Zapier Task vs Make Credit vs n8n Execution: How Automation Platforms Count

Four ways to meter the same automation

Every automation has the same shape: a trigger (the event that starts it), then steps that act on the data. The platforms differ in which part they meter. Zapier counts successful action steps.¹ Make counts module runs, item by item.⁵ n8n and Activepieces count whole runs, however long.¹⁰¹³ Pipedream counts compute time,¹⁵ and IFTTT counts how many automations you keep.¹⁴ So a short workflow that runs often and a long one that runs rarely can cost very differently on each.

Zapier: a task is a successful action

Zapier's definition is short: "A task is any successful action that runs in Zapier." The trigger never uses tasks, and neither do the polls Zapier makes while it checks an app for new data, even at more than 20,000 polls a month. Filter and Paths steps are free, and so is any step that doesn't run because a filter stopped the Zap or an earlier step failed. Built-in helper apps (Formatter, Delay, Looping, Digest, Storage) and Zapier Tables and Forms steps are free too.¹

Zapier's own example: a Zap that uploads an email attachment to Dropbox and then messages Slack uses 2 tasks per run, so 20 tasks for ten runs. Some products bill at other rates: AI by Zapier depends on the model tier picked, Zapier MCP uses 2 tasks per successful tool call, and Lead Router uses 5 per lead routed. Zapier Agents don't use tasks at all; they have a separate activity quota.¹

When a paid plan runs out, pay-per-task billing keeps Zaps running, up to 3 times the plan's task limit; past that, runs are held until the billing cycle resets.³ The entry paid plan is $29.99 / mo · billed monthly · 750 tasks or $19.99 / mo · billed yearly · 750 tasks; the free plan is Free · 100 tasks / mo · two-step Zaps.²

Make: credits come from operations, and operations multiply

Make renamed its billing unit from operations to credits; plans and credit prices stayed the same. For ordinary apps, and for third-party AI apps on your own connection, 1 operation equals 1 credit.⁴ An operation is "a single module run to process data or check for new data", and a module runs once for every bundle (one email, contact or file) it receives. A trigger is the exception: it runs once per check, however many bundles it returns.⁵

That is where Make's counts grow. In Make's example, a form trigger returns 10 responses and three more modules each run 10 times: 1 + 30 = 31 operations for one scenario run. A Text aggregator that combines items into one cuts another example from 11 operations to 3.⁵ Filters sit on the connecting line between two modules, not in a module, and Make's tutorial says they "save operations because they stop the flow when your conditions aren't met".⁶

Some modules break the 1-to-1 rule. Certain Make AI Content Extractor modules use 2 or 10 credits per operation, and Make's built-in AI apps on Make's AI Provider charge by tokens: on the Large tier (GPT-5 mini), 1 credit buys 3,616 input tokens or 452 output tokens.⁴ Extra credits bought beyond a plan cost 25% more than the plan's own credits.⁸

n8n: one execution per run, whatever is inside

n8n defines an execution as "a single run of your entire workflow. It doesn't matter how many steps are in the workflow or how much data it processes."⁹ Only production executions count: runs started by triggers, schedules or polling. How often that happens depends on the trigger. A Schedule Trigger counts every time it fires, whatever the outcome; a polling trigger counts only when it finds new data; a webhook counts every inbound request, including one with an empty body.¹⁰

Test runs from the editor don't count, and neither do sub-workflows (only the parent run counts), error workflows or malformed webhook requests.¹⁰ n8n gives its own rule of thumb: a workflow on a 5-minute schedule comes to about 8,600 to 8,900 executions a month.⁹ The entry Cloud plan, Starter, includes 2.5K executions a month for $24 / mo · billed monthly or $20 / mo · billed annually.⁹

Activepieces, IFTTT and Pipedream

Activepieces also bills by run: "A run is one credit however many steps it has." AI steps are the exception, because Activepieces pays the model provider: 2 credits on a fast model, 10 on a smart one, 20 on a frontier one, and 1 credit per agentic action. On Plus and up, connecting your own AI provider keys brings every AI step down to 1 credit, with the model cost moving to your provider's bill.¹³ Plus includes 10,000 credits a month for $20 / mo · billed monthly, and runs past the allowance cost $0.007 per credit; on the free plan, runs past the quota wait for the credits to refresh.¹³

IFTTT doesn't meter runs on its plans page. It prices by how many Applets (its name for an automation) you can keep: 2 on the free plan, 20 on Pro and unlimited on Pro+. Multi-action Applets start on Pro, and filter code (conditions written in code) is a Pro+ feature.¹⁴

Pipedream charges Workflows by compute: one credit per 30 seconds per workflow segment at the default 256 MB of memory, so a 1-second or a 15-second run is 1 credit and a 35-second run is 2. A delay or a branch splits a run into segments, each billed separately, and doubling memory doubles the credits. Testing is free, and so is a trigger source from Pipedream's public registry. Pipedream says Workflows and String shut down on March 31, 2027; Pipedream Connect continues.¹⁵

A worked example: 5 steps, 1,000 runs a month

Take one workflow: a new order arrives by webhook (the trigger), a filter lets only paid orders through, and four actions follow: add a row to a spreadsheet, create an invoice, email the customer and post a note in a chat app. Say it runs 1,000 times a month, each order is one item of data, there are no AI steps, and every order passes the filter. The table applies each vendor's published counting rule to that run. It is our calculation, not a vendor's quote.

Our calculation: each vendor's published counting rule applied to one run of the workflow described above, then multiplied by 1,000 runs where the rule is per run. It assumes one item of data per run, no AI steps, no failed steps and no retries. These counts are not comparable across rows, because the units differ. Zapier's free plan allows two-step Zaps only, so this 5-step Zap runs on a paid plan, and 4,000 tasks is above the 750-task entry tier.² Make's rule for a scheduled trigger is "one per check", so the total then depends on the schedule.⁵

The filter shows the difference most clearly. When it stops an unpaid order, Zapier counts nothing, because the trigger and filter are free and no action ran.¹ Make still counts the trigger's operation.⁵ n8n and Activepieces count the run, because their unit is the run itself.¹⁰¹³ And if the trigger delivered ten orders at once, Make's four modules would each run ten times, while n8n would still count one execution.⁵⁹

The time side: who runs the servers, and how fast triggers check

Price per unit is half the picture; the other half is what stays on your plate. On every cloud plan here, the vendor runs the servers, applies updates and keeps the platform up. n8n's free Community edition is the exception: n8n's own comparison lists installation as "Requires setup (npm, Docker, or server)", technical expertise as "Required for installation and configuration", hosting as infrastructure the user provides and manages, and maintenance as "Your responsibility". On n8n Cloud, all of that is "Handled by n8n".¹¹ n8n's quickest self-hosted route is a one-line setup script that needs Docker on Linux or macOS (or WSL on Windows).¹²

Speed is the other effort cost. A polling trigger only notices new data when it checks. Zapier checks every 15 minutes on its free plan, every 2 minutes on Professional and every minute on Team and Enterprise.² A Make scenario runs every 15 minutes by default, and the shortest interval depends on the plan.¹⁷ A webhook trigger, on any of these platforms, starts the run as soon as the event arrives.