Do you actually need custom software?

The three questions, in order

Does a package already do it? If a £30-a-month subscription covers the job, buy the subscription. Someone else pays for its maintenance, security and improvements; you just pay the fee. Custom only wins where the packages genuinely stop.

Is it actually an integration problem? Most "we need something custom built" briefs we see are two systems that don't talk, wearing a trench coat. A small connection or a report that builds itself costs a fraction of a new system and carries none of the ownership burden.

Will the process outlive the build? Custom software fits how you work today, exactly. If the process it serves changes every year, exact fit is a liability; you'll pay for alterations forever. Custom belongs on the processes that are stable and genuinely yours.

Cheaper to build, no cheaper to own

The cost of writing small software has fallen sharply, which is why more firms are asking the question; a tool that would have been quoted at agency prices three years ago can now be built in weeks. The costs that haven't fallen are the owning kind: hosting, maintenance, and someone who understands it after the person who built it has gone. Compare quotes on the owning, not the building. A cheap build with no handover document is the most expensive kind there is.

When custom is the right answer

When the process is the business: the pricing logic that wins you jobs, or the workflow a competitor couldn't copy without hiring your people. A package forces that into someone else's shape. Build it, keep the build small and boring, and buy the rest off the shelf.