A software company telling you that you need custom software is like a barber telling you that you need a haircut. So here is the framework we actually use internally — including the cases where we tell clients not to hire us.
Buy when the process is not your advantage
Payroll, accounting, email, helpdesk, CRM basics — thousands of companies have solved these and your version will not be better. Buying means someone else pays for maintenance, security patching and compliance updates forever. The correct response to “our payroll is unusual” is almost always to make your payroll less unusual.
Build when the process is your advantage
Build when the workflow is how you actually win — your particular way of routing deliveries, pricing risk, or matching supply to demand. Build when no product supports an integration you depend on. Build when per-seat licensing becomes punitive at your scale. And build when the data is genuinely too sensitive to sit in someone else’s multi-tenant database.
The three-year total cost comparison
One-year comparisons always favour buying; three-year comparisons often do not. Model both properly.
- Off-the-shelf, 3 years — licences × users × 36 months, plus implementation, plus per-integration fees, plus annual increases
- Custom, 3 years — build cost, plus roughly 15–20% of build cost per year in maintenance, plus hosting
- The crossover usually lands between 25 and 60 users, sooner if you need several paid integrations
The hybrid answer almost everyone should choose
Buy commodity systems, build the one thing that differentiates you, integrate them. A logistics company should buy accounting and build routing. A clinic should buy payroll and build the patient journey. Trying to build everything is how a five-person company ends up maintaining eleven applications.
Three questions that settle it quickly
Would a competitor beat you if they had this exact same system? If no, buy it. Will this process still work the same way in three years? If yes, buy it. Does any existing product handle 80% of it acceptably? If yes, buy it and adapt. Three noes means build.
The short version
Build the thing that makes you money. Buy everything else. Most bad software decisions come from getting that sentence backwards.
Frequently asked questions
Upfront, yes. Over three to five years, often not — especially past 25–60 users, where per-seat licensing and paid integrations compound. Model total cost of ownership, not purchase price.
Five to seven weeks for a focused single-workflow tool, ten to fourteen weeks for a full platform, sixteen weeks or more for enterprise systems with compliance requirements.
Nothing catastrophic, if you set it up correctly: code in your repositories, infrastructure in your cloud accounts, documented architecture and no proprietary dependencies. Insist on all four from day one.
Yes, and it is often the smartest sequence. Use a product to learn what you actually need, then build the parts that matter once requirements are proven rather than guessed.
Not sure which side you're on?
Book a free 30-minute call. We will tell you plainly if buying is the better answer — that conversation costs you nothing either way.
Filed under
