For most small businesses the right place to start is low-code: tools like n8n, Zapier and Make that connect the software you already use. They're quick, cheap and easy to change. They stop being the right answer in some specific situations. This page shows you how to tell, and what to measure so you aren't guessing. There's no magic number, and nobody who tells you otherwise has one.
What's inside
- A side-by-side of what each approach is good at
- Seven signs you've outgrown low-code
- Two mistakes people make, in opposite directions
- Three things to measure, and the questions to ask
Side by side
| Question | Low-code fits when… | Custom software fits when… |
|---|---|---|
| How many moving parts? | A handful of steps across one to three apps | Dozens of steps, many branches, many systems |
| What if it fails? | A missed reminder is an annoyance | A failure costs real money or breaks a promise to a customer |
| Do the apps connect? | Yes, or a simple web request works | An old, closed or unusual system with no usable way in |
| Who looks after it? | You or a helper can open the editor and fix it | A developer owns it, with tests, version control and monitoring |
| What kind of data? | Ordinary business data | Regulated or sensitive data, or an audit trail you must keep |
| How much volume? | Occasional to moderate | High volume, where speed or cost per run matters |
| How fast to start? | Days | Weeks to months |
| What does it cost? | Cheap to start. Cost grows with use | More up front. Can be cheaper at high volume |
Seven signs you've outgrown low-code
One of these is a reason to look. Three or four is a reason to act.
- It breaks more and more often, and fixing it has become routine.
- The logic has grown tangled. Nobody can explain it without opening every step.
- You've hit a hard limit: a step that takes too long, uses too much memory, or can't reach what it needs.
- A system you depend on has no connector, or has an unusual login or an undocumented interface.
- A mistake would be expensive, and you need checks and a record that a visual builder can't enforce.
- The data is sensitive, or rules require it to stay in specific places.
- What you spend on the tool, and on fixing and watching it, has grown past what it would cost to rebuild properly.
Two mistakes, opposite directions
- Building custom too early. Before the process is stable, or before anyone can maintain code, you pay a lot to build something you'll change next month. Prove the process in low-code first.
- Staying too long. Patching, adding workarounds and treating failures as normal. The cost is quiet: your time, missed follow-ups and lost trust.
The usual healthy path is a hybrid. Start in low-code. Keep the cheap, simple automations there. Replace only the one or two that are fragile, expensive or critical.
Measure these three things
A number from your own business beats a rule of thumb.
- Failures per month. How often did it break, or do something wrong?
- Hours per month spent fixing, watching or working around it, counted at what your time is worth.
- What you pay per month for the tools, plus those hours.
Compare the total with a quote to rebuild it properly. If staying costs more than rebuilding over the next year or two, it's time.
Questions to ask yourself
Stay with low-code if
- The process is still changing.
- A failure is annoying, not costly.
- Your apps have connectors.
- You or a helper can still follow how it works.
Start planning custom if
- It runs something that must be right every time.
- You keep hitting limits or adding workarounds.
- You need a record of what happened and why.
- More people depend on it than can safely maintain it.
Choosing between the low-code tools
| Tool | Tends to suit |
|---|---|
| Zapier | The simplest automations and the widest range of apps |
| Make | Visual, branching workflows at a lower cost per run |
| n8n | Data-heavy and AI workflows, and people who want the option to run it themselves |
Prices and plans change, so check each tool's own pricing. n8n counts a whole workflow run as one execution, while Zapier counts each step as a task. Running n8n yourself removes limits, but you take on the updates, backups and keeping it running.