Make vs. n8n vs. Zapier: The Billing Unit Decides Your Bill

The decision rule

These three platforms will all build your automation. They meter it completely differently, and the difference is not marginal — it is an order of magnitude on the same workflow.

• Zapier bills per successful step. A ten-step Zap that runs once consumes roughly ten tasks.
• Make bills per module action. A ten-module scenario that runs once consumes roughly ten credits.
• n8n bills per workflow run. A ten-node workflow that runs once consumes one execution.

So the question that decides your platform is: how many steps does your typical workflow have, and how often does it run? Many steps, run often, and n8n's unit is dramatically in your favour. Few steps, run rarely, and the difference stops mattering and other things decide.

What each vendor says the unit is

Quoted from each vendor's own documentation and pricing page.

Zapier — a task. Counted "whenever Zapier successfully completes a unit of work for you. Failed actions are not counted." Zapier's own page notes this spans both automation (a successful Zap step) and programmatic access such as an MCP tool call. Importantly, a long list of steps do not consume tasks: "triggers, polling for new data, and Zapier built-in data tools (Formatter by Zapier, Paths by Zapier, Filter by Zapier, Delay by Zapier, Looping by Zapier, Sub-Zap by Zapier, Digest by Zapier, Zapier Manager, Storage by Zapier, Tables by Zapier, Forms by Zapier)". That is a bigger carve-out than most people assume, and it materially improves Zapier's effective rate — a ten-step Zap where four steps are filters and formatting bills as six, not ten.

Make — a credit. "Each module action in your scenario, like adding a Google Sheet row or fetching Gmail account data, counts as one credit." Make's page also flags that some advanced features using its AI provider "may use more" than one credit per action, without publishing multipliers.

n8n — an execution. "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 last sentence is the whole competitive position, and n8n knows it.

What each one costs

List prices from each vendor's own pricing page.

Two things to notice before the arithmetic. Make's plan tiers are not volume tiers — Core, Pro and Teams are all quoted at the same 10,000-credit level, so you are choosing features and then choosing volume separately. And n8n prices its cloud in euros, which means your dollar cost moves with the exchange rate.

The arithmetic

Take one realistic workflow: a new form submission is enriched, scored, written to the CRM, routed, and an alert is sent — about twelve steps after the trigger, of which three are filters or formatting. It runs 1,000 times a month. (Triggers are on Zapier's exempt list too, which is why they are excluded from the count on all three platforms here.)

Same automation. Three answers that differ by a factor of twelve.

This is why "which is cheapest" has no answer in the abstract and a very clear answer for your specific workflows. Count the steps in your three busiest automations and multiply by their monthly run count. That single number, applied to the three pricing pages above, decides this for you in about fifteen minutes.

Where the unit stops being the whole story

Zapier's free utility steps are a real discount. Filters, Paths and Formatter not counting means a Zap that does a lot of routing and reshaping bills far below its step count. If your workflows are mostly conditional logic rather than API calls, Zapier's effective rate is better than the raw step count suggests.

Zapier's agents are billed in a different unit entirely. Zapier Agents are metered in activities, not tasks, on their own allowance. If your plan is to replace Zaps with agents, the task maths above does not describe your bill and you need to price the agent product separately.

Make's AI steps may cost more than one credit each. Their page says so without saying how much. If your scenario is AI-heavy, treat 1 credit per action as a floor, not an estimate, and watch actual consumption during the trial.

n8n's free unit has a cost somewhere else. The Community edition is free and self-hosted, which means you own the server, the upgrades, the backups and the pager. That is a genuine engineering commitment, not a footnote. n8n Cloud exists precisely so you can buy the execution model without the operations, and that is what its €20 and €50 tiers are for.

Who each is genuinely wrong for

Zapier is wrong for you if you run high-volume, many-step workflows. The per-step unit is the most expensive model on this page at scale, and no amount of plan-shopping changes the multiplication. Zapier is right when breadth of integrations is the constraint — it connects to more things than anything else, and for a business whose problem is "our niche scheduling tool has no API friends", that breadth is the entire value.

Make is wrong for you if you want the cheapest plan to also be the plan with the features. Its tiers separate capability from volume, so the honest budget is "the feature tier I need, at the volume I need", and those are two multiplications rather than one. It is also the wrong choice if visual complexity bothers you — a large Make scenario is a genuinely dense picture.

n8n is wrong for you if nobody on the team will own infrastructure and you also do not want to pay for Cloud. The per-execution model is the best economics on this page for real workloads, and self-hosting is how most people get it — but "free" software running your revenue-critical automations on a box nobody patches is not a saving, it is a deferred incident.

All three are wrong for you if the workflow you are automating is not yet written down. The most expensive automation is the one that faithfully executes a broken process a thousand times a month.

The recommendation

• Fewer than a dozen automations, low volume, breadth of integrations matters most: Zapier. Take the annual rate if you are past experimenting.
• Moderate volume, cost-led, willing to learn a visual builder: Make. The entry economics are the best of the two hosted options.
• High volume, many steps, and somebody who will own it: n8n. The execution unit is worth more than any discount the other two will offer you.
• You want the execution model and no server: n8n Cloud, and price it in euros.

What breaks in month three

The silent failure. A workflow stops firing — a credential expires, an API changes a field name — and nothing announces it, because the failure is the absence of an event. Build an alert for "this has not run in 24 hours" on every automation that matters. This is the single most common way automation quietly stops delivering value while continuing to be paid for.

The task meter runs out mid-month. On per-step platforms, an unusually busy week can exhaust the plan, and then automations simply stop. Set the usage alert during setup, not after the first outage.

One person built everything. Automation platforms concentrate institutional knowledge in whoever made the workflows. Document the trigger, the destination, and the owner for each one, in a place that is not that person's head.

The unreviewed edge case compounds. Automations handle the happy path immediately and the awkward path never. Sample ten real runs a month and check the ones that took an unusual branch.

The objection you will hear

"Just use whichever one your team already knows."

This is better advice than it sounds, and it is still wrong at scale. Platform familiarity is worth real money in weeks one to eight. The billing unit is worth real money every month forever. If the maths above says your workload is twelve times more expensive on the platform you know, the learning curve is the cheaper problem.

The reasonable version: keep what you know for the long tail of small automations, and put your two or three highest-volume workflows on the platform whose unit fits them. Running two automation tools is not a failure of discipline; paying twelve times over to avoid it is.

Next step

Our earlier Make vs Zapier comparison covers the builder experience and integration breadth; this page supersedes its pricing analysis and adds the third option. If what you actually want is an agent rather than a workflow, Lindy vs n8n vs Gumloop and AI agents vs Zapier cover that distinction properly — they are different products, and picking the wrong category is more expensive than picking the wrong vendor within it.

Vendor detail: Make.com, n8n, Zapier, plus the Lindy vs Make head-to-head. The workflow library has the automations themselves.

If you are choosing a platform before you have chosen a workflow, start the other way round — the free AI Agents Starter Kit walks through picking the first automation that pays for itself.