n8n vs Zapier vs Make: which should a service business run on?
Pick by who will look after the workflows. If nobody on your team is technical, Zapier suits short workflows and Make suits ones that branch. If someone technical looks after them, n8n charges per run instead of per step and can run on your own server. I build on n8n, so I list where it fails too.
All three tools move data between the apps you already pay for. A form comes in, a contact lands in the CRM, someone gets told. Where they differ is what you pay for as volume grows, who can fix a workflow when it breaks, and whether a person can check something before it goes out to a client.
I build on n8n, so read its column with that in mind. Every price and plan limit below comes from the vendor's own page, listed under Sources with the date I read it.
| Question | n8n | Zapier | Make |
|---|---|---|---|
| What you pay for | Workflow executions. One execution is one full run of a workflow, however many steps it has. Cloud plans start at 20€/mo, billed annually. The self-hosted Community edition has no plan fee. | Tasks. A task is used each time a Zap completes an action, and failed actions don't count. The Free plan includes 100 tasks per month. Professional starts at $19.99/month, paid yearly. | Credits. Each module action counts as one credit. The Free plan gives up to 1,000 credits/mo, and paid plans are priced from 10k credits/mo. |
| How cost grows with volume | With the number of runs. Adding steps to a workflow doesn't add executions. Pro is 50€/mo, billed annually, for 10K workflow executions. | With runs times steps. A long Zap on every new lead uses a task for each action it completes. | With runs times steps times items. If a trigger returns ten new rows, every later step runs ten times. |
| Who can change a workflow | Someone comfortable reading the data each step returns. An owner can follow a workflow, but fixing one usually takes a technical person. | Most staff. Each step is a short form, which makes it the easiest of the three to hand to a non-technical team. | Staff who will learn its visual map. Simple scenarios are easy to follow, and branches and lists take practice. |
| Where your data lives | On n8n's cloud, or on a server you run yourself. | On Zapier's cloud. | On Make's cloud. |
| What happens when a step fails | The run stops and is logged as failed. An error workflow, set per workflow, can send the alert to email or chat. | Autoreplay, on Professional, Team and Enterprise plans, can retry an errored step up to five times, and the error email goes out after the last attempt fails. | Without an error handler the run stops, and storing failed runs for a retry is off by default. Repeated failures turn off the schedule. |
| A person approving before something sends | Built in. A Slack or email message waits for approve or decline, then the workflow carries on. | Built in on Professional, Team and Enterprise plans. Human in the Loop pauses a Zap until a reviewer approves or edits it, and each reviewer needs a Zapier account with the Zap shared to them. | Its Human in the Loop app is on the Enterprise plan and in closed beta. On other plans it has to be built by hand. |
| Moving off it later | Self-hosted, the workflows and their run history sit on your server. Moving to another tool still means rebuilding each one. | Each Zap gets rebuilt in the new tool and tested again. | Blueprints export as files, but they only import back into Make. Moving means rebuilding. |
Where each one fails
n8n
n8n fails when nobody technical is around. A field gets renamed in your CRM, a run fails, and someone has to read the error and fix the mapping. If that person is a freelancer you can't reach, the workflow stays broken until you find someone else.
Self-hosting adds upkeep on top. n8n's own docs say self-hosting needs technical expertise and that maintenance is your responsibility: updates, backups, security patches, and the server going down at a bad moment. The free licence covers running it for your own internal business. It doesn't cover offering n8n to your customers to build in.
Zapier
Zapier fails on cost once workflows get long and busy. Every completed action is a task, so a Zap with many steps that runs on every lead burns through a plan quickly. When you hit the limit, Zaps pause until the next period unless pay-per-task billing is on, and that charges a higher rate per task. Logic with several branches also gets awkward to build and to read.
Make
Make fails on the multiplier. A trigger that returns many items runs every later step once per item, so a busy week can use far more credits than a quiet one. When credits run out, scenarios stop until more are added. Its error handlers are good, but someone has to add them to each scenario, and an approval step is only on the Enterprise plan today.
What I'd pick, per situation
- If nobody on your team is technical and each workflow is a trigger and a couple of actions, pick Zapier.
- If nobody is technical but your workflows branch, loop through lists or touch many apps per run, pick Make, and add an error handler to every step that writes to a client record.
- If someone technical looks after the workflows, in-house or hired, and they run often with many steps, pick n8n Cloud.
- If someone technical looks after the workflows but they're short and don't run often, keep them on whichever tool that person already knows.
- If client data has to stay on a server you control, pick self-hosted n8n, and name the person who updates and backs it up before you start.
- If a person has to approve every outbound email or invoice before it sends, pick n8n or Zapier, where that step is built in on the plans a small firm buys. On Zapier, each reviewer needs a Zapier account with the Zap shared to them.
- If Zapier already runs your business and the bill doesn't bother you, stay. Moving means rebuilding every workflow and testing it again.
Questions
Is n8n free?
The self-hosted Community edition has no plan fee, and its licence covers running it for your own internal business. You still pay for the server and for someone's time to update it, back it up and keep it secure. n8n Cloud starts at 20€/mo, billed annually.
Which one is cheapest for a small agency?
It depends on how many steps each workflow has and how often it runs. Zapier counts a task for each completed action, Make counts a credit for each module action on each item, and n8n counts one execution per run. Take your busiest workflow, count its steps and its runs in a month, then price it on all three pricing pages.
Do I need a developer to use n8n?
To keep it running well, you need someone comfortable reading the data a step returns and fixing a broken field. That can be a hired builder. To self-host it, you also need someone to run the server, which n8n's docs list as your responsibility.
Can I move my Zaps to n8n or Make later?
Not by import. Each tool keeps workflows in its own format, so moving means rebuilding each one and testing it against real data. List every Zap before you start, including the ones nobody remembers making.
What happens when a workflow fails overnight?
All three log the failure. Zapier's autoreplay, on Professional, Team and Enterprise plans, can retry an errored step up to five times and emails you after the last attempt fails. Make stops the run unless an error handler is set. n8n runs an error workflow, if you've set one, and that can alert you by email or chat.
- n8n pricing: Starter 20€/mo, billed annually. Pro 50€/mo, billed annually, for 10K workflow executions. An execution is a single run of your entire workflowREAD
- n8n docs, Choose how to use n8n: self-hosting requires technical expertise, and maintenance is your responsibilityREAD
- n8n docs, Community license: the Sustainable Use License allows use for your own internal business purposesREAD
- n8n docs, Handle errors gracefully: an error workflow runs when an execution failsREAD
- n8n docs, Slack approvals: Send and Wait for Response lets approvers approve or decline before the workflow resumesREAD
- Zapier pricing: Free plan with two-step Zaps and 100 tasks per month. Professional starting from $19.99/month, paid yearly. Tasks are used when a Zap completes an action, and failed actions don't count. With pay-per-task billing off, Zaps pause at the plan limit until the next period; with it on, extra tasks bill at a higher rate than plan tasksREAD
- Zapier help, What is replay?: autoreplay retries an errored step up to five times, on Professional, Team and Enterprise plansREAD
- Zapier help, Human in the Loop: Request Approval pauses a Zap until a reviewer approves or edits it, on Professional, Team and Enterprise plans, and each reviewer needs a Zapier account with the Zap shared to themREAD
- Make pricing: Free plan up to 1,000 credits/mo, paid plans priced from 10k credits/mo, each module action counts as one credit, and scenarios stop when credits run outREAD
- Make help, Operations: each item a trigger returns runs the later modules once moreREAD
- Make help, Errors and warnings: an unhandled error stops the run, and repeated failures turn off the scheduleREAD
- Make help, Incomplete executions: storing failed runs is off by defaultREAD
- Make apps, Human in the Loop: Enterprise plan only, in closed beta for invited customersREAD
- Make help, Scenario blueprints: export a scenario as a file to import into MakeREAD
Related
- Should I hire an operations coordinator or install a system?What an operations coordinator costs a year, set against installing a system: where each one fails, and which to pick first when your agency is growing.
- How do I stop website form duplicates in HubSpot?Where the handoff from your website form to HubSpot makes duplicates or loses leads, which field is the key, what stays with a person, and what I'd pick.
- How do I approve email replies from Slack before they go out?Approving email replies from Slack, Telegram or Google Chat: what the built-in apps do, where an approval loop breaks, the send-once rule, and what I'd pick.
- WebhookA webhook is a message one app sends another the moment something happens. What that means in plain English, how it fails, and how a system catches it.
- Questions to ask anyone building your automationsQuestions to ask an automation builder before you hire them, with what a careful answer sounds like and the answers that should worry you.
- Automations that fail without telling anyoneAn automation can stop for weeks before anyone notices. The usual causes, what not knowing costs, and what an alert should say when a step fails.