Six years of building automation for SMEs has taught us where n8n earns its keep and where it quietly turns into a fragile grey box nobody wants to open. This is the honest guide.
| Zapier | Make | n8n (self-hosted) | Code | |
|---|---|---|---|---|
| Best for | Marketing-owned | Visual branching | Volume + control | Custom logic, performance |
| Owner | Anyone | Ops person | Ops + engineer | Engineer |
| Cost @ 100k ops/mo | ~$700 | ~$300 | ~$10 hosting | Hosting only |
| Version control | ❌ | ❌ | ✅ (JSON in git) | ✅ |
| Data residency | US-only mostly | EU option | Wherever you host | Wherever you host |
| Failure visibility | Email digest | Email + retry UI | Custom (Slack, PagerDuty) | Whatever you build |
The visual layout stops helping and starts hiding logic.
You're coding in a text box with no tests — port it out.
Reuse deserves a real function, not copy-paste JSON.
n8n's per-node overhead adds up. Native code wins.
One customer's failed workflow shouldn't slow another's.
Auditors ask 'where's the test suite for this?' — n8n has none.
Zapier for the person in marketing who owns the workflow (best UX, most expensive at volume). Make for complex branching visually. n8n when volume is high enough that per-op billing hurts, or when self-hosting for data residency matters. All three are fine — the wrong choice is 'whatever the last consultant used'.
Tell us your goals — we'll recommend the right technology honestly, even if it isn't this one.