Make vs n8n vs Zapier at Scale: TCO at 10k, 100k and 1M Executions

Why list prices mislead at scale

Each platform meters a different unit, so comparing headline prices is meaningless:

Platform	Billing unit	What counts
Zapier	Task	Each successful action step. Triggers, filters and paths don’t count.
Make	Credit (formerly operation)	Each module that runs, including the trigger. Some AI features consume extra credits.
n8n	Execution	One complete workflow run, regardless of node count.

The formula that matters: monthly billable units = runs per month × billable units per run. Everything else — tiers, discounts, overage — is applied to that number.

The reference workflow

To make the comparison concrete, we modeled a typical B2B workflow: webhook trigger → enrich lead → AI classification → CRM upsert → owner lookup → Slack alert → spreadsheet log → status update. That is one trigger plus seven actions, with no loops.

Billable units per run	Zapier	Make	n8n
Reference workflow (1 trigger + 7 actions)	7 tasks	8 credits	1 execution

Monthly billable units

Runs per month	Zapier tasks	Make credits	n8n executions
10,000	70,000	80,000	10,000
100,000	700,000	800,000	100,000
1,000,000	7,000,000	8,000,000	1,000,000

These numbers are exact for the reference workflow. The cost estimates below are not — read the assumptions.

TCO model: assumptions

• Platform fees are illustrative monthly ranges derived from each vendor’s published tier pricing at our last check (see the table at the end), billed annually. Above the largest published tiers, vendors quote Enterprise contracts; we show plausible ranges, not quotes.
• n8n self-hosted infrastructure: a single VM and Postgres at 10k runs; queue mode with Redis, two to four workers and managed Postgres at 100k; horizontally scaled workers and highly available Postgres at 1M.
• Operations labor for self-hosting at a loaded rate of $100 per hour: about 2, 6 and 16 hours a month for upgrades, monitoring, backups and incident response.
• Build and maintenance effort for the workflows themselves is excluded because it is broadly similar across platforms — slightly lower on Zapier for simple flows, slightly lower on n8n and Make for complex ones.

Total monthly cost of ownership

Runs per month	Zapier	Make	n8n Cloud	n8n self-hosted (infra + ops)
10,000	~$500–$900	~$60–$110	~$60 (Pro tier)	~$230–$260
100,000	~$2,000–$4,000	~$400–$700	Business / Enterprise quote	~$750–$900
1,000,000	$10,000+ (Enterprise quote)	~$2,500–$5,000 (Enterprise quote)	Enterprise quote	~$2,200–$2,800

At 10,000 runs a month

Make and n8n Cloud are the clear cost leaders. Self-hosting n8n is a false economy here: the server is cheap, but a few hours of engineering time each month costs more than the cloud subscription. Zapier’s premium is real but small in absolute terms — for a business team that values speed and the app catalog, it can still be the right call.

At 100,000 runs a month

The gap widens. Make (opens in a new tab, affiliate link) is typically the cheapest on platform fees, often by several times versus Zapier. Self-hosted n8n becomes competitive if you already run Kubernetes or a platform team, because the incremental operations effort is low. If you’d be standing up infrastructure just for n8n, Make usually wins.

At 1,000,000 runs a month

At this volume every vendor is negotiating Enterprise pricing, but the unit math still dominates. n8n (opens in a new tab, affiliate link) self-hosted bills nothing per execution, so its cost is driven by infrastructure and people — and it typically lands at or below Make Enterprise. Zapier at seven million tasks a month is a material line item that needs a strong governance or adoption argument to justify.

The hidden multipliers that break budgets

The reference workflow is deliberately clean. Real workflows rarely are:

• Loops and iterators. Processing 50 invoice lines in a loop multiplies per-step billing. An eight-step workflow with a five-step loop over 50 items is roughly 250+ Make credits or Zapier tasks per run — still one n8n execution.
• Polling triggers. In Make, a scheduled polling trigger consumes a credit every time it checks, even when there is nothing new. Checking every minute is about 43,000 credits a month per scenario before any work is done. Use instant webhooks wherever possible.
• Longer workflows. A 20-step workflow roughly triples Zapier and Make units versus our reference; n8n’s cost doesn’t change.
• Retries and errors. Failed steps that are retried consume units again on Make and Zapier.
• AI steps. Make’s built-in AI features can consume additional credits, and Zapier’s AI steps count as tasks. On every platform, bringing your own LLM API key moves model costs to your provider bill — budget them separately.

Sensitivity check: if your average workflow has loops or more than ten steps, move one column to the right in the TCO table above — n8n’s relative advantage grows quickly.

Comparison at a glance

Make vs n8n vs Zapier compared

Tool	Best for	Starting price	Free plan	Deployment	AI capabilities	Score	Link
Make Top pick	Ops teams that need complex, multi-branch logic at low cost	$9/mo		cloud	AI Agents, OpenAI/Anthropic modules, AI-assisted scenario building	4.6	Visit Make (affiliate link, opens in new tab)
n8n	Engineering-led teams with data-residency or volume needs	$24/mo		cloud, self-hosted	LangChain-based AI Agent nodes, vector stores, tool calling, MCP	4.6	Visit n8n (affiliate link, opens in new tab)
Zapier	Business teams that want breadth and simplicity over control	$19.99/mo		cloud	Zapier Agents, AI by Zapier steps, Copilot, Chatbots	4.5	Visit Zapier (affiliate link, opens in new tab)

Beyond cost: what else should drive the decision?

Factor	Zapier	Make	n8n
Who builds	Business users	Ops / RevOps specialists	Engineers and technical ops
App catalog	Largest	Large	Smaller native; HTTP node covers gaps
Complex logic and data mapping	Good	Excellent	Excellent, plus code nodes
Data residency / self-hosting	No	No	Yes
Version control and environments	Limited	Limited	Git-based source control on paid tiers
AI agent building	Zapier Agents	Make AI Agents	AI Agent nodes, tool calling, vector stores

A decision rule you can apply this week

1. Count units per run for your three highest-volume workflows on each platform’s billing unit.
2. Multiply by projected runs 12 months from now — not today.
3. Price it on each vendor’s current pricing page; request Enterprise quotes above the largest published tier.
4. Add operations labor honestly if you’d self-host.
5. Then weigh the non-cost factors in the table above.

As a rule of thumb: business users and modest volume → Zapier; specialists and growing volume → Make; engineers, compliance or very high volume → n8n. Our AI workflow automation tools guide covers capabilities in depth, and the ROI estimator helps quantify the labor you’ll free up.

The bottom line

Automation costs are a function of units per run × runs, and the three platforms count units very differently. Make (opens in a new tab, affiliate link) is the cost leader for most teams up to six-figure monthly volumes; self-hosted n8n wins at seven figures for teams that can operate it; and Zapier is a deliberate premium for breadth and ease of use. Re-run this model every time a workflow’s volume grows tenfold.