Make.com vs n8n: What Actually Breaks First

I run n8n in production and I've shipped client work on Make. Here's what actually fails, what actually costs money, and how I'd pick between them today.

The Workflow That Worked Fine Until It Didn't

You build a workflow on a Friday afternoon. A new lead comes in, it gets tagged, a Slack message fires, a row lands in your CRM. You watch it run twice, it works both times, you close the laptop feeling like a genius.

Three weeks later a field is empty in your CRM and you have no idea why. Somewhere between the form and the database, something changed shape and nobody noticed.

That is the actual question behind "make com vs n8n." Not which one is prettier to build in. Which one you can trust when nobody is watching, and which one tells you the truth when it breaks.

I run n8n on my own VPS for the automations behind CallCrewHQ and Grease Trap Quotes, and I built plenty on Make for client work before that. Both are good tools. They are not the same tool, and the difference shows up exactly when you need it to, at 2am, when a workflow quietly stops.

What Each One Actually Is, Fast

Make (formerly Integromat) is a visual, node-based automation builder that runs entirely in the cloud. You drag modules onto a canvas, connect them, and Make's servers execute the whole thing. There is nothing to host, patch, or back up.

n8n is also visual and node-based, but it is built to be self-hosted, usually in Docker, on a VPS or your own server. n8n Cloud exists too if you do not want to run infrastructure, but the entire pitch of n8n is that you can own the box it runs on.

That one architectural difference explains almost every other difference between them.

Where the Data Actually Lives

On Make, every execution, every payload, every credential passes through Make's cloud. Fine for most businesses. It matters more if you are moving anything sensitive, like patient records or financial data, where your compliance requirements or a client contract specify exactly where data can and cannot touch third-party servers.

On a self-hosted n8n instance, the data never leaves infrastructure you control. Nobody outside your business sees the payloads unless you send them somewhere on purpose. That is a real answer to a real question some clients ask, not a technicality.

Pricing: The Part Everyone Gets Wrong by Only Reading the Homepage

Make prices by operations, meaning every module execution inside a scenario counts. A workflow with eight steps that runs 500 times a day chews through your monthly operation quota fast, and Make's published tiers scale up quickly once you are running anything with real volume, per Make's own pricing page.

n8n Cloud prices by workflow executions, not individual operations. That is a materially different unit, and one reason a workflow with a lot of internal steps can be cheaper on n8n's hosted plan for the same task. n8n's pricing page lays out the execution-based tiers directly.

But the number that matters most for a lot of builders is the one both comparison guides skip: what self-hosting costs when you strip away the marketing. A VPS with enough RAM to run n8n comfortably, alongside its Postgres database, runs somewhere in the $10 to $40 a month range depending on the provider and how much you are pushing through it. That is it. No per-execution ceiling, no operations meter ticking up as your business grows.

The catch, and it is a real one: self-hosting n8n is cheap and reliable, but somebody has to own the updates, the backups, and the secrets. That is the actual trade-off against a managed tool like Make, and it is worth being honest with yourself about whether that somebody is you.

The Cost Nobody Puts in a Pricing Table

Here is a decision rule most comparisons skip entirely: do not price the tool, price the failure. If a workflow breaks and nobody notices for two days, what did that cost you? For a lead-routing workflow, the answer is usually "a lead that went cold and never converted." That number is almost always bigger than the gap between Make's and n8n's monthly fee, which is why the platform choice matters less than whether you have built in a way to notice when something breaks, on either tool.

Where Make Wins, Honestly

Make's canvas is easier to read at a glance, especially for branching logic with a lot of conditional paths. The visual data mapper, where you click a field from a previous step instead of writing an expression, is friendlier for someone who has never automated anything before. If you want zero infrastructure to think about and you are comfortable with usage-based pricing, Make is a completely reasonable first tool.

Make also ships with a large native app library, so common SaaS integrations, think Google Sheets, Shopify, Stripe, tend to have well-maintained modules with less fiddling than building an HTTP request by hand.

Where n8n Wins, Honestly

n8n's node library is deep, and its support for calling arbitrary code, AI models, and custom HTTP endpoints is more flexible than Make's, particularly for the kind of AI agent workflows that have exploded since 2023. If you are building something that calls the Anthropic API as part of a workflow, which is exactly the stack I run for AI-driven products, n8n's native AI nodes and community integrations tend to move faster than the equivalent on Make.

Self-hosting also means no vendor can change your pricing tier overnight, throttle your executions, or sunset a feature you depend on. You are running the software, full stop.

A Worked Example: The Lead Follow-Up Workflow

Say you are deciding which platform to build your lead follow-up automation on. A form fill triggers a text message within five minutes. On Make, that workflow probably runs as four or five modules: webhook trigger, a filter, a CRM lookup, an SMS send. At even 300 leads a month running through that chain, you are in the mid-tier operations plan.

On n8n, the same workflow is one execution per lead regardless of how many internal nodes it touches. Run it on a $20-a-month VPS and you could scale to several thousand leads a month before the infrastructure cost moves at all. The trade is that you are now the one who restarts the container if it crashes and applies the security patch when one ships.

That is the actual decision, not "which is more powerful." Do you want a predictable line item, or a lower ceiling on cost in exchange for owning the box?

Where the Comparisons Stop and Reality Starts

Most small businesses that ask me about automation have the same three problems: leads that do not get followed up fast enough, data typed twice into two different systems, and calls that go unanswered. Neither Make nor n8n solves any of those by existing. They are both just the pipe. The workflow design is the actual work, and it is identical thinking whether you build it in Make's canvas or n8n's.

The first automation that pays for itself is almost never a chatbot. It is lead follow-up within minutes instead of hours, on whichever tool you are already comfortable with. Pick the platform based on who is maintaining it and what it costs at your real volume, then spend your actual effort on the thing that drives revenue: making sure nothing falls through the cracks.

If you want a live example of that principle instead of a hypothetical, Grease Trap Quotes exists because commercial kitchens were waiting days for grease-trap cleaning quotes. The automation behind it delivers three quotes by text in under a minute. Nobody cares which workflow tool runs underneath. They care that it is fast.

So Which One Should You Actually Pick

If you never want to manage a server, and you are fine with usage-based pricing scaling alongside your growth, use Make. It is mature, well-documented, and the visual builder is approachable for a first automation.

If you are comfortable with basic server maintenance, or willing to learn it, and you want AI-heavy or highly custom workflows, or you need to guarantee data stays on infrastructure you control, self-hosted n8n is the better long-term bet. Budget real time for updates and backups, because that responsibility does not disappear just because the software is free to run.

Either way, build in a way to notice when a workflow quietly stops. That is the actual difference between an automation that pays for itself and one that costs you a lead you never see.

If you would rather skip the platform debate and just get the follow-up, routing, or intake problem solved, get in touch and I will tell you straight whether Make, n8n, or something else entirely is the right fit for what you are trying to fix.

---

*Written by Mario Lucas. Mario runs Utomat, an automation studio that builds AI voice agents and workflow automation for small businesses.*

Related reading: The Automated Tool Nobody Buys First (And Should).

Make.com vs n8n: What Breaks First When You Actually Run Them