Make vs Zapier vs n8n comes down to one thing most comparisons skip: the three platforms bill different units. Zapier charges you per action. Make charges per module run. n8n charges per workflow run, no matter how many steps that workflow has. So one automation, at one volume, can cost $489 a month on Zapier, roughly $85 on Make, or $50 on n8n. Same work, same result, three very different bills.
Two notes before the numbers. Every price here is the US-dollar, annual-billed monthly rate; month-to-month costs more on all three. And every one of these platforms prices on a configurable plan, so changing volume or plan settings moves the number, and the figures below are the published starting points, not quotes.
Below is the math, using each vendor’s own published prices as of July 2026. If you only take one thing away: compare cost per workflow run, never cost per “task”, because the word means something different on each platform.
The billing unit is the whole comparison
| Platform | Unit | What consumes one | What’s free |
|---|---|---|---|
| Zapier | Task | Each successful action step | Triggers, polling, Filters, Paths, Formatter, Delay, failed actions |
| Make | Credit (renamed from “operation” in 2025) | Each module run, 1 credit by default | Routers and error-handler modules |
| n8n | Execution | One entire workflow run | Everything inside the run |
Zapier is explicit that a task counts “whenever Zapier successfully completes a unit of work for you,” and that triggers, polling and its built-in data tools don’t count. Make counts each action a scenario performs (reading, searching, creating, updating, iterating a row) at 1 credit each by default.
n8n is the outlier, and says so on its pricing page: “An execution is a single run of your entire workflow. It doesn’t matter how many steps are in the workflow or how much data it processes — it’s still a single execution.”
That single sentence is why the numbers diverge so hard. Add a step to a Zapier Zap and your bill goes up. Add a step to an n8n workflow and it doesn’t.
The published prices
Zapier (Professional plan, monthly rate when billed annually, month-to-month is about 50% more): 750 tasks $19.99 · 5,000 tasks $89 · 10,000 tasks $129 · 20,000 tasks $189 · 50,000 tasks $289 · 100,000 tasks $489. The free plan gives 100 tasks a month, two-step Zaps only, and a 15-minute polling interval; Professional polls every 2 minutes.
Make: Core $9/month billed annually ($10.59 month-to-month), Pro $16, Teams $29. All three include 10,000 credits, and the plan tier buys features, not volume. Higher volumes come off a slider that runs to 8M+. The free plan includes 1,000 credits a month, which is genuinely enough to build and test something.
n8n: Starter $20/month for 2,500 executions, Pro $50 for 10,000, Business $800 for 40,000, billed annually. Month-to-month those become $24, $60 and $960. There’s no free cloud tier, only a trial, but the self-hosted Community Edition is free software you run on your own server.
The same workflow, three bills
Take a normal lead-intake automation: form submission arrives, junk gets filtered out, the company is enriched by an API call, an AI step classifies the lead, a CRM contact is created, a deal is created, a notification email goes out, and a message posts to Slack.
That’s 6 billable Zapier tasks (the trigger and the filter are free), 7 Make credits (every module counts), and 1 n8n execution, per run.
| Runs per month | Zapier | Make | n8n Cloud |
|---|---|---|---|
| 500 | 3,000 tasks → $89 | 3,500 credits → $9 | 500 executions → $20 |
| 2,000 | 12,000 tasks → $189 | 14,000 credits → ~$20 | 2,000 executions → $20 |
| 10,000 | 60,000 tasks → $489 | 70,000 credits → ~$85 | 10,000 executions → $50 |
The Make figures above 10,000 credits are estimates, and we’d rather flag that than pretend otherwise: Make publishes only the 10k price and puts the rest behind a slider. The estimate assumes extra credits at the published +25% markup, so treat it as a ceiling, since the slider tiers are usually cheaper than buying packs.
The shape of the table is the point. At 500 runs a month the gap is pocket change. At 10,000 runs a month you’re choosing between a $489 bill and a $50 bill for identical work, and the only reason is which unit the vendor decided to count.
The gotchas that break the estimate
Make bills per row, not per step. A module that iterates 200 rows spends roughly 200 credits, not one. Scenarios that loop over spreadsheet rows or API pages burn credits far faster than the module count suggests. Some AI modules also carry a fixed cost of 2 or 10 credits instead of 1.
Make punishes overage. Since November 2025, extra credits bought mid-month cost 25% more than the credits included in your plan, whether you buy them manually or on auto-purchase. Sizing the plan right is worth real money.
On Zapier, retries cost. Failed actions aren’t billed, but Autoreplay re-running them is billed again. A flaky third-party API quietly inflates your task count.
On n8n, concurrency is the hidden ceiling. Starter allows 5 concurrent executions, Pro allows 20. If 40 webhooks land in the same second, executions aren’t your problem. The queue is.
Self-hosted n8n is free software, not a free system. You’re running a server, upgrades, backups and a database. If nobody on the team wants that job, the $20 cloud plan is cheaper than the outage you’ll eventually have to fix on a Sunday.
What we do on real builds
We build client automations on whichever of the three fits, deployed on accounts the client owns. The platform decision comes after counting billable units, not before. We write out the step list, mark which steps bill on each platform, multiply by the client’s realistic monthly volume, and only then pick. It takes twenty minutes and it’s the difference between a $20 subscription and a $200 one for the same automation.
The other habit came from getting it wrong on our own stack. When we launched our outreach system, we planned volume against what the technology could handle, and the subscription’s unit, contacts per month rather than emails, turned out to be the lower ceiling. We cut our launch volume by a quarter because of a billing definition. Same lesson as this article, learned the expensive way; there’s more on that in our cold email warmup guide.
So: check the unit before you commit to a platform. It binds sooner than any technical limit.
When none of this matters
If you’re running two or three workflows a few hundred times a month, which is most small businesses starting out, every option here costs between $9 and $30 and you should pick on something else entirely: which builder you can actually maintain, and whether the apps you need are supported. Zapier’s catalogue is the widest and its setup is the easiest; Make is the best visual builder for branching logic; n8n is the one you can host yourself and extend with real code.
Switch to counting units when a workflow gets long, gets looped, or crosses a few thousand runs a month. That’s where the difference stops being a rounding error.
FAQ
Is n8n always cheaper than Zapier? At volume with multi-step workflows, yes, and by a lot, because it bills per run rather than per step. Below a few hundred runs a month the difference is a few dollars, and Zapier’s entry plan is cheaper than n8n’s.
Does Make count the trigger as an operation? Yes. Every module run costs a credit by default, including the trigger. Only routers and error handlers are free. Zapier is the opposite: triggers, filters and paths are free, and only successful actions bill.
Can I self-host to avoid subscription costs entirely? With n8n, yes. The Community Edition is free. You’re trading a subscription for server costs and maintenance time. It’s a good trade if someone on the team already runs infrastructure, and a bad one if this would be your first server.
Which should I pick if I don’t know my volume yet? Start on Make’s free plan (1,000 credits) or an n8n trial, build the real workflow, and let it run for two weeks. Your actual unit consumption will be visible in the dashboard, and then this table gives you the answer directly. Guessing volume in advance is how people end up on the wrong plan.
If you’d rather not do this yourself: we map the process, build the automation on accounts you own, and hand over an operator guide, live within 7 business days. That is Never Do It Twice.