Why is my Zapier bill so high?
Eight quick questions about what your automations do. You get back where the billable steps are going, how many of them a platform billed per run would not charge for at all, and the steps running before a filter throws the result away.
- Where your billable steps are actually going, biggest first.
- The same work counted both ways: per step, and per whole run.
- The steps running before a filter that throws the result away.
- Whether moving platform would help you, or just move the bill.
The problem this settles
The bill goes up, nobody changed anything, and the dashboard shows a number of tasks with no clue which automation ate them. So the conversation turns into whether to move to something cheaper, which is the wrong question asked at the worst possible moment, because a rebuild on a new platform carries every structural mistake across with it and then charges less per mistake.
The cause is usually the unit, not the volume. Zapier and Make charge for every acting step, so a ten step automation running a thousand times is ten thousand of something. n8n charges once per run, so the same work is a thousand. Nothing about the automation changed; the thing being counted did. On top of that sits the one real waste: a filter far enough down the chain that you pay for most of the steps above it on runs that were always going to be discarded. Worth knowing what is not the problem, because people assume it is: on Zapier a poll costs nothing, a failed step costs nothing, and a step skipped by a filter costs nothing.
This check separates those. It ranks your automations by billable steps a month, counts the same work both ways so you can see how much of the bill exists only because steps are the unit, says what share of those steps run before a stop that throws the result away, and tells you plainly when the honest answer is that your bill is just the work and moving platform would only move it. It quotes no monthly price, because the platforms publish ranges rather than tables and the middle of a range is not a fact.
- Your automations
- Your answer
- Your plan
Question 1 of 8
How to use it
- 1Answer eight questionsQuick picks about what your automations do, roughly how often they run, how they start, where the filter sits and what happens when a step fails. One short line about the business. Two minutes.
- 2Run the prompt in your own ChatGPTThe page writes the prompt and opens ChatGPT with it already sent. A free account is fine, and Claude or Gemini work too. Nothing comes to us, and you never paste a connection or a key anywhere.
- 3Bring the answer backCopy the code block at the end of its reply and paste it back on the page. Ctrl+V anywhere works. The page reads it and scores it against the rubric published here.
- 4Read the verdict and take the reportYou get where the tasks are going, the share ending in nothing, the cost of the same work on each platform, and what to cut first. The report is on the page in full. If you want the image file to keep or forward, that is where we ask for an email.
How the score works
This score is ours, not an industry standard. It is the same seven things we check before quoting a rebuild: is anything paying per step for what is one job, does the cheap decision happen first, do the billable steps end in anything, is something running on a timer to find nothing, can a failure go somewhere, do you know which automation spends the most, and does anybody own them. The billing rules behind it were read on 7 October 2026 from each platform's own pages: Zapier bills only successful action steps and never triggers, polls, failures or steps skipped by a filter; Make bills a credit per module action; n8n bills one execution per whole run however many steps it has. We do not quote a monthly price, because Zapier publishes its Professional plan as a range rather than a table and we will not invent the middle of it. It is published here so you can disagree with it.
- 20No automation is paying per step for what is one job
- 20The decision that stops most runs happens first
- 15Most billable steps end in something actually happening
- 15Nothing runs on a timer only to find there was nothing to do
- 15A failing step has somewhere to go
- 10You know which automation is spending the most
- 5Somebody owns them
Questions
Why did my Zapier bill go up when I did not change anything?
Usually volume found you. Zapier bills successful action steps, so an automation behind a form, an order or a new lead costs a step each time it fires, and a good quarter is a bigger bill with nothing changed. The second common cause is a replay: a failed run costs nothing, but when somebody replays it, every step that succeeds the second time is charged again. Both show up as task count rather than as an error, which is why the dashboard looks fine.
How is a Zapier task actually counted?
One per action step that succeeds, per run. Zapier's own help page is explicit about what is free: triggers never use tasks, polling for new data never uses tasks, a step that errors or halts does not count, and a step skipped because of a filter or a path condition does not count either. Make counts a credit per module action, with routers and error handlers free. n8n counts one execution per whole run however many steps it has. Read 7 October 2026 from each platform's own pages. That last difference is the one that decides most bills.
Should I move from Zapier to Make or n8n to save money?
It genuinely can help, and for one specific reason: n8n bills a whole run as one execution while Zapier and Make bill every acting step, so a long chain is where the saving is real rather than imagined. But fix the shape first, because a move carries it with you. Put the filter at step one, collapse the chains that are five steps doing one job, and then compare. If the bill is still the problem after that, the bill is the work, and the choice is about who maintains it.
Is self hosting n8n actually cheaper?
The community edition is free to download, and the server is not free, and neither is somebody keeping it patched and noticing when it stops overnight. It is genuinely cheaper at high volume and genuinely more expensive at low volume once you count the hour a month somebody spends on it. This check tells you which side of that line your run count puts you on, and it counts your work in executions as well as steps so the comparison is yours rather than ours.
Do you see my answers or what ChatGPT sends back?
No. The questions, the prompt and the scoring all run in your browser. You are never asked for a connection, a key or access to an account. Nothing is sent anywhere unless you fill in the form at the end, and that form shows you the exact note before it goes.
Do I need a paid ChatGPT account?
No. A free account works, and so does Claude or Gemini. The prompt asks for a code block at the end of the reply, which all three can do.
Who is this not for?
If you run a couple of thousand steps a month you are near the bottom of every price list and the saving is not worth an afternoon. The check will say so. It is for the bill that has become a line item somebody asks about.
How long does it take, and what do I get?
Eight questions, one prompt and one paste. You get your automations ranked by billable steps, the same work counted per run as well as per step, the share running before a stop that discards the result, and a fix list in order of saving.
What if the answer is that nothing is wrong?
Then you have an answer for whoever asked about the invoice, which is worth having. The bill is the work, and the only levers left are the volume and the number of steps. The page says that plainly rather than inventing a saving to keep you reading.
Tell us what you are building.
Thirty minutes, free, with the people who would do the work.
Book a call