100,000 users are different systems wearing the same interface.

Building from India for the US and EU

The same market has a second price sheet. Indian engineering bills$15–45 an hour on offshore contracts, and the discipline a US studio sells at $25,000–80,000 ships from India at ₹2.5–6 lakh. The geography changes the invoice; it does not have to change the engineering.

That cuts both ways. If you are an Indian founder selling into Western markets, your product will be judged against software built at Western budgets — so build to that bar, and let the rupee do the arithmetic in your favour. And if a quote from India looks too cheap to be true against these ranges, it usually is: the corners being cut are the ones you find in year two.

How to spend less without buying regret

Cut scope, never quality. A smaller product built properly beats a bigger one built badly, every time it's tried. Launch with the three features your users actually need; let real usage argue for the rest.

Phase deliberately. A ₹6 lakh idea is often a ₹3 lakh phase one and a ₹3 lakh phase two — with the enormous advantage that phase two is designed after real users have voted.

Pay for discovery even if you build elsewhere. A one- or two-week paid discovery produces a scoped, priced plan any competent engineer can execute. It's the cheapest insurance in software.

What we charge, for the record

SkyAstrall websites start at ₹1,00,000, premium builds at ₹2,50,000, and custom software at ₹1,50,000 — floors published on purpose. Every engagement covers strategy, design, engineering, testing, and launch, and it's priced with year two in mind, because we stay on the systems we ship. Our own numbers are public on thetelemetry page; we hold client work to the same standard.

If your budget is under that floor, tell us anyway — a tighter scope is often enough to launch a first version. If it truly isn't, we'll say so on the first call and tell you the honest cheaper path — a freelancer, a template, or a no-code start. An engineer's answer, either way.

Weighing a build like this? Ask the person who writes the code.