The Same Shopify Workflow Is 1,800 Zapier Tasks, 4,680 Make Operations and 1,000 n8n Executions

Somebody asks what it will cost to automate a piece of their Shopify store, and the answer arrives as three monthly prices sitting next to each other. That comparison is broken before it starts, because the three vendors are not selling the same unit. Zapier bills a task. Make bills an operation. n8n bills an execution. Those three words describe three different things, and the gap between them is not a rounding error — on one ordinary workflow it is more than four times.

Here is that workflow, then the four rules that produce the spread, then the parts of Shopify Flow that decide whether you need an external layer at all.

A format note: the canonical version of this article carries an interactive calculator that converts your own workflow shape into all four vendors' units, plus a routing quiz. Neither survives a plain markdown feed, so the arithmetic and the routing logic are written out as text below.

One workflow, three counts

Take a shape most stores would recognise. Something happens 1,000 times a month. A filter throws away 40% of those as irrelevant, so 600 survive. Each of the 600 runs three action steps. The tool checks Shopify every 15 minutes, which over a 30-day month is 2,880 checks.

Notice what did the damage, because it is not what people expect. The filter is free everywhere, so it is not the culprit. The three action steps are free on n8n and charged three times over on the other two. And the 2,880 checks — the same checks, on the same schedule — are worth nothing on Zapier and n8n and 2,880 operations on Make.

Then change exactly one thing. Poll every 5 minutes instead of every 15, and the checks go from 2,880 to 8,640 a month. Zapier still bills 1,800 tasks. n8n still bills 1,000 executions. Make goes to 10,440 operations, past the only allowance its published price covers. Same work, same result for the store, and one of the three vendors has just charged you for being impatient.

The four rules that produce it

Each of those is the vendor's own wording. Zapier's is the plainest: "A task is anything Zapier successfully completes on your behalf." Make's diverges at exactly one point, and it is the expensive one — its help centre states that a trigger module costs one operation per check, regardless of the number of bundles returned, which includes checks that return nothing at all. Make charges you for looking. Zapier and n8n do not, and n8n writes the opposite rule down explicitly:

Polling nodes (like Google Drive Trigger): Count one execution only when new data is found. Polls that return no results don't count as an execution.

— n8n, "Understand executions"

Now put prices on those counts — carefully

1,800 tasks lands on the 2,000-task step of Zapier's Professional ladder: $49 a month billed annually, $73.50 billed monthly. 1,000 executions sits inside n8n Cloud Starter at $20 a month billed annually or $24 month to month, which covers 2,500 executions. 4,680 operations sits under Make's $12-a-month Core price. All three verified on each vendor's own pricing page, 28 July – 1 August 2026.

Two of those three are rungs on a ladder. One is a floor. Zapier publishes an 18-step task ladder in both billing cycles, from 750 tasks up to two million. n8n publishes two tiers in two cycles. Make publishes a single usage slider that moves quota and price together, and the $12, $21 and $38 figures are all prices at the lowest position of that slider, labelled "10,000 credits/mo". Above that position Make publishes nothing, and no annual dollar figure appears for any tier. So "4,680 operations for $12" is an entry price sitting next to a count, not a quote — and any comparison chart that plots Make against rising volume is drawing a line somebody made up.

Two more things worth carrying out of the price section:

• Zapier's billing-cycle gap is 50%, not the 15–20% most SaaS charges — on every step up to 100,000 tasks. Any Zapier figure quoted without naming the cycle is potentially half the real number.
• The unit counts are not a ranking either. n8n produces the smallest number and is not the cheapest line; Make produces the largest and has the lowest published entry. The unit does not tell you who is cheaper today. It tells you which way your bill moves when your workflow changes shape.

Before any of that: where Shopify Flow stops

Flow is the layer people skip, and it is free on Basic, Grow, Advanced and Plus — Shopify publishes that list twice, on the Help Center and on its Flow marketing page, and Starter appears in neither. It reacts to events in the store and its connected apps, and it has a clock trigger. If a workflow starts and ends inside Shopify, no meter ever starts.

Two plan gates sit inside the app your plan already includes:

• Send HTTP Request — the one built-in action that talks to anything outside Shopify — is available only on the Plus, Advanced or Grow plans. On Basic it is not offered. At US list on monthly billing that gate costs $66 a month to clear by upgrading Basic to Grow, and it buys everything else Grow carries, not just the action.
• Tasks built by custom partner apps need Plus. That one affects a single capability, and it is the reason a bespoke build sometimes wants Plus underneath it.

Four more limits are not plan gates at all — no tier above yours removes them:

• Flow sends no general customer email. The only general email action is Send internal email, and Shopify's guidance on the action's own page is blunt: "the Send internal email action is best used to send emails to staff. To automate emails to customers, create a marketing automation." Four named actions do reach customers — B2B access, draft order invoice, order invoice, payment reminder — and that is the documented list.
• No built-in trigger starts Flow from outside Shopify. It starts on store and app events, or on the clock. An installed app can register its own triggers; the plan cannot.
• Send HTTP Request has no published payload or body-size limit. We fetched the full action page and found none, which means the limit is untested rather than generous. What the page does document: Flow waits up to 30 seconds for a response code, then closes the connection and retries, and 4XX, 5XX and 429 responses can be set to Retry for up to 24 hours, Fail, or Ignore.
• There is no published run allowance. This is the most-asked question about Flow with no published answer. Shopify instead says usage limits follow the Admin API rate limits of your plan — "Stores are given different usage limits in Shopify Flow according to the API limits governed by their plan." Those buckets are 100 points/second Standard, 200 Advanced, 1,000 Plus, 2,000 Shopify for enterprise. That is a ceiling on how hard the store can be worked, not on how many workflows you may own — and it is the closest thing to a planning figure that exists.

One naming collision does real damage here: marketing platforms call their email and SMS sequences flows, and a Klaviyo flow has nothing to do with Shopify Flow. If the automation you are picturing sends a series of emails to a customer, you are shopping on the wrong layer entirely.

Reach is two questions, not one

How much of Shopify a tool can touch, and how it finds out something happened, are separate — and the second one decides both your bill and your latency.

On catalogue, counted on each vendor's own reference pages (Shopify and Make in July 2026, Zapier on 4 August 2026): Flow, 83 triggers and 109 actions before anything an installed app adds. Zapier, 59 Shopify items — 20 triggers plus 39 actions. Make, 27 modules — 4 triggers, 18 actions, 5 search. n8n, two built-in nodes — a Shopify app node with 10 operations across orders and products, plus a separate Shopify Trigger node whose docs page links usage templates rather than enumerating the events it supports, so confirm your event before you plan around it.

On mechanism, there are exactly two ways for anything outside Shopify to learn that an order arrived, and Shopify is unusually direct about the limits of the first:

Webhook delivery isn't always guaranteed, and your app can miss or mishandle events for other reasons, such as handler failures or downtime.

— Shopify, "Webhooks"

Shopify's own advice to app builders is to run reconciliation jobs that periodically re-fetch, and that applies just as squarely to a Zap or a scenario. Any workflow that moves money, stock or fulfilment needs a second, slower workflow that checks the first one's work — a nightly sweep re-reading yesterday's orders. That sweep is a second workflow with its own unit count on whichever layer runs it, and it is the one most people forget to model at all.

Polling is the alternative, and it always lags by its own interval. The only lever that shortens the lag is polling more often — which costs operations on Make and nothing on Zapier and n8n. Speed is a paid upgrade on exactly one of the three. There are floors, too: Make's own comparison table puts the minimum scenario interval at 15 minutes on the free plan and 1 minute on Core, Pro and Teams; Shopify Flow's clock is documented down to a smallest repeat interval of 10 minutes.

Is self-hosted n8n really free?

The licence is. The running of it is not, and no vendor publishes that number — which is why the honest answer is a list of components rather than a total.

• The software: nothing. A standard self-hosted version is on GitHub.
• The server: your number, not ours. n8n's docs name an instance class rather than a price — a Hetzner CPX11 (2 shared vCPU, 2 GB RAM, 40 GB SSD) described as enough for most usage levels. Its monthly price is not published anywhere a tool can read it; we tried five independent methods on 28 July 2026, including Hetzner's own public cloud pricing endpoint and the Wayback Machine, and the server-rendered HTML ships a placeholder where the number belongs.
• Upkeep: hours × your rate. Updates, backups you have actually restored from once, and monitoring that tells you before your merchants do.
• The 2 a.m. question. Who is on the hook when a workflow stops silently during a sale. On a hosted plan that is the vendor; here it is a person you employ or are.

And one licence detail that matters if you build for clients rather than for yourself. Plenty of write-ups call n8n open source. n8n does not:

Although n8n's source code is available under the Sustainable Use License, according to the Open Source Initiative (OSI), open source licenses can't include limitations on use, so we do not call ourselves open source.

— n8n, "Sustainable Use License"

The licence permits use or modification for your own internal business purposes, or non-commercial and personal use, and permits distribution only free of charge for non-commercial purposes. Automating your own store is squarely internal business use. Hosting it for clients and reselling it is the example n8n itself names as not allowed. The word for this model is fair-code — source-available with restricted commercial use.

The exit is the price nobody quotes

On the import and export pages we read — each vendor's own, 28 July 2026 — no vendor documents a path that moves workflows to either of the others. Each reads only its own format. n8n describes saving workflows as JSON and importing JSON files into your n8n library, and nothing on that page or its siblings mentions another platform's format. The picture is the same in every direction.

So a tool at Make's $12 slider floor that takes three weeks to leave is not cheaper than Zapier's $49 annual-billing step that takes three weeks to leave. The subscription difference is rounding next to the rebuild. Two practical consequences: keep a written description of what each workflow does, independent of the tool that runs it, so a rebuild is a translation rather than an archaeology project. And weight the first choice toward the unit that fits your workflow shape, because that is the thing you cannot renegotiate later without paying for it twice.

Five things nobody publishes

Every gap below is a verified absence rather than an assumption — we went looking, and this is what was not there. Listing them is the honest alternative to filling them in:

1. Any Make price above the lowest position of its usage slider, and any annual dollar figure per tier (checked 28 July and again 1 August 2026 by capturing the pricing page's raw HTML in full).
2. The monthly price of the Hetzner CPX11 instance n8n's docs recommend (five methods, 28 July 2026).
3. A run allowance for Shopify Flow (the Flow hub plus its getting-started, create, manage and reference pages, the Send HTTP Request page, and shopify.com/flow — 28 July 2026).
4. A payload or body-size limit for Send HTTP Request.
5. A documented migration path between Zapier, Make and n8n.

The bottom line

The decision is smaller than it looks once the two halves are separated. Put everything that stays inside Shopify on Flow, where it costs nothing beyond the plan you already pay for — on most stores that is the majority of the list. Then take what is left, work out its real unit count on each layer, and buy the one whose unit is kindest to your shape. A workflow with many action steps and few runs punishes Zapier and Make. A workflow that polls frequently punishes Make specifically. A workflow with enormous volume and someone already running servers argues for n8n on your own hardware, provided you count the hours honestly.

The sorting is what saves the money. The vendor choice only decides how much.

---

Originally published at shopify.ecom-store.pro, where the same comparison carries an interactive unit-cost calculator. All prices and counts were verified against each vendor's own pages between 28 July and 4 August 2026; check them before you commit, because they move. Our content is AI-generated and fact-checked against official Shopify and vendor documentation.