Know which software bills are worth replacing.
Practical verdicts for common SaaS infrastructure: what it costs, what can replace it, and how much operational work the swap actually creates.
The default host for Next.js. Trivial to leave if your app is a standard Node server — Vercel's value is DX, not anything proprietary.
Postgres plus batteries. The Postgres part is easy to self-host; the Auth/Storage/Realtime surface is what you'd be rebuilding.
Auth0 charges per user. For most SaaS, a session-cookie library plus an email/password (or OAuth) flow is a weekend of work and $0/mo.
Drop-in auth UI that bills per MAU. The convenience is real; so is the bill once you cross 10k users.
Cheap at the unit price, expensive in operational tax. Replacing it means owning deliverability (IPs, DMARC, bounces).
Don't. Card processing, PCI scope, dunning, tax, and 3DS are why every sane founder pays Stripe's ~3%. There is no savings here worth the risk.
Self-hostable (it's open source), but the self-hosted Sentry docker stack is heavy. GlitchTip is the lean alternative.
Excellent relevance, expensive at scale. Postgres full-text search or a self-hosted Meilisearch/Typesense covers 90% of use cases.
You're paying for branching and a managed proxy over MySQL. Plain MySQL or Postgres on a VPS is a direct swap if you don't need branching.
The all-in-one observability tax. Powerful, but priced for enterprises. Piecing together the open-source equivalent is real work.
Convenient per-request pricing. A self-hosted Redis or a KeyDB instance on your existing box is free and plenty fast.
S3 storage is cheap; S3 egress is not. S3-compatible alternatives with free egress (paired with Cloudflare) cut the bill dramatically for read-heavy workloads.