Custom software vs off-the-shelf: how to decide

Buy off-the-shelf software when a product already fits at least 80 percent of how you work and the remaining 20 percent can be absorbed by changing your process. Build custom software when the process is what sets your business apart, when you are paying for three or four tools that do not talk to each other and staff copy data between them, or when no product fits because your work is unusual. Most businesses should buy most of their software and build one or two things. The mistake is building what you should buy, or, more often, buying five things and running the real process in spreadsheets around them.

This post gives a method for the decision, with costs as of October 2026, and the questions that catch the expensive errors in both directions.

What do the three options actually look like?

The first column is where accounting, payroll, email, CRM for a standard sales team, and e-commerce belong. Building any of these is a mistake unless you are in that business. The third column is where the work specific to your company belongs: the way you quote, schedule, track jobs, serve clients or price. Our client portals and internal tools service is built for the third column and, honestly, a good part of our first conversation with a client is working out whether their need is in the first.

When should you buy?

The function is standard. Invoicing is invoicing. Payroll has rules set by the government, not by you. A CRM for a sales team that calls leads and tracks deals is a solved problem; HubSpot vs Zoho vs Salesforce compares the usual choices for small teams.

A product fits most of the process. Trial it with real data for two weeks. If the team can run the actual workflow in it, with no more than a few workarounds, buy it and change the workarounds into process.

The team is small. Per-seat pricing is cheap at five users and expensive at fifty. Under ten people, almost every subscription beats a build.

You need it next week. Custom takes months. If the pain is acute, buy something now and reconsider in a year when you understand the process better.

You cannot maintain software. Custom software needs someone to keep it running, apply updates and fix things. If there is nobody and no budget for a care arrangement, buying is safer.

When should you build?

The process is your advantage. If customers choose you because of how you handle their orders, projects or cases, software that forces you to work like everyone else erodes the reason they chose you. This is the strongest argument for building, and the one we see most in logistics, manufacturing, professional services and specialist trades.

You pay for tools that do not talk. Three subscriptions, a spreadsheet to join them and a person who spends Fridays reconciling. Sometimes the fix is an integration between the existing tools, which is cheaper than a build; API integration explained and Zapier vs Make vs n8n vs custom code cover that route. When the spreadsheet in the middle is where the real work happens, the spreadsheet is the system, and when to replace spreadsheets with a custom tool describes the point at which it should become one.

Clients or partners need access. Giving customers a login to see job status, documents and invoices is rarely done well by generic tools, and client-facing software reflects on your brand. Client portal: what it is and what to include covers the usual scope.

The seat count makes subscriptions expensive. Fifty staff on a tool at ₹2,500 or $30 a seat a month is ₹15,00,000 or $18,000 a year, every year. A custom tool at ₹12,00,000 or $30,000 to build and ₹2,00,000 or $5,000 a year to run pays back in under three years and then keeps paying.

Nothing fits. You have trialled four products and each one fails on the same requirement. That requirement is probably the thing that makes your business different, and it will not get solved by the fifth product.

How do you put numbers on it?

Four figures, for five years.

1. Subscription cost: seats times price times sixty months, with a 10 percent annual rise, because that is what happens. Two or three tools if that is what it takes.
2. Workaround cost: hours a week spent copying, reconciling and chasing, times an hourly cost, times 250 working weeks. Our automation hours-saved calculator does this arithmetic for a task, its frequency and the people involved.
3. Build cost: a fixed-scope quote for a first version, plus 15 to 20 percent a year for hosting and maintenance.
4. Risk cost: harder to quantify, so list it. For buying: price rises, a feature removed, the vendor acquired. For building: the developer disappearing, scope growing, nobody maintaining it.

When the subscription plus workaround total over five years is well above the build plus maintenance total, build. When they are close, buy, because buying is faster and reversible. When buying wins clearly, buy and stop thinking about it.

What makes custom software go wrong?

Three things, all avoidable.

Scope. A tool meant to replace one spreadsheet grows to replace the ERP. Keep the first version to one process, with the roles and reports that process needs. How to build an internal dashboard for operations shows what a well-scoped first version looks like, and roles and permissions covers the access design that is cheap to get right at the start and painful to retrofit.

Data. Old data in old formats has to come across, cleaned. It is always more work than estimated; data migration from old systems is the checklist. Decide what not to migrate.

Ownership. The code, the hosting account, the database and the documentation must be yours, in your accounts, with your agency as a collaborator. If a proposal does not say that plainly, ask.

A fourth, for integrations specifically: anything that connects two systems will break when one of them changes. Build with logging, retries and alerts, or the first you hear of a failed sync is a customer complaint. Why integrations break, and how to build ones that do not is the detail.

A decision in five questions

1. Is this a standard business function with established products? If yes, buy.
2. Did a two-week trial with real data cover at least 80 percent of the process? If yes, buy and adapt the rest.
3. Is the process one that customers choose you for? If yes, lean build.
4. Over five years, is subscriptions plus workarounds clearly more than build plus maintenance? If yes, build.
5. Can you fund and staff the maintenance? If no, buy, or build only with a care arrangement in place.

Vendor pricing pages are the right source for the first number; HubSpot's pricing page, for example, shows how quickly per-seat costs climb across tiers. Put your own seat count in and do the sixty-month multiplication before the sales call, not after.