A closer look at three workflow automation platforms: pricing, self-hosting, integration count, vendor lock-in and five concrete use-case recommendations.
Platform selection process
Workflow inventory, volume estimate, data-sensitivity mapping. One to two weeks in discovery.
Two to three mid-complexity workflows on each platform. Free tier or one paid month is enough.
Task, operation and execution costs at real volume on a three-year horizon, including infrastructure.
Two-week parallel run in shadow mode, then a gradual cutover — typically 20-40 hours per 50 workflows.
Workflow automation in 2026 runs mostly on three platforms: Zapier (the pioneer, with the largest integration catalog), Make (the cleanest visual builder, successor to Make.com / Integromat), and n8n (self-hostable, the developers' favorite). All three are worth knowing — but the choice is project-level, not company-level. At an SMB it's entirely normal for the marketing team to run on Zapier while the tech team runs on n8n in parallel.
This article is a deep dive into all three platforms — their strengths and weaknesses — plus five concrete use cases with a recommendation for each. The numbers and benchmarks reflect Q1 2026; the market moves fast, so it's worth verifying against the official pricing pages.
The market has transformed completely since 2020. Three main trends:
The strategic position of the three platforms:
Zapier launched in 2011 and remains the market-leading SaaS integration platform. It has 6,000+ integrations, most of them fully solved (auth, webhook, error handling). The no-code interface is genuinely no-code: a business user can assemble a "new lead lands in HubSpot → Slack message → Google Sheet row" workflow in ten minutes.
Zapier works with Zaps: one Zap = one trigger plus one or more actions. Pricing is task-based: every action run counts as one task. If your Zap is "Stripe → Slack + Google Sheet + HubSpot", one trigger consumes 3 tasks.
Trigger: Stripe new charge (0 task, free)
Action 1: Send Slack message (1 task)
Action 2: Add row to Google Sheet (1 task)
Action 3: Update HubSpot contact (1 task)
Total: 3 tasks / event
SMB-level traffic (50 customers / day × 3 tasks) works out to 4,500 tasks a month — that fits inside the $50 Starter tier. At 10x that volume you already need the Professional tier ($73-103/month).
| Tier | Tasks/month | Monthly fee | Use case |
|---|---|---|---|
| Free | 100 | $0 | Single user, hobby |
| Starter | 750 | $19.99 | Small business, 2-3 workflows |
| Professional | 2,000-50,000 | $73-289 | Standard SMB |
| Team | 50,000+ | $103-799 | 3+ users, sharing |
| Company | Unlimited | $1,000+ | Enterprise SLA |
Note: Zapier's task counting isn't always intuitive. A filter step can consume a task, and so can a lookup. For a real budget estimate, run a free week and check actual usage on the Tasks panel.
Make (formerly Integromat) has a more modular architecture. Its "operation"-based counting is more granular than Zapier's (one webhook trigger = 1 op, one data lookup = 1 op), and complex branching workflows are easier to manage in the visual scenario editor.
Make works with scenarios. A scenario is a freely connectable module graph — not necessarily linear. Multiple triggers, multiple paths, aggregators and iterators are all handled natively.
Webhook trigger (1 op)
↓
Router (0 op)
├─ Path A: Slack message (1 op)
├─ Path B: HTTP GET vendor API (1 op) → JSON parse (1 op) → DB insert (1 op)
└─ Path C: Filter (0 op) → Email send (1 op)
Pricing is operation-based, and every module execution is 1 op. It's generally cheaper than Zapier at a comparable volume.
| Tier | Ops/month | Monthly fee |
|---|---|---|
| Free | 1,000 | $0 |
| Core | 10,000 | $9 |
| Pro | 10,000 | $16 |
| Teams | 10,000 | $29 |
| Enterprise | 1M+ | Custom |
Cheaper than Zapier at comparable volume. A 20,000 ops/month workload (~200 scenario executions/day, 3 ops average) runs about $30-50/month. The equivalent on Zapier is about $80-100.
n8n (pronounced "n eight n", short for "nodemation") runs self-hosted or on n8n.cloud. It's available under the Sustainable Use License — open-source code, but with restrictions on competing CRM-like commercial use. For SMBs and end users it's 100% free to use.
n8n workflows can be exported as JSON and committed to git. This changes workflow management dramatically: code review, branch-based deployment and environment promotion all work.
# n8n workflow (simplified)
nodes:
- name: Webhook
type: n8n-nodes-base.webhook
- name: Validate
type: n8n-nodes-base.code
parameters:
jsCode: |
if (!$input.item.json.email) throw new Error("Missing email");
return $input.item;
- name: OpenAI Classify
type: n8n-nodes-base.openAi
- name: HubSpot Create
type: n8n-nodes-base.hubspot
connections:
Webhook: [Validate]
Validate: [OpenAI Classify]
OpenAI Classify: [HubSpot Create]
Self-hosted pricing is unlimited per execution. Only VM cost matters.
| Option | Cost |
|---|---|
| Self-hosted Community | $0 license + ~$10-50/month VPS |
| Self-hosted Enterprise | $0-X license (custom) + infrastructure |
| n8n.cloud Starter | $20/month (5 active workflows, 5k executions) |
| n8n.cloud Pro | $50/month (50 workflows, 50k executions) |
| n8n.cloud Business | $667/month (unlimited, advanced features) |
docker run -d --restart unless-stopped \
--name n8n \
-p 5678:5678 \
-v n8n_data:/home/node/.n8n \
-e N8N_BASIC_AUTH_ACTIVE=true \
-e N8N_BASIC_AUTH_USER=admin \
-e N8N_BASIC_AUTH_PASSWORD=$(openssl rand -base64 32) \
-e WEBHOOK_URL=https://n8n.example.com/ \
docker.n8n.io/n8nio/n8n
Caddy or Nginx as a reverse proxy for HTTPS, a Hetzner backup snapshot for disaster recovery. The whole stack runs under $10/month.
| Criterion | Zapier | Make | n8n |
|---|---|---|---|
| Integration count | 6,000+ | 1,500+ | 400+ (growing fast) |
| Self-hosting | No | No | Yes |
| Vendor lock-in | High | Medium | Low |
| No-code capability | Excellent | Good | Medium |
| Code extensibility | Limited (Code by Zapier) | Limited (Custom Functions) | Excellent (Code node) |
| Pricing model | Task-based | Operation-based | Per-execution / unlimited self-hosted |
| Entry monthly fee | $20 | $9 | $0 (self-hosted) / $20 (cloud) |
| Cost at 100K executions/month | $1,000+ | $200-400 | $10-50 (self-hosted VPS) |
| Complex workflows | Hard | Easy | Easy |
| Onboarding | Simple | Medium | Steep |
| GDPR / EU data | Enterprise tier | EU-host option | Self-hosted, full control |
| Version control (git) | No | No | Yes (JSON export) |
| AI / LLM support | Good | Good | Excellent (local LLM via Ollama) |
| Vector DB integration | Limited | Limited | Yes (Pinecone, Qdrant, pgvector) |
An SMB-level workflow (200 events/day × 5 steps = 1,000 executions/day, ~30K/month):
| Platform | Tier | Monthly cost | Vendor cost trend |
|---|---|---|---|
| Zapier | Professional 50K tasks | $103 | +12% a year |
| Make | Pro 25K ops | $16-29 | Stable |
| n8n cloud | Starter | $20 | Stable |
| n8n self-hosted | Hetzner CX22 | $5-10 | Stable |
A larger volume (100K executions/month, 500K-1M ops/month):
| Platform | Tier | Monthly cost |
|---|---|---|
| Zapier | Team 100K tasks | $389 |
| Make | Teams 1M ops | $99-149 |
| n8n cloud | Pro | $50 |
| n8n self-hosted | Hetzner CX32 + Postgres | $20-40 |
The difference is dramatic. At 100K tasks/month, Zapier costs 8-20x as much as self-hosted n8n.
Setup: CRM (HubSpot) → Slack notification → Google Sheet → email send. Standard integrations, no custom logic. The marketing team wants to manage it self-service.
Volume: 50-200 events/day, 3-5 steps/event.
Recommendation: Zapier. The no-code interface and the 6,000+ integrations win here. Members of the marketing team can build a new workflow in 30 minutes. The $50-100 monthly cost isn't a concern.
Alternative: Make's Core tier, if Zapier's cost is too high. But onboarding the marketing team takes 2-3x as long.
Setup: Stripe webhook → conditional routing → CRM update → analytics event → Slack alert. Heavy branching, custom logic in places. The tech team manages it; marketing only has visibility at a monitoring level.
Volume: 5,000-50,000 events/day, 5-10 steps/event, heavy branching.
Recommendation: Make. Operation-based pricing and the visual router are cheaper than Zapier here. The scenario architecture scales well, and the Make Pro tier ($16) is plenty to start.
Alternative: Self-hosted n8n, if the tech team wants git-flow for its workflows. More setup effort, but cheaper long-term.
Setup: Internal CRM → custom API → public webhook → audit log. Moving customer data (PII) that can't pass through a SaaS platform. The compliance team explicitly requires data residency.
Volume: 100-1,000 events/day, an audit log entry for every event.
Recommendation: Self-hosted n8n. Data residency is a hard requirement here. The SaaS alternatives offer an EU region on their enterprise tiers, but even then a US company operates it. With self-hosted n8n, the data never leaves your own server.
Alternative: Make's Enterprise tier with the EU-host option, if the compliance team accepts it. Zapier isn't a fit here either.
Setup: Inbound email → LLM classification → CRM update → Slack. The AI part uses a custom prompt that keeps changing. Prompt tuning happens weekly, sometimes daily.
Volume: 100-2,000 events/day.
Recommendation: n8n or Make. Both have a strong LLM module. n8n's advantage: the prompt can be exported to git, and complex pre/post-processing can be handled in the code node. On Make, onboarding the marketing team is easier.
Alternative: AI by Zapier — it works, but prompt iteration is slower in the Zapier UI.
Setup: Google Analytics → data aggregation → email with a formatted table. One-off, low traffic. Roughly one new workflow a month.
Volume: 1-3 events/day.
Recommendation: Zapier. The free 750 tasks/month tier covers this need. Setup takes 30 minutes. If it outgrows that, $20/month won't hurt.
Alternative: Make's Free tier — also fine, slightly more complex setup.
Migrating between automation platforms isn't trivial, but it isn't impossible either. The practices below come from our experience running 3-5 migration projects.
The most common migration. Reason: cost reduction and/or GDPR.
Typical migration time: 20-40 person-hours per 50 workflows.
Less common, but similar logic. Make's scenario structure is closer to n8n's than Zapier's linear flow is.
Rare. Reason: when maintaining self-hosting turns out to cost more than a SaaS tier.
We steer clients toward self-hosted n8n when:
Otherwise: Zapier for the marketing team, Make for the tech team's workflows. For an SMB at low volume, Zapier is the safest choice — setup cost is minimal, and a pilot starting on the Free tier costs nothing.
Tip: If you're just starting out, don't pick a "company-wide platform". Start with 2-3 platforms, build one or two workflows on each, and decide based on actual usage. "Everything on Zapier" or "everything on n8n" are premature decisions.
Related articles of ours: building a RAG chatbot walks through n8n's LLM orchestration examples, and AI implementation for Hungarian SMBs covers the ROI math behind workflow automation.
There is no "best" platform — it depends on the project and the team. For SMBs at low volume, Zapier is the safest choice. For a strong technical team, self-hosted n8n scales better. For medium-complexity workflows, Make is the compromise.
Platform selection is a one- to two-week scope in the discovery phase: workflow inventory, volume estimate, data-sensitivity mapping, team skill audit. The actual build then takes four to eight weeks, and it's iterative from there.
If you're planning an AI or automation project, let's talk it through during discovery to find which platform fits your actual use case. Platform selection takes a week, implementation four to eight weeks, and ROI is measurable within 30-60 days.
About the author
Corevanix Kft.
Technology partner
Budapest-based technology partner — SAP/ERP integration, web development, AI automation and mobile app development. We work inside the client’s own environment, and the delivered code belongs entirely to the client.
Legal basis, a DPA with the AI provider, EU data residency, pseudonymisation, retention and training opt-out, the balancing test, the AI Act and a checklist.
Prompts are code: repo, versioning, review, template structure, few-shot examples, eval sets, regression tests, injection defence, cost and observability.
OCR + LLM pipeline, JSON-schema extraction, validation with human-in-the-loop, SAP/ERP integration, error-rate tracking and ROI for invoices and contracts.