All three platforms can build most SME workflows. They differ in how you pay, where your data runs, how much complexity they tolerate and how easy they are to leave. This comparison is written from the point of view of someone who has to keep the workflow running for years, not just demo it.
The short version
Zapier is the fastest way to connect two mainstream apps and the easiest to explain to a non-technical owner. Make is the strongest visual builder for multi-step logic with a generous free tier. n8n is the most capable for technical workflows, offers self-hosting and prices by execution rather than by step. None of them is best for everything, and the pricing model matters more than the feature list.
Pricing models: executions, credits and tasks
This is the difference that surprises people six months in. Check the official pricing pages linked below for current figures; the models themselves have been stable.
- n8n bills per execution. One run of a workflow counts once, regardless of how many steps it has. Hosted Starter is listed from €20/month and Pro from €50/month billed annually; a self-hosted Business tier is listed at €667/month for enterprise controls. A 40-step workflow costs the same to run as a 3-step one.
- Make bills per credit (operation). Every module that runs consumes an operation. Plans range from a free tier through Enterprise. A workflow that loops over 200 records and does three things to each consumes 600 operations per run.
- Zapier bills per task. Each action step that does something counts as a task; the free plan starts at 100 tasks a month, and paid tiers scale with task volume. Multi-step Zaps with several actions multiply the cost quickly.
The practical rule: low-volume, few-step workflows are cheap everywhere. High-volume or many-step workflows favour execution-based pricing or self-hosting. Whatever you choose, the platform fee should be paid in your own account so you can see it.
Decision table
| Criterion | n8n | Make | Zapier |
|---|---|---|---|
| Pricing unit | Execution | Credit / operation | Task |
| Self-hosting | Yes (community and licensed editions) | No | No |
| EU data location | Your choice when self-hosted; EU cloud region offered | EU region available | US-hosted cloud; assess transfer basis |
| Complexity tolerated | High: code nodes, loops, sub-workflows | Medium-high: visual routers, iterators | Low-medium: linear with paths |
| Integration breadth | Large, plus any HTTP API | Large | Largest catalogue |
| AI features | Native model and agent nodes, memory, tools | AI modules and assistants | AI steps and agents within Zaps |
| Lock-in | Low: JSON export, self-host option | Medium: blueprint export, cloud only | Medium: cloud only, export limited |
| Best fit | Technical workflows, ownership, volume | Visual multi-step logic | Fast SMB integrations |
Data-location and feature details change; verify against the vendor’s current documentation before relying on them for a data-protection assessment.
Self-hosting and data location
For DACH businesses, where the workflow runs is often the first question. n8n is the only one of the three that can run on your own server or in a cloud region you choose, which keeps personal data inside infrastructure you control and simplifies the processor inventory. It also makes you responsible for updates, backups and uptime, which is a real cost. Make offers an EU region. Zapier runs in the US, so workflows carrying personal data of EU or UK residents need a documented transfer basis. This is a privacy-by-design decision, not a compliance badge; we identify where specialist legal review is required.
Complexity and maintainability
Zapier’s linear model is a feature until your workflow needs a loop, a retry with backoff or a wait for an approval. At that point you are working around the tool. Make handles branching and iteration well visually, but large scenarios become hard to read and review. n8n lets you write small pieces of code where the visual approach gets awkward, which keeps workflows readable but requires someone who is comfortable with JavaScript.
Maintainability matters more than initial build speed. Ask: can a new person open this workflow in a year and understand what it does? Can we test it against sample data before deploying a change? Can we export it and version it? These questions favour n8n and Make over Zapier for anything beyond simple syncs.
AI features
All three now offer AI steps and some form of agent. The useful question is not whether a platform “has AI” but whether you can bound it: set a confidence threshold, route uncertain cases to a person, log every model output and cap cost. n8n’s native model nodes give the most control. Make and Zapier offer simpler AI modules that work well for classification and drafting inside a controlled workflow. For knowledge retrieval over your own documents, n8n or custom code is usually the better foundation.
Lock-in
Lock-in has two sources: the platform and the provider who built the workflow. n8n’s JSON export and self-hosting option make the first the smallest. On any platform, the second is solved by building in accounts you own, documenting every workflow and handing over credentials and exports. If a provider builds on their subscription and offers you “access”, you do not own the workflow.
When to choose Power Automate or custom code instead
Microsoft Power Automate is the right answer for businesses that live in Microsoft 365, need desktop or legacy automation (RPA), or must stay inside a tenant their IT department already governs. Public pricing lists Premium at $15 per user per month and Process at $150 per bot per month; licensing needs careful scoping because per-user and per-flow models can cost very different amounts for the same workflow.
Custom code wins when logic is unusual, volume is very high, unit cost matters, or the workflow needs to become part of a product. It gives the most control and the most maintenance responsibility. We recommend it when a platform would need workarounds in more than a few places.
Our platform-neutral rule
We choose the simplest platform that reliably meets the requirement, per workflow, in the client’s own account, with an export path. That usually means Zapier for a quick two-app sync, Make or n8n for anything with branching and approvals, n8n self-hosted where data location or volume matters, Power Automate in Microsoft-centric estates, and code where a platform would fight us. We do not resell platform subscriptions, and the choice is documented in every proposal so you can challenge it.
Next actions
- Estimate monthly runs and steps per run for your first workflow, then price it on each vendor’s page using their own unit.
- Decide whether personal data may leave the EU or UK in this workflow. If not, shortlist accordingly.
- Ask any provider which platform they would choose for your case and why, and whether the workflow will live in your account. See how we deliver.
Sources
- n8n, Pricing, 2026 — https://n8n.io/pricing/
- Make, Pricing, 2026 — https://www.make.com/en/pricing
- Zapier, Pricing, 2026 — https://zapier.com/pricing
- Microsoft, Power Automate Pricing, 2026 — https://www.microsoft.com/en-us/power-platform/products/power-automate/pricing
- ICO, AI and data protection guidance, 2025 — https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/artificial-intelligence/