n8n, Zapier or custom code: when an iPaaS is enough, and when it is not

Tools like n8n, Zapier or Make have made it possible for anyone to automate simple connections between apps without writing code. They are an excellent tool — until they are not, and that turning point is the one almost nobody sees coming in time.

Where an iPaaS is the right call

For flows with simple logic (“when a form comes in, create a row in the spreadsheet and send an email”), an iPaaS is unbeatable on speed: it is set up in hours, requires no infrastructure to maintain, and anyone on the team with a bit of technical curiosity can learn to adjust it.

For moderate volumes and one or two specific automations, the tool’s monthly cost is far lower than a custom build.

Where it starts to fall short

Complex conditional logic. Once the flow starts having many branches (“if the customer is type A and the amount exceeds X and it is not the weekend, then…”), the visual editor becomes hard to read and maintain, even for the person who built it.

High volume. Most of these platforms charge per execution or per task. A flow that works fine at 500 runs a month can get expensive at 50,000, and that cost growth is rarely in anyone’s initial spreadsheet.

The need for full control over errors. When a failure in the automation has real cost (a lost order, a badly generated invoice), you need retry logic, alerts and logging finer than what the standard visual editor offers.

Watch out: the moment to migrate is not when the flow fails once. It is when every small change requires rebuilding half the automation.

The clearest sign you have already hit the ceiling

If every minor adjustment requires reorganizing a large part of the visual flow, or if nobody on the team fully understands anymore why the flow does what it does, it is time to consider whether that process deserves custom code: more control, more performance, without the growing per-execution cost.

It is not an all-or-nothing decision

Many companies combine both: an iPaaS for simple, low-volume connections, and custom development for the two or three critical processes where volume or complexity already justify it. You do not need to pick a single approach for the whole company.

How we do it at Bitora

We start by evaluating whether the process fits a tool like n8n — we do not sell custom development where a tool already solves the problem. When volume or complexity justify it, we migrate that specific process to custom code, without touching the rest of the automations that already work well.

If you have flows that already cost more to maintain than they save, see the full approach on AI agents and automation.