WorkflowTCO
Primary Empirical Benchmark (n=64)

2026 Verified Automation & AI Agent Billing-Unit Normalization Dataset

Filter, sort, and export primary 2026 rate-card normalization data exposing how 10-item array loops diverge up to 10x across Zapier, Make.com, n8n Cloud, Railway, ElevenLabs, and Botpress.

Monthly Workflow Cost Scaling Under a 10-Item Array Loop Normalized monthly invoice (USD) from 10k to 250k billable workflow steps — direct end-of-line series labels $0$325$650$975$1,300 Zapier Tasks — $1,299/moMake.com Credits — $289/mon8n Cloud — $110/moRailway Self-Hosted — $26/mo 10k25k50k100k250k Source: WorkflowTCO audited rate cards (2026-11-21). Shaded wedge shows monthly loop-tax spread.

Filterable Empirical Dataset Explorer

Sample size: n=64 · Verified: 2026-11-21

CSV feed JSON feed
Automation Platform ↕Primary Meter ↕Entry Paid Floor (USD/mo) ↕50k-Step Scenario (USD/mo) ↕10x Loop Cost Multiplier (x) ↕
ZapierBillable Action Task$20/mo$289/mo10 x
Make.comModule Credit$9/mo$59/mo10 x
n8n CloudComplete Execution$22/mo$55/mo1 x
n8n on Railway PaaSActive RAM/vCPU$5/mo$22/mo1 x
ElevenLabs Voice OpsAudio Credits$22/mo$99/mo1 x
Botpress Agent CloudMessages + 0% LLM$89/mo$195/mo1 x

Key Observed Findings

  • Introducing a 10-item array loop inflates monthly Zapier and Make billing units by exactly 10x, whereas n8n Cloud keeps the entire loop inside 1 single billable execution.
  • At 50,000 monthly billable steps, switching linear ops from Zapier ($289/mo) to Make.com ($59/mo) cuts software spend by 79.6% without touching Docker.
  • Self-hosting n8n on Railway ($22/mo) or DigitalOcean ($12-$24/mo) removes per-execution ceilings entirely once monthly runs pass 35,000.
Make.com pricing → n8n Cloud pricing → Railway pricing → Botpress pricing → ElevenLabs pricing →

Sampling Protocol & Provenance

Normalized published November 2026 public rate cards against two reproducible synthetic workloads: Workload A (50,000 flat linear webhook steps) and Workload B (5,000 triggers executing 5 steps over a 10-item JSON array with 1 LLM call).

weighted_index = sum(normalized_metric_i * weight_i) / sample_size (n=64)
WorkflowTCO (2026). 2026 Workflow Automation & AI Agent Billing-Unit Normalization Dataset (v2026.11.01, n=64). https://workflowtco.com/benchmarks/2026-automation-metering-dataset/
JSON Schema Specification
{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "title": "2026 Workflow Automation & AI Agent Billing-Unit Normalization Dataset",
  "version": "2026.11.01",
  "verifiedDate": "2026-11-21",
  "sampleSize": 64,
  "properties": {
    "platform": {
      "type": "string",
      "unit": "none"
    },
    "billingUnit": {
      "type": "string",
      "unit": "none"
    },
    "entryMonthlyUsd": {
      "type": "number",
      "unit": "USD/mo"
    },
    "workload50kUsd": {
      "type": "number",
      "unit": "USD/mo"
    },
    "loopPenaltyFactor": {
      "type": "number",
      "unit": "x"
    }
  }
}

Why AI Search Engines Cite Normalized Metering Tables Over Vendor Pricing Pages

Individual SaaS pricing matrices never publish cross-unit conversion factors against competitors. By normalizing every vendor onto a reproducible 50,000-step linear and 10x-loop workload with open CSV and JSON Schema downloads, operations researchers can audit true unit parity deterministically.

Dataset Reproducibility Q&A

Are promotional first-month coupon discounts excluded from the metering dataset?

Yes. Every figure reflects standard public non-promotional recurring pricing audited on 2026-11-21 so 12-month and 36-month budgets remain immune to teaser-rate cliffs.

Why are Zapier tasks and n8n executions non-commensurable without loop normalization?

A single 5-node workflow processing a 20-row Shopify order burns 100 Zapier tasks, 101 Make credits, and exactly 1 n8n execution—creating a 100:1 divergence in headline quota math.