EN
Ihor NikolenkoThree platforms, three sites promising "automate everything, no code," and one question every business owner asks: which one? Getting it wrong doesn't cost you subscription money — it costs months: workflows built on one platform don't transfer to another, and when things get cramped or expensive six months later, everything gets rebuilt from scratch.
I've worked with all three, in the club and in client projects, and I've watched the exact same task cost pennies on one platform and an assistant's salary on another. The difference isn't "which one is better" — it's how each one bills you, where it keeps your data, and how it behaves once a workflow gets more complex than "it arrived, log it."
This article is an honest comparison of Make, n8n, and Zapier across six criteria that genuinely affect a decision for a business with no developer, three common situations with a concrete recommendation, and one piece of advice for not going wrong, regardless of which one you pick.
One Task, Three Ways to Bill For It
The first thing worth understanding is what you're actually paying for. All three platforms sell a subscription, but the unit they bill in is different, and that unit is exactly what determines how much your real-world workflow will cost a year from now.
Zapier bills by tasks: every action in a workflow — logging a row, sending an email — is one task. The trigger itself usually doesn't count, but every step after it does. A five-step workflow that fires a hundred times a day eats up five hundred tasks.
Make bills by operations: every module that runs is one operation. The logic is similar to Zapier's, with one nuance: when a workflow processes a list of twenty items, the module after it runs twenty times — and that's twenty operations.
n8n's cloud version bills by workflow executions: one run is one unit, no matter how many steps or items it processes. The same five-step workflow running a hundred times a day is a hundred units. And on a self-hosted server, n8n doesn't bill for volume at all: the license allows free internal business use, and you only pay for the server.
The practical consequence: for short, infrequent workflows, the difference is minor. For long, frequent ones — which is what working automations eventually become — the bill in Zapier and Make grows along with the complexity, while in n8n it doesn't.
Six Criteria in One Table
| Criterion | Zapier | Make | n8n |
|---|---|---|---|
| What you pay for | Tasks (every action) | Operations (every module) | Cloud executions; self-hosted, just the server |
| Where your data lives | The service's cloud | The service's cloud, with a choice of data-center region | The service's cloud, or your own server |
| Learning curve | Lowest: linear steps | Medium: a visual canvas with branches | Medium: a canvas, expressions, the concept of nodes |
| Complex logic | Branches and filters exist, deep logic is limited | Routers, iterators, aggregators | Any logic, plus code when you need it |
| Working with AI | Built-in model actions | Model modules | Model nodes plus a dedicated agent node with tools |
| Ready-made integrations | The largest catalog | A large catalog | A large catalog plus community nodes and direct HTTP requests |
I'm deliberately not putting numbers on the integration counts: they change every month, and what actually matters for a real business isn't "how many total" but whether your specific five services are on the list. Check that before you choose — all three catalogs are public.
What's Behind Each Row
Data. If you work with clients in Europe, or simply don't want personal data passing through a third-party service, Make offers a region choice, and n8n offers self-hosting, where data never leaves your infrastructure at all. With Zapier, you accept the service's terms as they are.
Learning curve. Zapier is the simplest: a trigger, a few actions, done. That's both its strength and its limit — once you need "if this, then that, otherwise something else, then bring it all back together," the convenience runs out. Make and n8n show a workflow as a diagram on a canvas, where you can see the branches with your own eyes; it takes about a day to get used to, and after that the diagram explains itself.
Complex logic. Make has convenient modules for lists: split, process each one, put them back together. n8n's nodes do the same, and when a ready-made node is missing, there's a code node and direct requests to any API. This is the point where n8n has no ceiling, but it also demands that you understand what you're doing.
Working with AI. All three can call language models. The difference is agents: n8n has a dedicated agent node that picks its own tools and holds a conversation, while in Zapier and Make, the model is usually just one step among others. That matters for a lead-qualifying bot or an assistant with memory.
What doesn't transfer. Workflows don't migrate between platforms. That sentence is worth repeating: when you choose a platform, you're choosing it for years.
Three Situations, Three Answers
"I need to connect two services and forget about it." A form submission into a spreadsheet, a new row into a messenger, an email attachment into a folder. Two or three workflows, infrequent runs. Go with Zapier: the fastest start, the least to learn, and the bill won't hurt at that volume. The condition is that your services are actually in the catalog.
"I have a dozen processes, and they're growing." Leads, reminders, reports, syncing a CRM with spreadsheets, processing lists. This is where the per-operation cost starts driving the decision. Choose Make if the team is non-technical and values a convenient canvas; choose n8n if you'll genuinely end up with a lot of processes and you're willing to spend a day learning the concept of nodes. Both will handle it — the difference shows up in the bill a year from now.
"I need bots, AI, and control over my data." A conversational Telegram bot, an agent with access to a knowledge base, handling personal data, integrations with services that have no ready-made module. That's n8n, and most often, self-hosted. This is exactly how I build workflows for the club and for clients; a ready-made example is the n8n Telegram Traffic Processor, a Telegram traffic-processing workflow you can use as a foundation.
Advice That Matters More Than the Choice
Before opening any of the three platforms, write the process out on paper: what comes in, what happens to it, where the result goes, what to do on error. If the process can't be described in five to seven steps, it isn't ready for automation yet, and no platform will fix that. If it can, the choice becomes obvious from the table above: by volume, by data, by complexity.
And second: whichever platform you pick, build your first workflow on test data, with a separate workflow for error notifications. A silent failure means lost leads, and that's equally possible on all three.
If you're not sure which process to start with, or you're wondering whether your chosen platform will hold up against your plans for the year, that's exactly the kind of question worth bringing to a consultation with a 30-day plan: we go through your processes and come out with a list of what to automate first, and on what.
The Short Version
- Zapier bills by tasks, Make by operations, n8n's cloud version by executions, and self-hosted n8n doesn't bill for volume at all.
- For short, infrequent workflows the price difference is small; for long, frequent ones it drives the decision.
- Data: Make offers a region choice, n8n offers self-hosting; Zapier means accepting the service's terms as they are.
- Zapier is the simplest; Make and n8n show logic on a canvas; n8n has no complexity ceiling, but it demands understanding.
- For agents and conversational bots, n8n has a dedicated agent node with tools.
- Workflows don't transfer between platforms — the choice is made for years.
- Describe the process on paper first; if it doesn't fit into five to seven steps, it isn't ready for automation yet.
Frequently Asked Questions
Can I use two platforms at the same time?
Technically yes, and some teams keep Zapier for simple connections and n8n for complex ones. But that's two places where something can break, and two subscriptions. It's better to have one platform that can handle your plans for the year.
Is n8n really free when self-hosted?
The license allows free internal business use with no volume fee; the restrictions apply to reselling n8n as a service. The real cost is the server itself and the time spent maintaining it.
What should I choose if the team has no technical person at all?
Zapier for your first steps. Once things start feeling cramped, move to Make or n8n, and by that point it's worth having one person you know who's already done this: an hour of their time saves you a week.
Will AI agents make these platforms obsolete?
Quite the opposite: an agent needs a framework that triggers it with events, hands it tools, and keeps a log. That's exactly what automation platforms do, and n8n already builds agents right inside its workflows.
