Zapier's per-task pricing looks innocent on the pricing page: ₹800 base + ₹2 per 100 tasks. At 100 tasks a month, you're at ₹800. At 500 tasks, ₹1,800. At 1,000 tasks, you hit ₹2,400. By 2,000 tasks, you're paying ₹4,400 monthly. It scales smoothly, which means it scales painfully. Make and n8n don't scale that way. They charge flat rates for operation volumes, and their per-dollar efficiency improves as you push more through. But that flat-rate promise breaks the moment your workflows demand error handling, multi-step approval chains, or conditional branching. We tested all three at real-world task volumes to show where the actual tradeoff lives. Zapier's task math: Where the bill accelerates Zapier defines a task narrowly: one trigger + one action. A workflow that reads a Slack message, looks up a contact in HubSpot, and sends an email is three tasks. A workflow that routes that email to different team members based on priority is four tasks. A workflow that logs to a spreadsheet, sends a Slack notification, and creates a CRM record is four tasks. At ₹2 per 100 tasks, the math feels manageable until your task volume reflects reality. We built a common scenario: lead ingestion with verification and routing. Trigger: new form submission (1 task) Look up company in external database (1 task) Check against do-not-call list (1 task) Route to correct sales rep based on territory (2 tasks, because Zapier counts conditional routing as separate operations) Send confirmation email (1 task) Log to Google Sheets (1 task) Create CRM record with metadata (1 task) That's 8 tasks per lead. At 125 leads monthly, you've crossed 1,000 tasks and hit the ₹2,400 tier. Add another common automation—invoice approval chains, contract signature routing, customer churn monitoring—and you're well past 2,000 tasks. Most teams underestimate their task volume by 40% in the first three months because they don't count conditional branches, retries, and failed automation attempts as tasks. Make's flat-rate model: When it stays cheap Make charges ₹600–₹1,200 per month (depending on feature tier) for 10,000 operations. An operation in Make is broadly equivalent to a Zapier task, but Make counts fewer of them. A conditional branch that splits a workflow into two paths is typically one operation, not two. We tested the same lead-routing workflow in Make: Trigger + webhook (1 operation) HTTP request to company database (1 operation) Array aggregator + router for multi-step logic (2 operations) Parallel branches: email + CRM + Sheets (counted as 1 per branch, 3 total) Error handling module (1 operation, optional but critical) Same workflow, 8–9 operations in Make. At 125 leads monthly (1,000–1,125 operations), you stay well within Make's 10,000-operation monthly allowance. At ₹600, your cost per 1,000 operations drops to ₹60. Zapier charges ₹200 for the same volume. The catch: Make's visual builder is denser than Zapier's. Complex workflows—especially those with error recovery, retries, and fallback paths—take longer to build and debug. But once built, they run cheaper and with more control. n8n's self-hosted escape hatch n8n operates on a different model entirely. The cloud version charges ₹600–₹1,200 per month (similar to Make's tiers), but n8n also lets you self-host for free. If you run n8n on your own server (AWS, DigitalOcean, Render), you pay only for infrastructure—typically ₹2,000–₹5,000 monthly for a small-to-medium workload, regardless of how many executions you run. For teams running 3,000–5,000 tasks monthly across 50+ automations, self-hosted n8n hits breakeven against Zapier by month two. By month twelve, the cumulative savings are substantial: ₹28,800 (Zapier annual at 3,500 tasks/month) vs ₹30,000–₹60,000 (n8n self-hosted with infrastructure). The tradeoff: you own the uptime, security patches, and database backups. You also inherit the ops debt of running a service. For most SMBs, this is a net negative. For teams with 1–2 ops engineers, it's a net positive. Where cheaper platforms stumble: The complexity test We built three real-world workflows in all three platforms to expose the friction points: Scenario 1: Multi-step approval chain Purchase request → Manager review → Finance review → Vendor payment → Accounting log. Zapier: Each branch (manager rejects vs approves, finance rejects vs approves) adds tasks. Error handling requires a separate webhook to catch rejections. At 50 approval chains monthly, you've hit 400–500 tasks just for conditional logic. Cost impact: +₹600/month. Make: Router module handles branching natively. All conditional paths count as 1–2 operations. Same 50 chains cost 150–200 operations. No added cost; stays within the 10,000-operation tier. n8n: Identical to Make's approach. Router + conditional output pins handle all branches in a single node. Cost: zero incremental. Scenario 2: Error retry and fallback An API call fails (network timeout, rate limit). The workflow needs to retry three times with expo