"Buy or build" is usually presented as a fork in the road. It isn't. Walk into almost any established enterprise and you'll find both: a base of licensed platforms doing the commodity work, and a handful of bespoke systems doing whatever makes the business different. The interesting question was never which side of the fork to take. It's which parts of the estate belong on which side, and most organisations get that allocation wrong for years before anyone stops to check it properly.
The debate usually gets flattened into two caricatures: buying is cheap, fast and safe; building is slow, expensive and a career risk for whoever signs it off. Both caricatures are wrong often enough that neither deserves to be trusted on its own. The honest version of this question has nothing to do with ideology and everything to do with where a piece of software sits relative to what actually makes the business money.
Where off-the-shelf genuinely wins
For a large share of what a business runs on, off-the-shelf isn't a compromise. It's the correct answer. Payroll, expenses, email, calendaring, accounting: capabilities where every organisation needs roughly the same thing, where the software has been refined across thousands of paying customers, and where being different confers no advantage whatsoever. Buying here isn't settling for less. It's recognising that the problem was solved a long time ago, by people with a much bigger budget for solving it.
A vendor with ten thousand customers has already met edge cases an in-house team won't encounter for years, and quietly folded the fixes into a release. Rebuilding that in-house means redoing work better-funded teams have already finished, then maintaining it indefinitely, for a capability that was never going to distinguish the business from a competitor. The test is simple: if a rival could buy the identical system and gain no advantage from doing so, that's the sign to buy it and move on.
The licence trajectory nobody prices in year one
Licence pricing is built around growth, and not the business's growth: the vendor's. Cost is charged per seat, and seats multiply with headcount; cross a usage threshold and the pricing tier jumps with it. A system that looked inexpensive when twenty people used it can become a serious recurring cost once three hundred do, without a single new feature being added.
The moment the business asks for something the standard package doesn't do (a custom field, a workflow tweak, an integration), the "off-the-shelf" system stops being off-the-shelf. It starts costing more through professional services, premium tiers and add-on modules, while remaining, in every sense that matters, something the business doesn't own.
None of this shows up in the year-one business case, because year one is precisely when the vendor is keenest to land the account and negotiating leverage sits with the buyer. The trajectory only becomes visible a few renewal cycles later, once the seat count has grown, the launch discount has quietly expired, and switching costs have become the vendor's best retention strategy. By then, the comparison being tested at renewal isn't the one that was made in year one.
When the process is the product
Some processes aren't commodity. They're the reason customers choose one company over another: a claims-handling method, a way of pricing complex risk, a sequencing logic that shaves a day off delivery. Force a process like that into generic software and it doesn't disappear. It gets bent to fit whichever fields the vendor decided everyone needs, until the work that actually wins business is happening somewhere the software can't see.
A method refined over years, and that genuinely wins work, deserves software shaped around it, not software that reshapes the method to match somebody else's assumptions about how the job gets done. Configuration only stretches so far before the process gets quietly redesigned to fit the tool, and the thing that made the business distinctive gets sanded down to whatever the platform's data model happens to allow.
The test here points the opposite way from the buying test: if a competitor could buy the same system and copy the process simply by configuring it the same way, the process was never proprietary: buy the system. If the process can't be replicated by configuring somebody else's software, no matter how generous the settings menu, that's the signal it's worth building for.
What you own at the end
There's a quieter argument for building, and it has less to do with process than with balance sheets. A licensed system is rented capability: stop paying and access stops, while years of configuration and data sit inside somebody else's product. Bespoke software, built properly, is an asset the business owns outright, and ownership changes what a handful of practical things look like.
It's not free: building means taking on responsibility a vendor would otherwise absorb, and the business becomes its own second line of support. What it buys in return is control over the things a subscription never grants.
- The source code sits in the business's own repository, not a vendor's
- No renewal negotiation stands between the business and its own data
- Changes are commissioned on the business's timetable, not a vendor's roadmap
- The system can be extended, resold or walked away from without starting again
The test that cuts through it
One question does most of the work, even if applying it takes judgement: does this system run a process that makes the business money in a way competitors can't easily copy? If a rival could buy the same package and match the process by ticking the same boxes, buy it, and spend the effort saved on the thing that's actually different. If the process is the differentiator, bespoke is very likely the cheaper option once the licence trajectory and the ownership question are both counted honestly.
Most estates need that question answered system by system, not once at board level. That's a mapping exercise before it's a decision, and it's usually where we start, because the right sequence only becomes obvious once somebody has looked at the whole estate, not just the one system currently up for renewal.
