In short: The bill follows the unit, not the logo. On Zapier, a successful action counts and a trigger, filter or failed step does not, with a change for Next Gen Zaps from 1 December 2026. On Make, most non-AI operations are one credit, and AI usage can vary. Do the count for your own volume. This page has no prices.
People ask whether Zapier is “worth it” and why it feels expensive. Both phrases appeared as search suggestions on 8 November 2026. They are not a volume and not a verdict. A self-described startup owner asked if Zapier was worth the cost because the pricing felt high. A reply from someone who identified themselves as an automation consultant said, in substance, that Zapier is easier for non-technical people and costs more. That is one opinion. A much older thread, from 2019, asked whether anyone pays for Zapier. A reply from a self-described boutique agency owner talked about attention rather than time saved. The prices in that year are not usable. Another owner described a multi-tool sync and named a monthly figure. That figure is not repeated. It is their bill, not a benchmark.
This page explains the unit so you can estimate your own usage. It does not rank products and it does not print a price. Open the vendor’s pricing page on the day you buy. The cost guide is a different job: comparing an implementation quote, licences and handover. This one is only how the meter runs after the workflow exists.
Zapier: a task is a successful action
Zapier’s help article “How is task usage measured in Zapier?”, updated 5 November 2026, is the source for everything in this section. It describes Classic Zaps, and it flags a change for Next Gen Zaps.
A task is a successful action. In a Classic Zap:
- A successful action step is one task each time it runs.
- Triggers do not use tasks. Polling on a timer does not use tasks, even if it checks thousands of times. The article’s own example is a Facebook poll every two minutes: the checks are not tasks; a successful send to a CRM is.
- Filter steps and Paths steps do not use tasks.
- Action steps that error or halt do not use tasks. Steps that never run do not use tasks.
- A successful step on an error-handler path does use a task.
- Replaying a whole run uses tasks again for the successful steps.
- Search steps have a special case: “proceed if nothing is found” uses a task; “do not proceed if nothing is found” does not.
- Testing inside the editor does not use tasks. The action can still happen in the connected app.
- Formatter, Delay, Looping, Digest, Storage, Sub-Zap, Zapier Manager, Zapier Tables and Zapier Forms steps do not count as tasks in the list on that page. Other products on the same page have different rates: AI by Zapier depends on a model tier, Lead Router is listed at five tasks per successful lead, Zapier MCP at two tasks per successful tool call. Those rates are Zapier’s, on that date. Recheck them.
The article’s worked example: a Zap receives an email, uploads a file, and sends a Slack message. The upload and the Slack message are two tasks. Ten runs are twenty tasks. The trigger is free. That arithmetic is Zapier’s, not a price.
From 1 December 2026, the same page says Next Gen Zaps and the Zapier SDK also use tasks, and that each completed Next Gen run uses at least one task. Filters, formatting, loops and branching do not use tasks on their own in Next Gen, but a run that only does those things still uses one task. If you are choosing a style of Zap in late 2026, ask which style the quote assumes. A filter-only run that costs nothing on Classic can cost one task on Next Gen after that date.
Zapier Agents, the page says, do not count as tasks. They use a separate activity quota. Do not compare an agent quote and a Zap quote as if the meters were the same. The agent guide is about when you want an agent at all.
When the plan limit is reached, Zapier’s “Task limits and pricing” section says extra tasks are charged if pay-per-task billing applies, with email at the plan limit and again as you approach and hit an extra allowance it describes as three times the plan. Beyond that, tasks are held. Plans without pay-per-task, or with it switched off, hold new runs at the plan limit. Held runs can be replayed after the cycle resets, an upgrade, or turning pay-per-task on. “The Zap stopped” may be a limit, not a bug. Check usage before you rebuild.
Make: a credit is not always one operation
Make’s help page on credits, checked for this article, says credits replaced “operations” as the name of the billing unit, and that existing plan pricing, including the cost of a credit, was unchanged by the rename. It also says the cost of a credit varies by plan. This article still does not quote that cost.
How the page says credits are used:
- Non-AI apps: one operation equals one credit.
- Third-party AI apps such as OpenAI, Anthropic Claude and Gemini, where you create a connection: one operation equals one credit, and you pay the AI provider for tokens.
- Make’s own AI provider, used in Make AI Agents and Make AI Toolkit: credits depend on tokens and operations. The page publishes token-to-credit tables and says those rates can change as models change. They are not copied here because they are a price list in another shape. Read them on the day.
- Some modules are tagged as more than one credit, or as varying with file size, pages or runtime.
A previous public summary of Make said each bundle processed by an action counts. The credits page checked now speaks in operations and, for AI, tokens. Do not plan on the older sentence. Count operations in the scenario you will actually run, and treat every AI module as “open the credit tag before you assume it is one”.
Only features triggered by a scenario run, or an AI agent’s chat, use credits, the page says. Building the scenario is not the meter. Running it is.
Power Automate: the meter is the licence, and it is easy to assume wrongly
Microsoft’s licensing is not a task counter you can copy from Zapier. A flow may be covered by a user’s licence, by a limited entitlement inside a broader Microsoft subscription, or by a Process allocation, and premium connectors are not something you should assume are included. The Power Automate approval guide is about the 30-day run limit, which is a different constraint from the bill. Before you design, open Microsoft’s current licensing page and match the connectors on the diagram to the licences you hold. If that match is not done, “we already pay for Microsoft 365” is a hope. This article does not state a price or a request allowance, because those figures go stale and were not the point of the count below.
n8n: you swap a task meter for a host
Self-hosted n8n does not bill you per task in the Zapier sense. You pay for the server, the time to patch it, and the risk of the data you store in executions. That trade is the subject of the n8n and PDPA guide. No hosting price is given here. A self-hosted tool is not free because the invoice is absent.
Illustrative count, not a quote
Illustrative calculation, not a client, not a tariff and not a recommendation to buy. An enquiry creates a CRM row, a Slack message and a spreadsheet row. That is three successful actions after the trigger.
At a made-up volume of 200 enquiries in a month, a Classic Zapier design uses 200 × 3 = 600 tasks, using Zapier’s own rule that the trigger is free and each successful action counts. If a filter drops 50 spam enquiries before those actions, the actions run 150 times: 450 tasks. The filter itself, on a Classic Zap, does not add tasks. On a Next Gen Zap after 1 December 2026, a run that is only a filter still uses at least one task, so the 50 dropped enquiries are not necessarily free. Ask which Zap style you are counting.
The same picture on Make, using the non-AI rule of one operation to one credit, depends on how many modules run per enquiry, including modules you forgot were in the scenario. If you add an AI step that uses Make’s own model, stop using the “one credit” rule and read the tag. The 200 is invented so the arithmetic is visible. Substitute your month. Then open the price page. The cost worksheet is where that usage meets an implementation fee. They are not the same number.
A polling trigger that checks every two minutes does not, on Zapier’s page, create a task per check. It can still be a bad design for other reasons: delay, duplicates, and load on the source system. “It does not cost a task” is not “it is harmless”.
Seven ways to be charged for less work
- Filter before the actions, and know whether your Zap style still charges a task for a filtered run.
- Use a native connection inside a product you already pay for when it does the job. The cost guide’s comparison of native, platform and custom is the decision.
- Digest. Five alerts in one message are fewer actions than five messages, where the product’s digest step is not itself metered. Check. Zapier lists Digest among steps that do not count as tasks. Make may differ.
- Do not poll in a loop that writes a row every time it wakes. Waking may be free and writing is not.
- Dedupe before the write, so a retry is not a second task and a second record.
- Stop syncing fields nobody reads. Each unnecessary update can be an action.
- Retire dead workflows. They still hold a connection and sometimes still run. The handover checklist is how you find them when the builder has left.
When the bill means the design is wrong
If the count, at your real volume, only works on a plan you will resent, the next conversation is not a coupon. It is whether this process should be a native feature, a smaller workflow, or someone’s job. That is the hire guide and the cost worksheet, in that order: what is being built, then what it costs to run.
Limits
Zapier’s article was updated 5 November 2026 and Make’s credits page was read for this piece. Both will change. Power Automate prices were omitted on purpose. Forum bills were omitted on purpose. Nothing here says what you should pay, or that one vendor is worth it.
When the bill is a reason to hire help, see when to hire a Zapier, Make or n8n expert and when to hire a Power Automate consultant.
To count one workflow before you pick a plan, use the workflow service or WhatsApp +65 9071 3976 (chat).
Frequently asked questions
Is Make cheaper than Zapier?
It depends on the shape of the workflow, not on a slogan. Count successful actions on Zapier and operations or credits on Make for the same volume, including filters and AI steps, then read each vendor’s current price page. This guide does not say one is cheaper.
Do failed Zapier steps use tasks?
Zapier’s help, updated 5 November 2026, says action steps that error or halt do not count, and steps that never run do not count. A successful step inside an error-handler path does count. Read the page again before you rely on it. The rule is about Zapier, not about other products.
Does testing a Zap use tasks?
Zapier says testing in the editor, including a single step, all steps, or an end-to-end test, does not use tasks, even though the action really runs in the connected app. That can still create or email something in the app. Testing is not free of side effects. It is free of Zapier tasks, on the terms of that page.
What happens when I hit the Zapier task limit?
Zapier says that once the plan’s task limit is reached you are charged for extra tasks if pay-per-task billing applies. It emails you, then again around 80 percent and 100 percent of that extra allowance, which it describes as three times the plan limit. After that, further tasks are held until the cycle ends. If the plan has no pay-per-task billing, or you turned it off, new runs are held at the plan limit instead. Upgrade, a new cycle, or turning pay-per-task on are the ways the page says held runs can proceed.
Are there free plans?
Vendors change what a free plan includes. Open the pricing page on the day you decide. Do not use a number from a forum, including this one, because this one deliberately has none.
