Zapier changed how it bills. Most advice hasn’t caught up.

Zapier bills per successful action. Triggers do not count. Filters and Paths do not count. Neither do Delay, Looping, Storage, Digest, Sub-Zap, Tables, or Forms. Most internet advice still talks as if “tasks” were a vague tax on every step in a Zap — or, worse, as if a filter always made the whole run free the way people remember from older posts. If you are comparing Zapier to Make or n8n on a blog post from before mid-2025, you are probably comparing three different meters with outdated names.

What Zapier counts

The unit is a successful action. When a step that counts completes successfully, it increments usage. When a step is in the free set, it does not.

Do not count:

• Triggers
• Filters
• Paths
• Delay
• Looping
• Storage
• Digest
• Sub-Zap
• Tables
• Forms

Do count the actions that actually do work against other apps — the creates, updates, searches, and sends that succeed. A Zap that triggers, hits a Filter that stops the run, and never reaches an action-heavy step should not burn the same budget as a Zap that fan-outs into several successful app steps.

That is why step order still matters, but not for the reason many tutorials claim. You are not trying to “hide” billable work behind a filter so the filter magically erases earlier charges. You are trying to avoid running successful billable actions when a cheap Filter or Path could have stopped the workflow first. Put the gate before the expensive successful actions. Leave the free control steps to do their job.

Paths are in the free set. Use them to branch without treating every branch scaffold as a billable event. The billable events are still the successful actions inside the branches that run.

What this does to old advice

Three pieces of folk wisdom need an update:

1. “Filters make Zaps free.” Filters themselves do not count. Successful actions that already ran before a stop still can. Design for fewer successful actions, not for superstition about the Filter step.
2. “Every step is a task.” Triggers and the free control/utility set above are not successful actions in the billing sense Zapier documents now.
3. “Just move to Make — operations are cheaper.” Make’s unit changed. Comparing Zapier actions to Make “operations” as if that word still defined the product will mis-price the migration.

Re-read your own Zap history with the successful-action definition in mind. Sort by Zaps that produce many successful actions for little business outcome. Those are the redesign candidates. Zaps that are mostly trigger → filter → stop are not the villains older essays made them out to be.

Make: operations became credits

Make switched from “operations” to “credits” on 27 August 2025. If your comparison spreadsheet still has an operations column, it is stale.

Under the credits model, routers, filters, and error handlers are free. That is the important structural rhyme with Zapier: control flow is not where Make wants to charge you. The work modules are.

When people say “Make is cheaper for branching,” check whether they are citing pre- or post-27-August-2025 rules. Free routers and filters change the economics of fan-out and guarded flows. A scenario that looked expensive when every router hop burned an operation can look different when those hops are free and only the credit-bearing modules matter.

Entry plan to know: Make Core is $9/month for 10,000 credits. That is the landmark for a small store testing whether Make’s credit math beats Zapier’s action math on the same workflows — after you remap each scenario into credits, not after you paste an old operations estimate.

n8n Cloud: one execution per run

n8n Cloud bills per execution — one whole workflow run. It does not meter the same way Zapier meters successful actions inside a run, and it is not Make’s credit ledger.

That difference dominates architecture choices:

• On Zapier, a single run with many successful actions can be expensive even if you only “meant” one business event.
• On n8n Cloud, the run is the unit. A dense workflow that does a lot inside one execution is a different cost shape than the same logic exploded into many small Zapier actions.
• On Make, you are back to summing credits across modules (with routers, filters, and error handlers free).

None of these is universally cheaper. They are different shapes. A chatty, step-heavy integration style is punished hardest on an action meter. A “one big workflow per event” style maps more naturally onto an execution meter. Branchy control flow is less scary on Make when routers and filters are free — provided the credit-bearing modules stay in check.

Entry plan to know: n8n Cloud Starter is €20/month for 2,500 executions. Note the currency. Do not paste €20 into a USD-only budget sheet without converting.

Entry prices side by side

For a first pass on a one-person Shopify stack:

Platform	Entry landmark	Unit
Zapier Professional	from $19.99/month	successful actions (see free list above)
Make Core	$9/month for 10,000 credits	credits (routers, filters, error handlers free)
n8n Cloud Starter	€20/month for 2,500 executions	executions (one whole workflow run)

This table is not a recommendation. It is a unit converter. The cheapest logo on the sticker is often the wrong winner once you express your real workflows in each unit.

How to re-price a workflow you already run

Take one Zap you care about — order to Slack, refund to sheet, VIP tag to email tool — and write it three ways:

1. Zapier: Count expected successful actions per business event. Ignore triggers, Filters, Paths, Delay, Looping, Storage, Digest, Sub-Zap, Tables, and Forms. Multiply by monthly events. Compare to a Professional plan starting from $19.99/month and your action allowance.
2. Make: Rebuild with routers/filters/error handlers treated as free. Sum credits for the modules that bill. Compare to Core at $9/month for 10,000 credits.
3. n8n Cloud: Count executions — how many times the workflow should run per month — against Starter at €20/month for 2,500 executions.

If the three estimates disagree wildly, you have found a workflow whose shape favours one meter. That is useful information. Migrate the shaped ones first; leave the ambiguous ones until you have measured.

A Shopify-shaped example without fake volumes

Think about a common store pattern: new paid order → enrichment → internal notify → maybe a sheet row → maybe a helpdesk tag. On Zapier, the trigger is free; Filters and Paths that decide whether this order is worth the rest of the Zap are free; Delay and Storage you use to debounce are free. The successful actions against Shopify, Sheets, Slack, or Gorgias are what move the meter. On Make after 27 August 2025, the routers, filters, and error handlers that express the same branching are free; credits attach to the modules that do the work. On n8n Cloud, the question collapses to how many times the workflow runs — one execution per whole run — against Starter’s 2,500 executions for €20/month.

You do not need invented order counts to see the design lesson. If your Zap performs several successful actions for every order, Zapier Professional from $19.99/month may still be fine until those actions multiply. If your Make scenario was priced in your head using pre-credits “operations,” rebuild it against Core’s $9/month for 10,000 credits with free routers and filters before you declare a winner. If n8n lets you keep the logic in one workflow per event, you are buying executions, not a tour of every internal step.

What to stop doing

• Stop trusting migration Twitter threads that still say “operations.”
• Stop treating every Zap step as billable.
• Stop assuming n8n’s execution price maps one-to-one onto Zapier’s action price.
• Stop designing Filters as moral cleansing for earlier successful actions; place free gates before billable work instead.

And when Zapier, Make, or n8n publish another metering change, assume the blog posts will lag by months. Re-read the vendor’s billing docs, not the SEO results for your own query.

If you want those platform choices placed next to helpdesk and email metering in one operational picture, that is what Your First Ten Hires is for.