Strategy6 May 20267 min read

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

Buying is right more often than developers admit and wrong more often than software vendors admit. Five questions that settle it, plus the honest five-year cost comparison.

A workspace with several systems open side by side

This decision gets made badly in both directions. Companies buy a subscription product for a process that is genuinely how they compete, then spend years working around it. Others commission a bespoke system to do something Xero already does perfectly well.

The framework below takes about twenty minutes and gets it right most of the time.

Question one: is this process how you compete?

The single most useful test. If the process is standard across every business in the country — payroll, accounts, email, calendars — buy it. Somebody has already solved it better than a bespoke build ever will, and they maintain it for a fraction of what it costs you to.

If the process is the thing that makes customers choose you rather than the firm down the road, that is a very different conversation. Forcing your differentiator into a generic product does not just cost efficiency — it slowly makes you the same as everybody else using the same product.

Question two: how many people are bridging gaps by hand?

Count the hours per week spent exporting from one system to import into another, retyping, reconciling two lists that should agree, or maintaining a spreadsheet that exists purely because a product cannot do something.

That number is the real running cost of the current setup, and it is invisible on every invoice. Ten hours a week is roughly £13,000 a year in wages — often more than the entire cost of removing it.

The spreadsheet nobody mentions in the meeting is usually the clearest evidence that a product does not fit.

Question three: what does it cost at three times your size?

Per-seat pricing is designed to be painless at the size you are now. Model it at the size you intend to be. A product at £45 per user per month is £8,100 a year for fifteen people and £32,400 for sixty — before the annual increase you have no power to refuse.

Custom software has effectively no per-seat cost. Its running cost is hosting, which scales with usage rather than headcount and is usually a rounding error by comparison.

Question four: how much of the product do you actually use?

Enterprise products are sold on feature lists. Most companies use a small fraction of them while paying for all of them, and the breadth carries a real cost in complexity, training and the sheer difficulty of finding the four screens your team needs.

If a demo of your bespoke replacement would be four screens and a report, you are paying an enterprise licence fee for four screens and a report.

Question five: how badly does it hurt when the vendor changes something?

Every subscription product will eventually reprice, redesign, deprecate the integration you depend on, or be acquired and quietly wound down. If that would be an inconvenience, buy. If it would stop you trading, that is a dependency worth thinking about more carefully.

The five-year comparison, done honestly

Compare like with like across five years rather than comparing a build fee against a monthly fee.

Off-the-shelf: subscription × users × 60 months, plus implementation and configuration, plus annual increases of five to ten per cent, plus the staff hours spent on workarounds, plus data extraction if you ever leave.
Custom: build cost once, plus hosting at £15 to £150 a month, plus optional support, plus a realistic allowance for changes as the business changes.

Custom usually overtakes somewhere between year two and year four for teams above ten users — and past that point the gap only widens.

The third option most people miss

It is not a binary. The best answer is frequently to keep the off-the-shelf products that genuinely work — accounting, payroll, email — and build only the layer that connects them and handles the part that is specific to you.

That is the cheapest project on our price list and often the one with the largest measurable return, because it removes manual work without asking anybody to abandon a system they already know.

The short version

Buy standard processes. Build the process that makes you competitive.
Count the hours currently spent bridging gaps by hand — that is the real cost of the status quo.
Model per-seat pricing at three times your current headcount before signing anything.
Compare total cost over five years, not build fee against monthly fee.
Integration between the products you already own is often the highest-return project available.

Questions people ask about this

Can I start with off-the-shelf and move to custom later?

Yes, and it is a sensible path. Use a product to prove the process works, then build once the requirements are known and the volume justifies it. Just check the product lets you export your data in a usable form before you commit.

Is custom software harder to maintain?

Only if it is built badly. Mainstream frameworks, conventional patterns and real documentation mean any competent developer can maintain it. The risk is exotic in-house architecture, which is why we deliberately avoid it.