The short answer
If you're not sure which tool fits your team, start with an automation tool. Move to custom software only when you hit a real wall: logins, roles, approvals, or a system of record that needs to outlive any single workflow.
That's the whole framework. The rest of this article explains why, and how to tell which situation you're in.
What each option actually is
Zapier is a broad app connector. It has the largest library of pre-built integrations and the simplest setup. Good for straightforward, linear workflows: "when X happens in app A, do Y in app B." It's usually the easiest place to start if your team has no technical background.
Make (formerly Integromat) is a visual scenario builder. It shows your workflow as a flowchart with branches, routers and loops. It's a step up in complexity from Zapier, but it gives you more control over conditional logic and error handling without writing code.
n8n is developer-friendly workflow automation. It supports the same kind of app-to-app connections as Zapier and Make, but it also lets you drop in code nodes (JavaScript, Python) for custom logic. n8n can run in the cloud or be self-hosted on your own infrastructure, which matters for teams with data residency or cost-control concerns.
Custom software is not an automation tool. It's your own application: a database, user logins, role-based permissions, and a purpose-built interface. You'd build this when the "workflow" is really a piece of your business that needs a home, not just a connection between two other tools.
"The question isn't which tool is more powerful. It's whether you need a workflow or a system of record."
Decision table
| Factor | Zapier / Make | n8n | Custom software |
|---|---|---|---|
| Volume of runs | Fine at low-to-moderate volume; cost scales with usage | Fine at any volume; self-hosted removes per-run fees | Volume-independent once built; infrastructure cost instead |
| Logic complexity | Simple to moderate branching | Moderate to complex, including custom code | Unlimited — you write the logic |
| Need for logins/roles/UI | Not designed for this | Not designed for this | This is the point of custom software |
| Data ownership/residency | Data passes through vendor's cloud | Can self-host to keep data in-house | Fully yours, wherever you host it |
| Cost behavior at scale | Rises with tasks/operations processed | Flat subscription or infrastructure cost if self-hosted | Upfront build cost, then hosting and maintenance |
| Who maintains it | Vendor maintains the platform | You or a partner maintain workflows and, if self-hosted, the server | You or a partner maintain the app long-term |
| Vendor lock-in | Moderate — workflows tied to platform | Lower if self-hosted; workflows are portable | None — you own the code |
| Audit trail needs | Limited, vendor-dependent logging | Better logging, especially self-hosted | Full control over audit logs and history |
Cost behavior: per-run fees vs. infrastructure
Zapier and Make both use consumption-based pricing — you pay based on the number of tasks or operations your workflows run each month. That's simple to start with, but it means your bill grows with your usage. Check each vendor's current pricing page before committing, since these plans and limits change.
n8n takes a different approach. n8n Cloud starts around $20/month on the Starter plan and $50/month on Pro, and n8n can also be self-hosted, which shifts cost away from per-run fees and toward infrastructure and maintenance instead (n8n Pricing).
Custom software follows this same logic, one step further. You pay to build it, then you pay for hosting and upkeep — not for every workflow run. As an illustrative anchor, a hosted database like Supabase Pro runs $25/month (Supabase Pricing). That's a small piece of what a real infrastructure bill looks like, but it shows the shift: instead of a fee per automation run, you have a predictable line item for the systems that store and serve your data.
| Cost model | How it scales | Typical fit |
|---|---|---|
| Zapier / Make (consumption) | Rises with tasks/operations | Low-to-moderate volume, simple workflows |
| n8n Cloud | Flat subscription tiers ($20/mo Starter, $50/mo Pro) | Teams wanting predictable cost without self-hosting |
| n8n self-hosted | Infrastructure + maintenance, no per-run fee | Higher volume, data residency needs, technical team |
| Custom software | Build cost + hosting/maintenance (e.g., Supabase Pro at $25/mo) | Long-term system of record, UI, roles, approvals |
Note: pricing for all vendors changes. Confirm current numbers directly on each provider's site before budgeting.
Signals you've outgrown an automation tool vs. signals you should stay
| Outgrown an automation tool | Stay on an automation tool |
|---|---|
| You need customers or staff to log in and see different things based on role | Everything happens between apps you already use |
| You're building spreadsheets or side databases to track state the tool can't hold | Data lives fine inside your existing apps |
| You need approval steps with permanent history and accountability | Your workflow is a straight chain of triggers and actions |
| Multiple workflows are duct-taped together to fake a real application | One or two workflows solve the problem completely |
| You're hitting per-task cost ceilings for a process that's core to the business | Task/operation costs are a minor, predictable line item |
| You need a real audit trail for compliance or client-facing reporting | Vendor-provided run history is good enough |
The hybrid pattern
Most teams don't have to choose one or the other forever. A pattern that works well: use an automation tool as the glue between apps, and a custom app as the system of record.
For example, n8n might watch for new leads across your ad platforms, enrich them, and route them into your CRM. But if you need a client portal where customers log in, see their project status, and approve deliverables, that's not a job for an automation tool — that's a custom application with its own database and logins. The automation tool keeps talking to the custom app through its API, feeding data in and pulling data out.
This keeps your automation layer simple and your system of record solid, instead of stretching a connector tool past what it was built for.
Where we fit in
We build with n8n, Make, Zapier and custom software, depending on what the job actually needs — not what's easiest for us to sell. When we build custom software, you own the code. No lock-in to us.
If you're not sure which side of this line you're on, an Ops Teardown ($490) is a fast way to find out: we map your current workflows and tell you plainly whether an automation tool or a custom build is the better fit.
From there:
- Automation Sprint — from $3,500, 2-4 weeks, for teams ready to build or fix automation workflows.
- Custom Software Build — from $9,000, 4-10 weeks, when you need a real application with logins, roles and a UI. You own the code.
- Run & Improve — from $1,500/mo, for ongoing maintenance once something is live.
- Fractional Automation Lead — from $3,000/mo, if you want ongoing strategic oversight without a full-time hire.
See full details on pricing, or get in touch to talk through your specific situation.