Cost · n8n vs make vs zapier
n8n, Make or Zapier: the billing unit decides the cost, not the sticker price
Zapier bills per step, Make per module action, n8n per workflow run. On a 12 step workflow that is a 12 to 1 difference. Prices checked August 14, 2026.
Lasse Pettersen
Three platforms, three sticker prices, and the sticker price is the least useful number on any of their pages.
Zapier bills you per step. Make bills you per module action. n8n bills you per workflow run. On a twelve step workflow that runs 1,000 times a month, the first two charge you roughly 12,000 units and the third charges you roughly 1,000. That single difference moves the annual cost more than any discount either of the first two will ever offer you.
All prices below were checked on August 14, 2026. All three change them, so verify before you commit.
The prices, as published
| Platform | Entry paid plan | What that includes | Billing unit |
|---|---|---|---|
| Zapier | $19.99 a month billed annually, $29.99 monthly | 750 tasks | One task per action step |
| Make | $9 a month for Core | 10,000 credits | One credit per module action |
| n8n Cloud | €20 a month billed annually for Starter | 2,500 executions, unlimited steps | One execution per workflow run |
Free tiers: Zapier 100 tasks a month, Make 1,000 credits a month, n8n self-hosted Community edition at no licence cost.
Higher tiers: Zapier Team from $69 a month billed annually for 2,000 tasks with up to 25 users; Make Pro at $16 and Teams at $29 for 10,000 credits; n8n Pro at €50 a month for 10,000 executions and Business at €667 for 40,000 with self-hosted deployment.
Two things a Canadian buyer should notice immediately. Zapier and Make publish in US dollars and n8n publishes in euros, so all three cost you the exchange rate on top, and none of these figures is a CAD price. And Make moved its billing unit from operations to credits in late 2025, which is the kind of change that quietly reprices an existing workflow.
Why the billing unit is the whole decision
Take an ordinary job: an order form arrives, the data is cleaned, a customer record is looked up, a record is created in the CRM, an invoice draft is created in accounting, a notification goes to a channel, and a row is appended to a sheet. Call it twelve steps once you count the branches.
Run it 1,000 times a month:
- Zapier: about 12,000 tasks. That is well past the entry tiers and into the volume where the annual number becomes a real line item.
- Make: about 12,000 credits, so the 10,000 credit plans do not cover it.
- n8n: 1,000 executions, comfortably inside a Starter plan at 2,500.
Now run a two step job 1,000 times a month and the ranking inverts in practical terms: 2,000 tasks on Zapier is cheap, and the setup time you save is worth more than the subscription difference.
The rule that falls out of this: short workflows at low volume favour Zapier. Long workflows at any volume favour n8n. Make sits between them and wins on the visual builder when somebody in the business is going to maintain it themselves.
The cost nobody puts in the comparison
Your own time, and the time of whoever maintains it after you.
Zapier is the fastest of the three to build in and the most forgiving of somebody who does not think in data structures. Make is more capable per action and demands more of the person building. n8n assumes the most and gives the most back, and self-hosting it adds a server, updates, backups and credential storage to somebody’s job description.
A business in Waterloo with a technical founder and a business in Barrie with none should not choose the same platform, and the subscription is not why.
Where a Canadian business actually differs
PIPEDA does not require your data to stay in Canada
This belief costs Ontario businesses money, and it is wrong. The Personal Information Protection and Electronic Documents Act does not prohibit transferring personal information outside Canada for processing, and there is no private-sector data localisation requirement.
What the Act does impose is accountability. Your organisation stays responsible for personal information in a processor’s hands, the protection is achieved principally through the contract with that processor, and the Office of the Privacy Commissioner of Canada expects you to be transparent with customers that their information may be processed in another jurisdiction.
So the question is not “is it in Canada”. The question is what the contract says, what the processor is permitted to do with the data, and whether your privacy policy tells your customers the truth. That reasoning is worked through in what PIPEDA means for the AI tools your business uses.
When self-hosting is still the right answer
Not because the law demands it, but because it shortens the work.
If your workflows carry client health information, financial records, or anything that identifies a person in a way that would matter if it leaked, self-hosted n8n lets you put the processing on infrastructure you have already assessed, under a contract you have already signed. That removes a diligence exercise rather than a legal obstacle, and on a professional services or clinic workflow it is frequently worth the server.
If your workflows carry order numbers, stock levels and internal notes, self-hosting is a hobby with a server bill.
Licensing, which matters if you resell
n8n is not open source in the conventional sense. It is released under a fair-code licence, with a self-hostable Community edition on GitHub. That is fine for running your own business on it and it has conditions if you intend to offer it to others as a service. Read the licence before building a product on top of it. Zapier and Make are proprietary hosted services and the question does not arise.
What I would actually choose
- Zapier, if you have fewer than about five workflows, each under five steps, and nobody technical. You are buying speed of setup and you should not optimise the subscription.
- Make, if somebody in the business enjoys building things and you want them to own it. The visual model holds a complex scenario in one view better than the other two.
- n8n, if you run long workflows, high volume, or data you would rather keep on your own infrastructure. Self-host only if a named person owns the server.
One thing none of the three controls is the cost of the compute underneath them, and that has stopped being a settled question. Ontario has proposed a separate electricity rate class for the large data centres this industry runs on, which is set out in the province’s data centre consultation and the 1 MW threshold in it.
And the answer that is not on the list: none of them, if what you are automating has not been written down yet. A platform choice made before the process is documented is a decision about tooling made in place of a decision about work, and it is the most common way these projects produce a subscription and no saving. The test is in which jobs to automate first.
What a build costs on top of the subscription
The platform is the small number. On this site a workflow automation build is $2,500 to $12,000, driven mostly by how many systems have to talk to each other, and the full picture including subscriptions and maintenance is in what AI consulting costs in Ontario.
Statistics Canada put Canadian business AI use at 19.2% in the second quarter of 2026. Most of the businesses in the other 81% do not need a platform recommendation. They need somebody to name the job first, which is what the $999 assessment does, and the platform choice falls out of the job rather than the other way round. If you already know the job and want it built, that is workflow automation.