CRM & Systems4 min read
Custom software vs off-the-shelf: the arithmetic
Build or buy is the wrong question. Split the system into three layers, model five years rather than one, find the seat count where a build overtakes per-seat licensing, and keep the UAE compliance layer firmly on the buy side.
Danial AhmedFounder & engineering manager
On this page
"Build or buy" is the wrong question, because almost every working system in a UAE mid-sized company is both. The useful question is which layer you build. Buy the ledger, the mailbox and the tax engine; build the slice of workflow that is the reason customers choose you. This is the model we apply on custom software work, and the arithmetic that usually decides it.
Split the system into three layers before you price anything
Most build-or-buy arguments are two people describing different layers of the same system.
- System of record. Ledger, payroll, tax filing, inventory master. Buy. These are commodity, heavily regulated, and expensive to keep compliant. Nobody should be maintaining their own VAT engine.
- System of differentiation. The process that makes you faster or cheaper than the competitor across the street — dispatch logic, pricing rules, underwriting, clinical scheduling. Build, or configure very deliberately.
- System of engagement. Portals, apps, dashboards, the customer-facing surface. Usually build, because this is where off-the-shelf products feel most obviously generic and where the switching cost is lowest.
If the thing being argued about sits in layer one, buy it and stop. If it sits in layer two, the arithmetic below applies.
Model five years, not one
One-year comparisons always favour SaaS, and that is why vendors present them. Model five years, and put the same six lines on both sides: year-one implementation, integration work, licences, hosting and infrastructure, support and administration, and change.
The shape of the answer is always the same. Per-seat licensing is cheap at a small headcount and compounds with every seat added; a build is expensive in year one and then barely moves, because it does not charge per seat. Somewhere the two lines cross — and where they cross for your headcount in three years is the only number in the exercise that decides anything, which is precisely the growth most buyers refuse to model when signing a three-year licence.
The costs each side leaves out
SaaS proposals omit configuration debt (the consultant-days that accumulate every time the process changes), API call limits that turn integration into a subscription tier decision, upgrade windows you do not control, and the cost of getting your data out in a usable shape.
Custom proposals omit on-call, dependency upgrades, the security patching cadence, documentation, and bus factor. A custom system with one person who understands it is a liability with a login page. Budget a retainer from the start or do not build.
The UAE compliance surface belongs on the buy side
There is one category where building is almost always wrong here, and it is the one clients most often propose: statutory compliance.
The Federal Tax Authority's requirements move. Corporate tax at 9% above the AED 375,000 threshold arrived with its own filing mechanics. The Ministry of Finance's Peppol-based e-invoicing programme phases in with formats and accreditation requirements that vendors are contractually obliged to track and you are not. Wage Protection System files have a specification that changes without consulting you.
Buy the compliance layer from a vendor whose business is keeping up with it, and integrate. Build the workflow that sits on top.
Five questions that settle it faster than a spreadsheet
- Would a competitor buying the same product get the same advantage? If yes, buy it — you are not differentiating, you are catching up.
- How many exceptions does your process have? Products handle the happy path cheaply and exceptions expensively. Count them honestly; a long list is a build signal.
- How many systems must it integrate with? Above four, the integration layer becomes the project regardless of which side you pick.
- What is headcount in three years? Per-seat pricing compounds; a build does not.
- What happens if the vendor raises prices sharply or is acquired? If the answer is "we pay", you have priced the licence but not the risk.
The pattern that usually wins
Buy the record layer, build a thin custom workflow over it, and put a real integration layer between them with its own tests and monitoring. It costs more than pure SaaS in year one, far less than a full custom platform ever, and it keeps the compliance burden where it belongs. If the layer in question is customer relationship management specifically, the CRM approach for UAE SMEs walks through the same decision with product-level detail.