Business Systems
How Much Does Custom Software Development Cost?
Planning ranges and cost drivers for custom software — discovery, integrations, quality, and maintenance — without a fake price menu. Use estimators to bound a conversation, then scope the work.
Palmate Solutions Editorial · Published 18 March 2026 · Updated 20 July 2026 · 5 min read
Nobody can honestly print a single number for “custom software.” The phrase covers a three-screen internal form, a marketplace sync with reservations, and a product with mobile clients. Treating any agency’s blog as a price list is how you get a quote that matches a fantasy scope.
This article is a planning guide: what drives cost, which ranges are useful for budgeting conversations, and what Palmate will not pretend. Figures below are order-of-magnitude bands in INR for Indian SME work as of 2026 planning — not Palmate Solutions’ official menu, not a promise, and not a substitute for a written proposal after discovery.
If the work is still website-shaped, start with the website cost estimator. If the work is pipes between systems, use the API project estimator. Both exist to structure questions, not to replace a statement of work.
What you are paying for
Custom software development is not “writing screens.” A maintainable system includes:
- Domain modelling — entities, states, who may change them.
- Application engineering — web UI, APIs, background jobs.
- Integrations — payments, GST-relevant exports, warehouses, CRMs.
- Environments — development, staging, production, secrets, backups.
- Quality — tests where they pay off, observability, access control.
- Handover — a repo someone else can extend.
A cheap build that omits the last four items is not cheap. It is a deferred invoice plus operational risk. Compare that path to staying on SaaS in custom software vs SaaS before you assume a build is the bargain.
Cost drivers that actually move the number
Ambiguity. If the approval path cannot be written in a page, engineering will discover it in production. Discovery and IT consulting are cheaper than rebuilding screens twice.
Integrations. Each third-party API adds auth, pagination, rate limits, sandbox cutover, and failure behaviour. Two systems in one direction is a different project from five systems with bidirectional stock. Estimators help you count flows before anyone guesses weeks.
Roles and audit. “Admin sees everything” is cheap. Maker-checker, location-based access, and export trails are not. If money or stock moves, budget for it.
Offline, mobile, and poor networks. A desktop CRUD app is not a field app. Mobile clients expose pagination, token refresh, and store release work. Do not hide that inside a “website” estimate.
Data migration. Spreadsheets with three years of informal SKUs are a project. Cleaning them is often larger than the first UI.
Non-functional requirements. Uptime targets, restore time, and “must work during a sale” change hosting and engineering. A business-hours internal tool should not be priced like a public checkout.
Who operates it afterwards. Include a change budget. Software that cannot be changed will be replaced, which is the most expensive maintenance.
Planning bands (not quotes)
Use bands to test whether the problem and the budget are in the same universe. Widen or narrow after a workshop.
| Shape of work | Typical planning band (INR, 2026, SME) | What this usually includes |
|---|---|---|
| Narrow internal tool, one role, few integrations | A few lakh | Discovery, one web app, basic auth, staging, handover |
| Operational system with roles, exports, one or two APIs | High single-digit to low tens of lakh | Domain model, jobs, observability, real UAT |
| Multi-channel sync, reservations, or several vendors | Tens of lakh and up | Integration service, idempotency, lag monitoring |
| Product with public users, mobile, and ongoing roadmap | Programme, not a one-off project | Team, releases, support, store operations |
These bands assume a competent small team, not an enterprise programme with six vendors. They assume you can provide a product owner. They do not include years of feature work after launch.
Hourly or monthly team rates in India vary widely by seniority and whether you hire staff, a studio, or a freelancer. A lower rate with no staging environment is not a saving. How to choose a software development company is about that operational test.
International rate cards (USD/EUR) are a different market. If you are comparing an overseas bid to a Bengaluru studio, compare scope and operating model, not a single hourly number.
Why “fixed price for everything” goes wrong
Fixed price is reasonable when scope is fixed: a known object model, known vendors, known environments. It becomes theatre when the brief is “like that app, but for us.” Someone will absorb the uncertainty — usually quality, timeline, or change requests billed later.
A healthier commercial shape:
- Paid discovery — process, data, integrations, success metrics.
- A build proposal with explicit in/out, environments, and acceptance tests.
- A maintenance retainer or change budget so the system does not freeze on launch day.
If a vendor quotes a large fixed number from a one-page email, they are guessing, or they have a template they will squeeze you into.
Hidden costs buyers forget
- Your people’s time. UAT, data cleanup, and policy decisions. If managers cannot attend workshops, the project will stall regardless of engineering speed.
- Licences and infrastructure. Hosting, SMS, maps, transactional email, app store accounts. Small monthly lines that were never in the software quote.
- Compliance and legal. Privacy notices, DPDP-related process, payment PCI boundaries. Engineering cannot invent your policy.
- The opportunity cost of delay. Sometimes SaaS plus glue ships a year earlier. That year has a value. Cost the delay, not only the invoice.
For website and brochure-plus-CMS work, the website cost estimator makes those buckets visible. For integration-heavy work, the API project estimator does the same for systems and directions of data.
How Palmate talks about money
We do not publish a fake menu of “ERP: ₹X, dashboard: ₹Y.” Those numbers would be wrong for your warehouse packs, your marketplace, and your audit needs. We will:
- Push back if SaaS already fits.
- Separate discovery from build.
- Name integrations as first-class work.
- Include environments and handover in the build, not as goodwill.
If you need a decision before a build — build versus buy, sequence, current-state review — that is technology consulting, which is cheaper than the wrong custom project.
A budget conversation you can have this week
Write down: users and jobs, systems that must connect, what “done” means for v1, who will own the backlog, and what you can spend this financial year including hosting. Run the relevant estimator. Then ask for a discovery proposal, not a guess.
Custom software costs what the process, the integrations, and the operating standard cost. Anything presented as a universal price is marketing. Treat ranges as a filter. Treat a scoped proposal as the number that matters.