n8n, Zapier or a custom build? Where the crossover point actually is

Every automation platform, n8n, Zapier, Make and the rest, prices the same way: you pay per run. Every custom build prices the opposite way: you pay once, and then very little. That single structural difference decides which is cheaper for you, and it is arithmetic rather than opinion.

This post is only about that decision. If you are still working out whether to buy a ready product at all, the broader version is custom software vs off-the-shelf.

The two cost shapes

Platform pricing changes often enough that quoting a 2026 tier here would be wrong within months, so check the current page for whichever tool you use. The shape is what matters and the shape does not change: their cost is a function of your volume, ours is not.

Working out your own crossover point

One line of arithmetic:

Months to break even = build cost ÷ (current monthly platform bill − monthly running cost of the build)

If a platform is costing you $400 a month, a custom build of that same workflow is $6,000, and running it costs $100 a month, then $6,000 ÷ ($400 − $100) = 20 months.

How to read the answer:

• Under 18 months: building usually wins, and the case gets stronger if your volume is still growing, because the platform bill grows with it and the build's does not.
• 18 to 36 months: it is a judgment call, and it should be decided by the three limits in the next section rather than by the money.
• Over 36 months: stay on the platform. Anyone telling you to build at that ratio is selling a build.

The trap in this arithmetic is forgetting that the platform bill is a moving number. Run it again with the volume you expect in a year, not the volume you have today, and the answer often changes.

Three things no subscription tier fixes

Money is the usual reason to switch, but it is not the only one. These do not get better on a higher plan:

• Logic that a node graph cannot hold. Branching that depends on state from previous runs, retries with their own rules, anything where the workflow diagram stops being readable. When the canvas becomes the documentation nobody can follow, the platform has stopped helping.
• Latency and reliability you control. A platform's queue is shared and its incidents are not yours to fix. If a delay of minutes costs you money, that belongs inside your own infrastructure.
• Data that cannot leave. Personal data, payment data, or a client contract that says where processing happens. This is a hard stop rather than a preference, and it usually decides the question on its own.

Start on a platform, then decide

For most businesses the right sequence is not one or the other:

1. Prove the process is worth automating on a platform. Low cost, hours to build, and if the process turns out to be the wrong one, you learned it cheaply.
2. Let it run long enough to see the real volume and the real edge cases. This is the part that cannot be shortcut, and it is exactly what makes step three cheap.
3. Build when the bill or one of the three limits says so. By then you have a specification written by reality instead of a guess, and the build is faster and less risky because of it.

Skipping step one is the expensive mistake. Building an automation before anyone has used it means paying to discover requirements that a $30 subscription would have taught you in a fortnight.

What we build, and what we tell people not to build

We sell the custom side, so it is worth being explicit: on a decent number of calls the honest answer is that a ready tool covers it and the client should buy that instead. That is not generosity, it is that a build which fails the arithmetic above becomes an unhappy client in month six.

When a build does make sense, it usually looks like one of these:

• Sewing Lab plans production daily, prices the materials to buy, reconciles cash-on-delivery payments and answers buyers. The logic spans a CRM, two ad accounts, two postal APIs, spreadsheets and Telegram, which is well past what a node graph holds comfortably.
• AI SMM Automation turns internal team wins into brand-aligned posts and publishes them across platforms: a content pipeline where the judgment about tone is the product, not the plumbing.

The pricing behind both is in what an AI agent costs, and the wider list of what gets automated first is in business process automation.

Frequently asked questions

Is n8n cheaper than a custom automation? At low volume, almost always. Platforms charge per run or per task, so a workflow that fires a few hundred times a month is far cheaper on a subscription than as a build. The arithmetic flips when your monthly platform bill approaches the running cost of a custom build, because the build's cost is mostly one-off and does not grow with volume.

How do I calculate my own crossover point? Take the one-off build cost, divide it by the difference between your current monthly platform bill and the monthly running cost of a custom build, typically $30 to $800. The answer is the number of months to break even. Under 18 months usually justifies building; over 36 usually does not.

What can a custom automation do that a platform cannot? Three things: logic a node graph cannot express without becoming unmaintainable, latency and reliability guarantees inside your own infrastructure, and data that is not allowed to leave your systems. No subscription tier fixes any of those.

Should we start with a platform even if we will outgrow it? Usually yes. A platform proves the process is worth automating for a small monthly fee and no build cost. Once it is proven and the bill is growing, the custom build has a known specification, which makes it cheaper and less risky than building the same thing from a guess.

What does a custom automation cost at DForce? A single-task agent is $1,000 to $3,000, delivered in 3 to 7 days. A workflow across several systems is $3,000 to $10,000, typically 2 to 4 weeks. A production system across several tools is $10,000 to $30,000. Running costs are $30 to $800 a month depending on volume.

If you have a workflow on a platform and a bill that keeps climbing, book a discovery call with our automation team and we will run the arithmetic with your real numbers, including the case where the answer is to stay where you are.