Choosing the right automation stack for your team
n8n, Make, Zapier, or custom code — how we decide what to reach for before writing a single workflow.
Ask five automation engineers which tool to use and you'll get five different answers, most of them defensible. The honest answer is that the right tool depends less on the tool and more on three questions we ask before touching any of them.
1. How often does this change?
If a workflow is going to be tweaked weekly by someone on your team, a visual builder like n8n or Make earns its keep — non-engineers can read the flow and make small edits without opening a pull request. If it's a stable, high-throughput pipeline that rarely changes, custom code is often simpler to reason about and cheaper to run at scale.
2. What's the failure cost?
A marketing automation that occasionally double-sends an email is annoying. A billing automation that occasionally double-charges a customer is a fire. The higher the cost of a silent failure, the more we lean toward code we can unit test, version, and roll back — with visual tools reserved for lower-stakes, easily reversible steps.
- Does a non-engineer need to edit this workflow?
- Would a silent failure cost real money or trust?
- Does it need to run thousands of times a day?
- Is it talking to more than three external APIs?
3. What does "done" actually look like?
Most automation projects fail not because the tool was wrong, but because nobody defined success before building it. We always start with the manual process fully mapped out, end to end, before deciding what to automate first.
Automate the process you understand, not the process you assume.
Where we usually land
- n8n or Make for anything a client team needs to see and tweak
- Custom TypeScript or Python services for high-volume or high-stakes logic
- A thin orchestration layer connecting both, so neither side has to do the other's job
The tool is rarely the interesting decision. Mapping the actual process — and agreeing on what success looks like — is the part that determines whether the automation still makes sense six months later.