Cost & ROI
Zapier vs Make vs n8n vs Custom: An Honest Comparison
Four ways to automate a workflow, what each genuinely costs at real volume, and the specific point where building your own becomes the cheaper option.
· 9 min read · Flowmatix
The short answer
Zapier is fastest to start and most expensive at volume. Make is cheaper per operation with a steeper interface. n8n is cheapest at scale if someone will host and maintain it. Custom code wins when volume is high, logic is unusual, or a system has no connector — typically past a few thousand operations a month.
Every automation project starts with the same question and it is rarely asked plainly: should we wire this together with a tool, or should someone write it. Here is the comparison without the affiliate links.
The short version
| Option | Best when | Where it hurts |
|---|---|---|
| Zapier | Few hundred runs a month, common apps, nobody technical available | Cost climbs steeply with volume; limited handling of unusual logic |
| Make | Moderate volume, multi-step branching, budget-conscious | Visual editor gets dense fast; still per-operation pricing |
| n8n | High volume, technical person available, data must stay in-house | Someone owns the hosting, upgrades and outages |
| Custom | Unusual logic, no connector exists, volume is large or growing | Upfront cost and a real handover; wrong choice for a simple two-app link |
How the pricing actually behaves
The tools are priced per task or per operation, and the word operation is doing more work than most buyers realise. A single business event — a lead arrives, gets enriched, scored, routed, logged and acknowledged — is not one operation. It is six. At a hundred leads a day that is eighteen thousand operations a month from one workflow.
This is why automation bills surprise people. The pilot ran fifty leads and cost nothing. Production runs three thousand and the plan tier moved twice. Model your real volume before choosing, not your pilot volume.
Where each one genuinely wins
Zapier
The largest connector library and by far the gentlest learning curve. If your workflow is a straight line between two popular apps and it runs a few hundred times a month, this is the right answer and building anything custom would be indefensible.
Make
Meaningfully cheaper per operation and much better at branching, loops and error paths. The trade is interface density — a complex scenario becomes a diagram that only its author can read six months later. Document it as you build.
n8n
Self-host it and the marginal cost of an operation approaches zero, which changes the arithmetic entirely at volume. You also keep the data inside your own infrastructure, which matters for regulated work. The cost has simply moved from a subscription to a person: someone updates it, monitors it, and is responsible when it stops at 2am.
Custom
Worth it when at least one of three things is true. Your logic does not fit a node graph — real pricing rules, exceptions, approvals. A system you depend on has no connector, or a bad one. Or your volume is high enough that per-operation pricing has become a permanent tax.
The hidden costs nobody quotes
- Every extra step is another thing that can silently fail. Tools rarely tell you loudly when a run stops halfway.
- Connectors change. A vendor deprecates a field and your workflow breaks quietly on a Tuesday.
- The person who built it leaves. Undocumented scenarios are the most expensive kind of institutional knowledge.
- Per-operation pricing punishes exactly the success you are hoping for.
How to choose in one afternoon
- 1Write the workflow out in sentences. Every step, every exception, every approval.
- 2Count the steps and multiply by realistic monthly volume. That is your operation count.
- 3Price that on each platform's calculator, at twelve months, not one.
- 4Check every system involved actually has a working connector. Test it, do not trust the logo grid.
- 5If the answer is close, choose the tool. Reversibility is worth more than a small saving.
The honest position
We build custom automation, and we still tell most enquiries to start on a tool. Most workflows are simple enough that a subscription beats a build. The ones worth building are the ones where the tool has already failed you in a specific, describable way — and if you cannot name that way yet, you are not ready to build.
Frequently asked questions
Is n8n really cheaper than Zapier?
At volume, substantially — self-hosted, the marginal cost per operation approaches zero. At low volume it is usually more expensive once you count the hours someone spends hosting, updating and monitoring it. The crossover depends on whether that person's time is free.
When should a small business build custom automation instead of using a tool?
When the logic does not fit a node graph, when a critical system has no usable connector, or when per-operation costs have become a permanent tax on growth. Below a few thousand operations a month, a tool is almost always the right answer.
Can I move from Zapier to custom later?
Yes, and it is a common path. Starting on a tool is also the cheapest way to specify the eventual build, because a year of running the workflow teaches you the exceptions no requirements document would have captured.