AUTOMATION
n8n vs Zapier vs code — picking the right rung.
Automation tools sit on a ladder of power and cost. The mistake is staying on the wrong rung — either reaching for code too early, or stretching Zapier into territory it was never meant to handle.
Three rungs, in order of complexity: Zapier (and similar — Make, Workato), self-hosted workflow engines like n8n, and bespoke code. Each is the right answer somewhere. The trick is matching the workflow to the rung.
Rung 1: Zapier
Best for: simple linear flows between popular SaaS apps, owned by a non-engineer, where total run volume is modest. New deal in HubSpot → create a Slack channel → post the deal summary. That's a Zap, and you should not write code for it.
Where it breaks: branching logic, anything that needs to loop over arrays of records, anything with serious error handling, and anything that runs thousands of times per day. The pricing curves get punitive past a few thousand tasks per month.
Rung 2: n8n
Best for: workflows that have outgrown Zapier but don't need a full engineering project. Self-hostable (so cost is flat, not per-task), supports real branching and looping, has an HTTP node that lets you call any API, and exposes the underlying data structures to the user.
It's the right rung when the same person who can use Zapier is willing to learn a little more, or when a developer wants a UI for the parts of a workflow that change often. It's also the right rung when run volume makes Zapier's task pricing painful.
Where it breaks: workflows with complex state, transactional guarantees, or behavior that needs unit tests. At that point you're using a workflow engine to dodge writing software, and you'll regret it.
Rung 3: code
Best for: anything customer-facing, anything that touches money, anything with strict reliability requirements, anything that needs to be tested before it ships. Code gets you version control, code review, observability, and the ability to refactor without clicking through a UI.
The cost is real: someone has to write it, someone has to deploy it, someone has to be on call. If the workflow doesn't justify those costs, don't reach for code.
How to pick
Start at the lowest rung that can do the job. Move up only when the workflow has proven itself important enough to deserve the upgrade. The most expensive automation is the one written in code that nobody uses, and the second most expensive is the Zap that breaks silently every time the upstream API changes.