Custom software vs off-the-shelf: what five years actually costs

Most custom-versus-package arguments are won by whoever picks the assumptions. This page picks them out loud. Every number below is an input you can replace with your own, and the sections after the model show which inputs actually change the answer.

Two mistakes that make the comparison meaningless

Comparing year one. A subscription spreads its cost over months; a build is paid up front against milestones, so over twelve months the subscription always looks better. Operational software stays in place for five to seven years, and that is the horizon the decision has to be made on.

Comparing unequal scope. A packaged product may cover invoicing, stock, CRM, reporting and a mobile app, built over many years. Rebuilding all of it is not a project. The only fair comparison is between the package and the part of your work you would actually build — and if you use most of what the package does, stop reading and keep it.

The assumptions behind the model

Worked five-year model at 25 seats

Under these inputs the difference is about $176,000 — a number only as good as the inputs, which is the point of the next section.

Change one assumption and watch the answer move

Each row below changes a single input and leaves the rest of the model alone.

Two inputs dominate everything else: how many seats you pay for and how much manual work the package leaves behind. The headline licence price barely moves the outcome by comparison. If you only measure one thing before deciding, measure the spreadsheet hours — the method is set out in the true cost of manual work.

A case where the package wins

A six-person consultancy runs client records, invoicing and time tracking on a $60-per-seat product and uses nearly all of it. Five years of subscriptions come to roughly $25,300, plus $2,000 setup and $3,000 of occasional configuration: about $30,300 in total. Building an equivalent system honestly costs $20,000 and produces about $42,500 over five years once maintenance and residual manual work are counted.

The package wins, and it is not close. Small team, standard process, high utilisation of the product — that combination beats a custom build every time, and we will tell you so during discovery rather than after the invoice.

Where off-the-shelf is the correct choice

• Statutory work. General ledger, payroll, tax filing and e-invoicing are legally defined and change with legislation. Packaged products track those changes for you. Do not rebuild them.
• Small teams with an evolving process. Below roughly ten people, and while the process is still changing every month, the flexibility of a subscription is worth its price.
• High utilisation. If you use most of the modules you pay for, the per-seat price is buying real work.
• Short-lived workflows. A process that will exist for six months does not deserve a build.

The hybrid pattern most companies end up with

The decision is rarely all or nothing. In practice, statutory accounting stays in the packaged product it belongs in, and the workflows that are specific to your business — production tracking, dealer ordering, field service, approvals — move onto custom modules that talk to the package through an integration. You get compliance from the vendor and fit from the build. How those connections are made in practice is covered on our integrations page.

Run the model on your own numbers

1. Pull the real invoice, not the list price

   Take the last twelve months of subscription invoices, divide by seats and by twelve. Include seats nobody uses.

2. Log one week of spreadsheet hours

   Ask every team to record hours spent working around the system. Multiply by the loaded wage, then by 52.

3. Count the modules you actually open

   List which parts of the package are used weekly. That list, not the whole product, is the scope of any custom comparison.

4. Get a quote for that list only, then add the custom-side costs

   Price the subset against the published bands, then add maintenance and residual manual work. A comparison that puts zero in those two rows is a sales sheet, not a model.

If you want the arithmetic done against your own invoices, send the numbers through the quote form, or read the decision framing on package versus custom first.

Frequently asked questions

Why compare over five years instead of the first year?

Because operational software stays in place for five to seven years. In year one a subscription looks cheaper, since the build is paid up front and the subscription monthly. From year two the subscription keeps compounding with headcount and annual uplifts while the build is already behind you.

What assumptions decide whether custom or off-the-shelf wins?

Seven: seats paid for, price per seat, annual uplift, one-off configuration, the yearly cost of change requests, the hours a month your team spends bridging gaps the package leaves, and your maintenance rate. Workaround labour moves the result more than the licence fee does.

What is workaround labour and how do you measure it?

It is the paid time staff spend exporting, cleaning and re-entering data because the package cannot do something your business needs. Ask each team to log the hours they spent in spreadsheets outside the system for one week, then multiply by the loaded wage and by 52.

When is off-the-shelf clearly the better buy?

When you use most of what the package does, when the process is standard and legally defined such as accounting, payroll and tax filing, when the team is small, or when the workflow will exist for only a few months. A package that genuinely fits is hard to beat.

Do I have to replace my accounting package to build custom software?

No. The common pattern is hybrid: statutory accounting stays in the packaged product and the workflows specific to your business run on custom modules connected to it. You keep compliance where it belongs and build only what the package cannot cover.

What does the custom column include over five years?

The one-off build, annual maintenance at 12%, 18% or 25% of build cost, hosting after the first year, and the residual manual work no system removes entirely. There are no per-user fees, and at the end of five years you own the code.