n8n vs Zapier vs Make vs Custom Code — When Each Wins for B2B Agencies
n8n, Zapier, and Make all solve the same problem — connecting apps without writing code. They differ on price, scaling behavior, and who owns the result. Here is the honest breakdown, plus where custom code beats all three.
The four options, plainly
Zapier is the easiest to start with. Pick a trigger, pick an action, you have a working automation in under 30 minutes. It bills per task — every successful action step counts. A five-step workflow burns five tasks per run.
Make.com is the visual middle ground. It bills per operation (now sold as credits since November 2025), and an operation is roughly one module processing data. More setup friction than Zapier, but meaningfully cheaper per run and better at branching logic.
n8n is the developer's no-code tool. It bills per execution — one workflow run is one execution, regardless of how many steps. That single billing difference is why a 10-step n8n workflow costs roughly 10× less than the same flow on Zapier. It can be self-hosted for the price of a small server.
Custom code — what Empirra ships — has no per-run billing at all. It runs on flat infrastructure (Vercel, Supabase) you control. Higher upfront build cost, near-zero marginal cost, and the agency owns the source.
We use n8n as a logic donor — it is excellent for prototyping a flow in 30 minutes. Production runs on TypeScript because per-task fees, vendor lock-in, and workflow-JSON-as-source-of-truth all become liabilities once an automation actually matters to revenue.
Pricing comparison — what you actually pay
All figures below are 2026 annual-billing pricing, taken from each vendor's published plans. Monthly billing runs 15–40% higher across every tool.
| Plan tier | Zapier | Make.com | n8n (Cloud) | Custom code (Empirra) |
|---|---|---|---|---|
| Free tier | 100 tasks/mo | 1,000 ops/mo | Self-host: unlimited | — |
| Entry paid | $29.99/mo — 750 tasks | $10.59/mo — 10,000 ops | €24/mo — 2,500 executions | $4k–$12k one-time build |
| Mid tier | $73.50/mo — 2,000 tasks | $18.82/mo — 10,000 ops | €60/mo — 10,000 executions | $50–$200/mo infra |
| Billing unit | Per action step | Per operation (credit) | Per workflow run | None — flat infra |
| Self-host option | No | No | Yes (~$3–7/mo server) | Yes — you own it |
| Vendor lock-in | High | High | Low | None |
| Time to first automation | ~30 min | ~1–2 h | ~4–8 h | 2-week build |
The headline numbers are misleading on their own. Zapier's $29.99 looks cheap until you remember a single 5-step Zap consumes 5 tasks per run — 750 tasks is 150 runs. Make's 10,000 operations and n8n's execution-based billing both stretch far further on the same workflow.
When each tool wins
Zapier wins when you need something live today, the workflow is simple (2–4 steps), and volume is low. The per-task cost is irrelevant at 50 runs a month. For a solo founder testing an idea, it is the right call.
Make.com wins when the logic has branches, filters, and iterators, and you are willing to spend an afternoon learning the canvas. At mid volume it is the cheapest no-code option per run, and the operation-based billing is more forgiving than Zapier's.
n8n wins when you have someone technical on the team, you want to self-host to kill the subscription entirely, and your workflows are step-heavy. It is the closest no-code tool to owning your automation outright.
Custom code wins when the automation is core to how the agency makes money — lead capture, qualification, proposal generation, client reporting. At that point per-run billing becomes a tax on growth, a broken connector becomes a revenue incident, and "the automation lives in a vendor's account" becomes a real risk. Code you own on flat infrastructure removes all three.
Not sure which tier you have outgrown?
The free audit maps your current automations against volume and cost, and names the ones worth rebuilding as code.
Book Free AuditThe hidden cost: per-task billing at scale
The pricing tables every vendor publishes hide the same thing — cost scales with usage, and usage scales with success. The better your automation works, the more it costs to run.
Take a lead-capture and qualification flow that fires on every inbound form submission. Say it is an 8-step workflow: receive webhook, validate, enrich, score with AI, write to database, notify Slack, send confirmation email, update CRM. On Zapier that is 8 tasks per lead. At 1,000 leads a month that is 8,000 tasks — well past the $73.50 Professional tier, into overage packages that climb steeply.
The same flow on n8n is 1,000 executions. The same flow as custom code is zero marginal cost — it runs on infrastructure you already pay a flat $50–$200/mo for, whether it fires 100 times or 100,000 times.
This is not an argument that no-code tools are bad. It is an argument that they are priced for experimentation, not for production load. The break-even point — where a one-time custom build is cheaper than years of subscription plus overages — arrives faster than most agencies expect, usually somewhere around 5,000 monthly runs on a multi-step flow.
Why Empirra ships code, not workflows
When Empirra builds an automation, the deliverable is TypeScript running on Vercel Edge Functions, with Supabase for state and Claude API for any AI step. There are four concrete reasons this beats handing over a workflow file:
No per-run billing
Flat infrastructure cost. The automation runs at any volume without the cost curve bending upward.
Real version control
Every change is a Git commit. You can see what changed, when, and roll back — not possible with a workflow JSON blob.
Zero lock-in
The code is yours. No vendor account holds your business logic hostage, no platform pricing change can break your budget.
Built to integrate
HubSpot, Airtable, Slack, Stripe — connected via official APIs, not fragile pre-built connectors that break on vendor updates.
The tradeoff is honest: a custom build takes two weeks and costs $4k–$12k, versus 30 minutes and $30/mo for a Zap. For a workflow you are testing, the Zap wins. For a workflow your revenue depends on, the math flips — and it flips permanently, because the custom build's cost is fixed while the subscription compounds.
How to decide for your agency
Skip the feature checklists. The decision comes down to three questions:
1. Is this automation core to revenue? If a failure costs you leads or clients, it belongs in owned code, not a vendor account. If it is internal convenience, a no-code tool is fine.
2. What is the monthly run volume, multiplied by steps? Low (under ~1,000 task-equivalents): Zapier or Make. Mid with a technical teammate: n8n self-hosted. High, or growing fast: custom code, because per-run billing becomes the dominant cost.
3. Who maintains it in six months? No-code tools assume whoever built the workflow is still around to debug the canvas. Code with version control and tests survives staff turnover.
Most agencies run a mix — Zapier for the throwaway internal stuff, custom code for the lead pipeline and client-facing automations. That is the right answer. The mistake is running revenue-critical automation on per-task billing because it was the fastest thing to set up.
Audit your automation gaps
Most agencies lose revenue to manual workflows — or pay per-task fees on automations they have already outgrown. Book a 30-minute audit and get a written system plan.
Book Free AuditFAQ
Is n8n cheaper than Zapier?
For multi-step workflows, yes — substantially. n8n bills per workflow execution while Zapier bills per action step, so a 10-step flow costs roughly 10× less on n8n. Self-hosted n8n (Community Edition) is free on a ~$3–7/mo server. Zapier is only cheaper when workflows are simple and volume is very low.
What is the difference between a task, an operation, and an execution?
A Zapier task is one successful action step — a 5-step Zap uses 5 tasks per run. A Make.com operation is one module processing data, billed as credits since November 2025. An n8n execution is one entire workflow run regardless of step count. This is the single biggest cost difference between the three tools.
When should an agency move off no-code automation tools?
When the automation becomes core to revenue, when monthly run volume multiplied by steps climbs past roughly 5,000 task-equivalents, or when nobody on the team can confidently debug the workflow canvas. At that point per-task billing is a tax on growth and a custom build pays back fast.
Does Empirra use n8n at all?
As a logic donor for prototyping — n8n is excellent for sketching a flow in 30 minutes. Production automations ship as TypeScript on Vercel Edge Functions, Supabase, and Claude API, so there is no per-task billing, no vendor lock-in, and the agency owns the source code in Git.
How much does a custom-coded automation cost versus a subscription?
A custom build runs $4k–$12k one-time depending on scope, plus $50–$200/mo flat infrastructure. A no-code subscription is cheaper month one but compounds — and overage packages on high-volume workflows climb steeply. The break-even point usually arrives within the first year for any revenue-critical, multi-step automation.
Can custom code integrate with HubSpot, Airtable, or Slack?
Yes — via each platform's official API rather than pre-built connectors. Webhook-based for real-time triggers, batch sync for reporting. Custom field mapping is handled in the audit phase, so there are no vendor-locked connectors and no fragile middleware to break on a platform update.
What does a custom automation build timeline look like?
Two weeks for most engagements: audit and scope (3 days), system design (4 days), implementation and handover (1 week). The code-first stack means no platform lock-in — the agency owns and controls the system the moment it ships.
Sources
- Zapier. Plans & Pricing. zapier.com/pricing (accessed May 2026)
- Make. Pricing & Plans. make.com/en/pricing (accessed May 2026)
- n8n. n8n Plans and Pricing. n8n.io/pricing (accessed May 2026)