n8n vs custom backend: when to pick each

The honest short answer

If your workflow is simple, low-volume, and changes weekly, n8n is great. If it is business-critical, high-volume, or has compliance edges, a custom backend will save you money within months.

We build both. We choose based on the workflow, not on what is trendy.

When n8n wins

• Marketing automations: lead routing, drip campaigns, CRM tagging
• Internal "glue" between SaaS tools where the logic fits in a few nodes
• Quick prototypes — you'll know inside a week if the idea works
• Volume under ~10k executions a month

n8n gets you live in a day. The UI is honest about state and you can hand the workflow to a non-engineer.

When a custom backend wins

• High volume — every execution above the plan tier costs real money
• Complex branching, retries, idempotency, and cross-job state
• Anything touching invoicing, payments, contracts, PHI, or auth
• Multi-tenant logic where each customer needs different rules
• Long-running jobs (>1 minute) or anything queue-driven

A FastAPI service with Celery + Postgres looks heavier on day one. By month three it is cheaper, faster, and the failure modes are observable.

The cost story nobody shows you

A 50k-executions-per-month n8n workflow costs around 50–200 USD on hosted plans, plus your team's time when it breaks at 2am. The same workload on a custom FastAPI + Celery service runs on a 20 USD VPS, retries itself, and pages you with context instead of "execution failed."

The break-even is usually between 15k and 40k executions per month depending on complexity.

How we actually decide for clients

We ask three questions:

1. What is the cost of a silent failure? If the answer is "a customer is charged twice," that workflow does not belong on a no-code tool.
2. Will this workflow change every week, or every quarter? Weekly → n8n. Quarterly → backend.
3. Who owns it in six months? If a non-engineer owns it, lean n8n. If an engineer owns it, lean backend.

Hybrid is fine

Most production stacks use both. n8n handles the marketing layer. The custom backend handles money, identity, and data. They talk through webhooks and a shared Postgres.

That is usually the right answer for an SMB — and it is the one we recommend most often.