n8n vs Zapier vs Make: pick one tool, then stop shopping

If you run a one-person business, the automation question is not “which platform is winning the internet.” It is “which one will still be running on a Tuesday when a client pays late and you are already in a call.” Zapier, Make, and n8n can all move data between the same apps. They fail in different ways, they bill you differently, and they ask a different amount of you when something breaks.

This is not a feature matrix. Feature matrices go stale in a quarter. It is a decision rule you can use this week: start with the job, then pick the cheapest reliable way to do that job, then leave it alone.

What you are actually buying

Zapier sells time. You pay per task, you get a huge catalog of maintained connectors, and when Google changes an API, Zapier’s team usually notices before you do. The editor is simple. The mental model is linear: trigger, then steps. If your work is “when a form lands, add a row, send an email,” Zapier will be live in an afternoon. You will also pay for every row and every email, forever, including the retries.

Make sells visual control. Scenarios can branch, loop, and aggregate in ways Zapier still makes clumsy. The price per operation is often lower than Zapier’s task price, which is why agencies like it. The tradeoff is that the canvas becomes a drawing. Six months later you will open a scenario you built in a hurry and spend twenty minutes remembering why a router exists. Make is at its best when one person owns the scenario and writes a one-paragraph note at the top of it.

n8n sells ownership. You can run it in their cloud or on a machine you control. Workflows are nodes and JSON. Self-hosting means the meter is hardware and your time, not tasks. The catalog is large, and you can write a function node when a connector is thin. The cost shows up as maintenance: updates, credentials, webhooks, disk, and the 2 a.m. question of whether a Docker volume still exists.

A pricing picture that does not require a spreadsheet

Count the runs, not the tools. A typical solopreneur stack is not thousands of Zaps. It is a handful of jobs that fire often: new leads, invoices, calendar events, email labels, a weekly digest. Zapier’s free and starter tiers run out faster than people expect because a “simple” Zap is rarely one task. A trigger plus three steps is four tasks per run. Two hundred form fills a month is already eight hundred tasks before you add the invoice sequence.

Make’s operations add up the same way, just with a different unit name. Filters and some modules still consume operations. If you are looping over a list of unpaid invoices, you are paying per invoice, per module. That is fine at fifty invoices. It is a surprise at five hundred.

Self-hosted n8n has no per-run fee. That sentence is the whole sales pitch, and it is true until you count the hours. A $6–12 VPS plus backups is cheap. The first time a Let’s Encrypt cert fails, or a webhook URL changes after a reverse-proxy tweak, you will spend an evening you could have billed. n8n Cloud exists for a reason: you keep the workflow model without becoming your own SRE.

Where each one quietly wins

Use Zapier when the connector quality matters more than the unit price. Accounting tools, CRMs, and niche SaaS apps are where Zapier’s catalog earns its keep. If you only have three automations and they must not die, paying Zapier is often cheaper than debugging n8n on a Sunday.

Use Make when the job is a branching scenario you can see. Multi-path routers, aggregators, and “wait for all of these” flows are clearer on a canvas than in a long Zap. Keep scenarios small. One job per scenario. Name the modules. Put the live version behind a status that you actually look at.

Use n8n when you will have more than a few workflows, when you want to version them as files, or when a workflow has to touch something Zapier will not: a local database, a custom HTTP API, a queue, a file on disk. n8n is also the right call if you already run other containers and a missed webhook is a problem you know how to find in logs.

The switching cost nobody puts on the pricing page

Moving a live automation is not an export button. Credentials have to be re-entered. Webhook URLs change, which means every form and Stripe event has to be repointed. Test data is different. Error paths you never wrote down will reappear. Budget a full workday to move one important workflow, not an hour.

That is why the useful advice is boring: pick one platform for production. You can prototype in another. You should not run invoices in Zapier, onboarding in Make, and backups in n8n unless you like three dashboards and three ways to miss a failure. If you are already in Zapier and it is under $30 a month, stay. If you are already self-hosting n8n and it has been quiet for two months, stay. Tool tourism is a hobby, not a business activity.

A decision rule you can use today

Write down the three jobs that, if they stop, cost you money this month. For most people that is lead capture, invoice follow-up, and something that files documents. Time how long each job takes by hand. If the three jobs together are under two hours a week, a paid Zapier plan is a reasonable purchase and you can ignore the rest of this article for a year.

If those jobs are growing, or you already have a server, or you want the workflows as files you can back up, start with n8n. Run it on n8n Cloud until you have a reason to self-host. Self-host only after you have a backup, a way to get error emails, and a calendar reminder to apply updates. Make sits in the middle: good if you think in diagrams and you will not outgrow the operation meter immediately.

Whatever you pick, write the workflow’s job in one sentence at the top. Name the owner (you). Note the last time it ran successfully. That note is worth more than any comparison chart.