Custom software is expensive to build and expensive to own. It is also, sometimes, the only sensible option. The decision deserves more rigour than it usually gets.
Buy when
- The process is standard and gives you no competitive advantage.
- A mature product covers eighty percent of your needs without heavy configuration.
- The remaining twenty percent can be handled by changing how you work.
Build when
- The process is the advantage, and forcing it into a product would erode it.
- Licence plus workaround costs already approach the cost of building.
- Integration requirements exceed what any product exposes.
- You need to own the data model because it changes frequently.
Change the process when
This option is skipped far too often. If the requirement exists only because of an accident of history, no software will fix it — you will simply encode the accident and pay to maintain it. Ask why the process is shaped this way before deciding what to build.
A useful test
Write down the three things the system must do that no product on the market does. If you cannot name three, buy. If you can name three and they are genuinely tied to how you compete, building is likely justified — and worth doing properly.