Stack tier explained: managed, serverless, or DIY
Three honest stacks for shipping your first SaaS — what each one costs, what each one breaks under, and how to know which one you're actually building.
STUB: the framing paragraph. Why the three-tier model exists (most founder advice is implicitly about one tier and silently assumes you're on the same one). The cost cliff between tiers. Why you can move up but rarely down.
Tier 1 — Managed: Vercel + Supabase + Stripe
STUB: Who this is for (most solo founders). Cost shape ($50-300/month all-in for a real-but-small product). What you trade (vendor lock, less control over scaling math) for what you get (zero ops burden, fastest iteration). When this stack breaks (heavy egress, multi-tenant noisy-neighbor patterns).
Tier 2 — Serverless: Cloudflare Workers + D1 + R2
STUB: The egress-free pitch. Why this tier costs less than Managed at scale but more in cognitive overhead. Real ceilings (no Postgres natively, KV instead of full SQL, cold-start considerations for some workloads). Who picks this and why.
Tier 3 — DIY: a VPS, Postgres, and nginx
STUB: The contrarian case. When DIY actually wins on cost (predictable load, beefy single-tenant workloads, anything where egress dominates). The hidden tax (you're now on call). When DIY loses (you have no ops experience and you're going to learn it during your first outage).
How to know which tier you're actually building
STUB: the diagnostic questions. Are you optimizing for iteration speed or unit economics? Do you have ops experience? Do you plan to hire ops or live without it for two years? Answers map to a tier.
The migration story (and why most people don't migrate)
STUB: moving up tiers is rare and expensive. Moving down is almost unheard of. Choosing your tier on day one is a decision that compounds for years. The calculator outputs a tier; this is why.