How Much Does Custom Software Cost? A Transparent Budgeting Guide

What actually drives the cost of custom software

Almost every software budget reduces to one formula: the number of people, multiplied by their rate, multiplied by how long they are needed. Everything else, from features to compliance, matters because it changes one of those three terms. A login screen is cheap; a login that supports single sign-on, audit trails and three user roles adds days to the timeline. Understanding which decisions move which term is the fastest way to control a budget.

Scope is the largest driver, but it is rarely the number of screens. It is the number of distinct workflows, the edge cases inside each one, and how much data must move between systems. Team shape matters next: a lean senior team often costs less in total than a larger junior one, because it makes fewer expensive mistakes. The delivery model, offshore, hybrid or onshore, changes the rate itself, often by a factor of two or three.

For AI products there is a newer driver: model strategy. Calling a hosted frontier model is the fastest way to start. Running private, self-hosted models adds infrastructure, evaluation and operations work before the first feature ships. Integrations with legacy or poorly documented systems and regulatory requirements such as HIPAA, PCI DSS or India's DPDP Act add effort that is easy to underestimate, because much of it is testing, documentation and review rather than visible features.

Fixed price, time and materials, or a dedicated pod

A fixed-price project agrees scope, timeline and price up front. It suits work that can be specified clearly, such as a defined MVP or a migration with known boundaries, and it moves delivery risk to the supplier. The trade-off is rigidity: every change goes through change control, and an honest supplier will include a contingency for the unknowns. Fixed price only works when the scope document is detailed enough that both sides would describe the finished product the same way.

Time and materials bills for the hours actually worked. It is flexible and transparent, and suits exploratory work where the problem is still being understood, but it moves cost risk to the buyer and needs active product ownership to stop the budget drifting. A dedicated pod sits in between: a stable team billed per month that works through a prioritised backlog. It suits products with an evolving roadmap, where the real question is how much capacity you want each month rather than what one release costs.

Our Budget Calculator offers the two models most of our clients choose between: a fixed project and a monthly pod. Pods have a minimum term of two months at the time of writing, because a team needs a few weeks to become productive in a new codebase and shorter engagements rarely deliver value. Whichever model you choose, insist on the same artefacts: a prioritised backlog, a demo at the end of every iteration, and a burn-down that shows spend against delivered scope.

How our Budget Calculator turns choices into a number

We built the calculator to show its working rather than hide it. You choose what you are building (an AI agent swarm, an enterprise RAG system, a SaaS platform or a cloud and DevOps engagement), a model strategy, a scale and a delivery model. Each project type has a base team size and a number of weeks to production for each scale. The MVP tier removes one engineer from the base team and the Enterprise tier adds one. Choosing frontier APIs takes a week off the timeline, while private models add two weeks and one engineer.

The rates are monthly, per engineer, and at the time of writing they are 2,000 US dollars for offshore delivery from our Bengaluru delivery centre, 3,200 dollars for hybrid delivery with an onshore lead and an offshore team, and 6,000 dollars for an onshore US or UK based team. The model strategy then applies a multiplier: 1.0 for frontier APIs, 1.05 for a hybrid gateway and 1.15 for private models, reflecting the extra infrastructure and evaluation work.

The monthly team cost is engineers multiplied by rate and uplift. For a fixed project it is multiplied by duration, converting weeks to months at 4.345 weeks per month. Early clients currently receive a 15 percent launch discount, and the calculator shows the regular figure struck through beside the offer price. Finally, it presents a range of 12 percent either side of the point estimate, because an honest early estimate is a range, not a number. Rupee figures use an indicative rate of 84 rupees to the dollar; we invoice clients in India in INR and international clients in USD.

• The timeline starts from base weeks for the project type and scale, is adjusted by model strategy, and never drops below three weeks.
• The team starts from a base headcount, is adjusted by scale and model strategy, and never has fewer than two engineers.
• The monthly cost is the number of engineers multiplied by the delivery-model rate and the model-strategy uplift.
• The project budget is the monthly cost multiplied by weeks divided by 4.345, less the launch discount, shown as a range of plus or minus 12 percent.

A worked example, step by step

Take a fixed project for a SaaS platform at Growth scale, which the calculator describes as up to 50,000 users with fast iteration, using a hybrid model gateway and offshore delivery. The base timeline for a growth-stage SaaS platform is eight weeks with three engineers, and a hybrid gateway changes neither. The regular monthly team cost is 3 engineers multiplied by 2,000 dollars multiplied by 1.05, which is 6,300 dollars. Eight weeks is 1.84 months, so the regular project cost is about 11,600 dollars.

Applying the 15 percent launch discount gives a point estimate of about 9,860 dollars, or 5,355 dollars a month for the team. Twelve percent either side produces a range of roughly 8,680 to 11,040 dollars, which the calculator displays as 8.7K to 11K dollars, or about 7.3 to 9.3 lakh rupees. The phase split shows where the effort goes: 8 percent discovery, 12 percent design, 50 percent build, 15 percent testing and 15 percent deployment, so roughly 4,930 dollars of the point estimate is the build itself.

Change one input at a time and the drivers become obvious. Switching to hybrid delivery raises the range to about 13.9K to 17.7K dollars; onshore delivery takes it to about 26K to 33.1K dollars. Keeping offshore delivery but moving to private models adds two weeks and an engineer, giving about 15.8K to 20.2K dollars. The same configuration as a monthly pod comes out at about 4.7K to 6K dollars a month, with the two-month minimum. All of these figures are indicative and correct at the time of writing; the calculator itself always shows the current rates.

The costs that sit outside any build quote

A build estimate, including ours, covers the team that designs, builds, tests and deploys the software. It does not cover the costs of running it, and those can match the build cost within a couple of years. Our calculator states this plainly: hosting and third-party licences are quoted separately. Ask for a run-cost estimate alongside every build estimate, broken down by month at your expected usage.

For AI products, model usage is the cost most often missed. Token spend scales with users, with the length of prompts and retrieved context, and with how many model calls each task makes. Estimate it bottom-up: calls per task, tokens per call, tasks per month and the provider's current price, then design it down with caching, smaller models for simple steps and tighter retrieval. Maintenance is the other large line. A common planning rule of thumb is 15 to 20 percent of the original build cost per year for updates, security patches, dependency upgrades and small improvements.

• Cloud hosting, managed databases, storage, backups, monitoring and log retention are billed every month the product runs.
• LLM tokens, embeddings and vector database usage grow with every active user and every extra model call.
• Annual maintenance is typically planned at 15 to 20 percent of the original build cost.
• App store developer accounts, store commissions on digital purchases and third-party services such as email, SMS, maps and payments all carry their own fees.
• Security reviews, penetration tests and compliance audits recur on a schedule rather than happening once.

How to cut scope without cutting value

The cheapest feature is the one you do not build yet. A real MVP is not a smaller version of the full product; it is the thinnest slice that lets real users complete the one job that proves the business case. Everything else, from admin dashboards to secondary user roles, can often be handled manually or with an off-the-shelf tool for the first few months, and built properly once usage shows which parts matter.

Phased delivery turns that principle into a plan. Split the roadmap into releases that each deliver usable value on their own, estimate the first in detail and the later ones roughly, and re-estimate each phase with what the previous one taught you. This is the reason our six-step process starts with a week of discovery and architecture and two weeks of prototyping before core engineering: decisions made on a clickable prototype are far cheaper than decisions made in production code.

• Write down the single job the first release must let a user complete, and cut anything that does not serve it.
• Replace low-volume back-office features with manual processes or existing tools until volume justifies building them.
• Use managed services for authentication, payments and notifications instead of building them.
• Start AI features on hosted models and move to private models only when cost, latency or data residency demands it.

How to read an estimate, and red flags in a quote

A useful estimate tells you who will do the work, for how long, at what rate, under which assumptions and with which exclusions. It breaks the work into phases or milestones you can track, and it names the risks it has priced in. When comparing quotes, line them up on those terms rather than the headline figure. A quote that is half the price of the others has usually assumed a smaller scope, a junior team or no testing, and you will pay for the difference later.

Be wary of precision that the information does not support. An exact figure for a project described in a single page is a guess presented as a commitment. Equally, a quote with no discovery phase is asking you to trust that the supplier understood your business from one call. The best suppliers will tell you what they do not yet know and how they plan to find out.

• The quote lists no assumptions or exclusions, so any surprise becomes a paid change request.
• There is no line for testing, QA or deployment, which usually means they will be squeezed at the end.
• The supplier keeps the intellectual property or source code instead of transferring it to you.
• Large up-front payments are not tied to delivered, reviewable milestones.
• Roles and seniority are not named, so the team you meet is not necessarily the team you get.
• Running costs, licences and post-launch support are not mentioned at all.

How we would approach your budget

Start with the Budget Calculator to get an indicative range in dollars or rupees in a couple of minutes, and to see how delivery model, scale and model strategy move the number. Then book a free discovery call: we turn the range into a fixed quote with a detailed scope, written assumptions and a separate estimate of running costs, and you own all of the code and IP from day one.

The calculator covers AI agent, RAG, SaaS and cloud work. For mobile apps, Chrome extensions, enterprise integrations or anything else in our services, including Custom Software Engineering and SaaS Product Development, a short scoping call gets you to the same kind of transparent estimate, built on the same principle of showing the team, the rate, the duration and every assumption behind the number.