I've spent the past 3+ years building automation systems for small and medium businesses, primarily in Israel. My stack has evolved over that time, but two platforms have remained constant: n8n (self-hosted) and Make.com (formerly Integromat).
This isn't a feature-by-feature comparison you can find on any SaaS review site. This is what I've learned from deploying both in production across 50+ projects.
The TL;DR
| Factor | n8n (Self-Hosted) | Make.com |
|---|---|---|
| Monthly cost | $10-20 (server only) | $9-99+ |
| Setup effort | 2-4 hours (Docker) | 5 minutes |
| Learning curve | Moderate-steep | Moderate |
| Custom code | Full JS/Python in any node | Limited |
| Data privacy | Your server, your data | Cloud (EU) |
| Integrations | 400+ (growing fast) | 1,500+ |
| Best for | Technical teams, agencies | Non-technical teams |
| API handling | Excellent (HTTP nodes) | Good |
| Error handling | Granular per-node | Good but less flexible |
| Version control | JSON export, git-friendly | No native git |
Pricing: Where the Gap Gets Real
Let's start with money, because for small businesses, this matters more than any feature comparison.
Make.com pricing (2026):
- Free: 1,000 operations/month, 2 active scenarios
- Core: $9/month — 10,000 operations
- Pro: $16/month — 10,000 operations + advanced features
- Teams: $29/month — 10,000 operations + team features
Sounds reasonable until you realize that a single moderately complex workflow can burn through 1,000 operations in a day. A WhatsApp bot handling 50 conversations daily uses roughly 5,000-15,000 operations/month depending on complexity.
n8n self-hosted pricing:
- Software: $0 (open source, fair-code license)
- Server: $10-20/month (Hetzner CX22 handles it beautifully)
- Unlimited workflows, unlimited executions
- Total: $10-20/month forever
n8n Cloud pricing:
- Starter: $20/month — 2,500 executions
- Pro: $50/month — 10,000 executions
For a single project, the difference might be negligible. But when you're running 10-20 active workflows (which is typical for a well-automated business), n8n self-hosted saves $50-200/month compared to Make.com at equivalent scale.
Over a year, that's $600-2,400 saved. For a small business, that's meaningful.
The Real Comparison: Building a WhatsApp Bot
Let me compare both platforms on a real project I've built on each: a WhatsApp appointment scheduling bot for a medical clinic.
The Requirements
- Receive WhatsApp messages via webhook
- Parse customer intent (book, reschedule, cancel)
- Check Google Calendar availability
- Create/update/delete calendar events
- Send confirmation and reminder messages
- Handle errors gracefully
- Log everything for debugging
Building it in Make.com
The good:
- Visual builder is clean and intuitive
- Google Calendar module works out of the box
- The Router module makes branching logic visual
- Error handlers are easy to set up
The frustrating:
- WhatsApp webhook parsing required a JSON module + multiple filters
- Date/time manipulation in Israel timezone (Asia/Jerusalem) needed 3 separate steps
- When I needed to calculate available time slots (30-minute intervals minus existing bookings), Make.com's built-in functions couldn't handle it cleanly. I ended up with a chain of 8 modules doing what would be 15 lines of JavaScript.
- No way to version control or diff changes. When something broke in production, figuring out what changed was painful.
- Running tests meant executing the scenario with real API calls every time.
Time to build: ~6 hours
Monthly cost: $16/month (Pro plan for the custom functions)
Building it in n8n
The good:
- HTTP Trigger node — done, webhook ready
- Code node for parsing: 20 lines of JavaScript, handles all edge cases
- Google Calendar node works great, plus I can fall back to the API directly
- Available slots calculation: one Code node, 30 lines of JavaScript, handles timezone, breaks, existing bookings — everything
- Workflows export as JSON — version controlled in git
- Test with manual triggers without hitting real APIs
The challenging:
- Initial setup: Docker, reverse proxy (Caddy/Traefik), PostgreSQL, environment variables — took about 3 hours for first-time setup
- Some n8n updates require manual migration steps (rare, but it happens)
- The UI, while improved a lot, is still less polished than Make.com
- Debugging complex workflows with many branches requires clicking into each node
Time to build: ~4 hours (faster because of code nodes)
Monthly cost: $0 (already had the server running)
The Verdict on This Project
n8n won here because the project required custom date logic that's natural in code but painful in visual builders. If the bot had been simpler (FAQ only, no calendar integration), Make.com would have been fine and faster to set up.
When I Choose Make.com
Despite my preference for n8n, I actively use and recommend Make.com in specific situations:
1. Client wants to manage it themselves
If the client's team is non-technical and wants to modify automations after handoff, Make.com's visual interface is more approachable.
2. Quick integrations between SaaS tools
Need to sync HubSpot contacts to Mailchimp and create a Slack notification? Make.com does this in 5 minutes.
3. The client doesn't want to manage infrastructure
Some clients explicitly don't want a server to maintain, even if I offer to set it up.
When I Choose n8n
1. WhatsApp bots and conversational systems
Conversation flows require complex branching, state management, and text processing. Code nodes make this manageable.
2. High-volume automations
When a workflow runs hundreds or thousands of times daily, n8n self-hosted wins purely on economics.
3. Clients with data privacy requirements
Israeli businesses handling medical or legal data need to know where their data flows. Self-hosted n8n on an Israeli or EU server keeps everything under control.
4. Multi-client agency work
I manage automations for multiple clients from a single n8n instance. Total infrastructure cost: $15/month for all clients combined.
5. Advanced AI integrations
n8n's AI Agent node and custom code support make it significantly more flexible for building AI-powered workflows.
Production Tips for n8n Self-Hosting
Use Queue Mode — Don't run n8n in the default "main" mode for production. Queue mode with Redis separates webhook processing from execution, making it much more reliable under load.
PostgreSQL, not SQLite — The default SQLite works for testing but doesn't handle concurrent writes well. Switch to PostgreSQL before going to production.
Back up your encryption key — n8n encrypts credentials at rest. Lose the encryption key, and you lose access to all stored credentials. Back it up separately.
Docker Compose is your friend — Pin your n8n version, use a proper compose file with restart policies and health checks. Don't just docker run.
Reverse proxy with SSL — Caddy or Traefik in front of n8n. Free SSL from Let's Encrypt. Non-negotiable for webhooks.
My Honest Recommendation
For most small businesses making their first foray into automation: start with Make.com. The free tier lets you experiment, the learning curve is gentler, and you'll be automating real work within an hour.
Once you outgrow it — either in volume, complexity, or cost — migrate to n8n. The transition isn't painless (you're rebuilding, not importing), but the long-term benefits are significant.
If you're technical, or you're an agency building for clients: start with n8n. The setup cost is a few hours; the savings are ongoing.
Resources
- Make vs Zapier vs n8n — Full comparison
- n8n Complete Guide (Hebrew)
- n8n Beginner's Guide — No Code Required
- How to Build a WhatsApp Bot — Technical Guide
- Business Automation Complete Guide
I'm Achiya Cohen, an automation specialist based in Israel building WhatsApp bots and business automation systems. 50+ projects, 5.0★ on Google. More at achiya-automation.com.
Top comments (0)