The build vs buy decision shapes budgets for years, yet most teams compare visible costs alone: license fee versus salary. The full math looks different. Building with your own team costs about 4x the license price once permanent maintenance and lost opportunity enter the ledger. Here is the decision tree, with numbers on both sides.
The true cost of building
An authentication system consumes 4 to 6 engineer-months before launch. Add the permanent tax: 15% to 20% of original effort per year on upkeep, fixes, and new requirements. Then add the biggest item: every month spent on auth is a month away from the feature that would differentiate your product.
The true cost of buying
Licenses scale with seats and usage, integration eats weeks, and lock-in erodes bargaining power. The surprise item: SaaS renewal prices climb 20% to 40% once your operation depends on the vendor. Dependency created, the hike arrives on schedule.
Question 1: does this differentiate your product?
Start the tree here. A layer customers pay for and competitors need months to copy deserves internal build. Authentication, email, and payments fail that test for 99% of companies; they are commodities served by excellent vendors.
Question 2: does a mature vendor exist?
A mature vendor has customers your size, references in your industry, and a public roadmap. Non-core problem plus proven vendor means buy, every time. The gray zone opens when the problem matters and the market stays young; there, a paid 90-day pilot beats a leap of faith.
The numbers on one example: authentication
In-house auth costs $180k in year one across loaded salaries, plus $30k per year in maintenance. A managed provider charges about $35k per year at the same scale. Auth will never be a competitive differentiator, so the $180k never pays itself back. Buy it and point the team at the layer that produces margin.
Gray zone rules
Buy the boring layers: identity, payments, email, observability. Build the differentiating layer. Connect the two with contracts that allow exit: guaranteed data portability, short notice periods, proportional termination fees. Clean boundaries between layers matter more than protective clauses.
Plan the exit before signing
Every bought component needs a documented replacement path: which tool takes over, how long migration takes, who owns it. Negotiate data export terms at signature time; after dependency sets in, you take whatever deal remains. Vendors know how to read a contract with a plan B attached.
