Before you compare the monthly prices of Zapier, n8n and Make, sketch one real workflow and count what each platform would bill. A form submission is one event to your sales team. To an automation platform it might mean one workflow execution, four successful actions, or several module runs per recipient. The number of boxes on the workflow canvas does not tell you which plan will fit.
Take a team that routes website enquiries to a CRM and sends an acknowledgement. The trial looks cheap because it uses one clean enquiry at a time. Then the team adds a duplicate check, two owners for some accounts and an alert on failure. Those changes affect each platform's meter differently. More importantly, they affect who needs to notice and fix a failed handoff.
What do Zapier, n8n and Make actually count?
| Platform | What to count | What can change the estimate |
|---|---|---|
| Zapier tasks | Successful billable actions, rather than every step drawn in the Zap | Conditional actions, multiple items and actions already completed before an error. Check special usage rules for the features you choose. |
| n8n workflow executions | Executions of the workflow; adding steps to a single execution does not itself add to that count | Workflows started separately for each event or on a schedule. Measure retries and related workflows on the plan you are evaluating. |
| Make credits | Credits consumed by modules; one operation generally uses one credit for non-AI apps | Modules can run once per bundle. Some features, including AI, use different rates or depend on consumption. |
Make's operations guide explains the effect of bundles. If one run reads multiple rows, downstream modules can run for each row. A workflow that checks for new requests every few minutes also behaves differently from one that runs only when a request arrives. Include the checks that find nothing when you measure usage; look at the actual module history rather than assuming an empty result is free.
Likewise, Zapier can count a successful CRM write even when the following notification fails. That matters to the budget and to the customer: restarting from the beginning can create a second record if the write has no duplicate protection. In n8n, look at how many production executions your design starts, rather than treating the number of items processed as the number of executions. An integration with a single scheduled batch is a different design from an integration that starts a workflow for every event.
What does a month of enquiry routing look like?
Here is a fictional example, with no assumed vendor prices. A form sends events to an automation. For each eligible enquiry, it looks up the CRM contact, writes the record, sends an acknowledgement and alerts the owner. An ineligible enquiry exits before the lookup. Some eligible enquiries require a second owner alert.
Assume 2,000 events arrive in a normal month, 1,500 qualify and 300 of those need the second alert. Assume all the actions described below succeed and there are no retries. For the busy-month estimate, use the same proportions with five times the traffic. These figures describe the example only.
| Meter in the example | Normal month | Busy month |
|---|---|---|
| Incoming events | 2,000 | 10,000 |
| Eligible enquiries | 1,500 | 7,500 |
| Zapier tasks, assuming four counted actions per eligible event and one per second alert | 6,300 | 31,500 |
| n8n executions, assuming each incoming event starts one metered workflow | 2,000 | 10,000 |
| Make credits, assuming one counted trigger operation per event and one credit per operation for the same downstream actions | 8,300 | 41,500 |
The arithmetic is deliberately visible. The normal Zapier total is 1,500 times four plus 300. The Make illustration adds 2,000 trigger operations to that total. Here the n8n runs include the events that exit early because the filter is inside the workflow. If filtering happens upstream, those events would not start it. The table is a starting estimate. The apps, modules, branches and trigger you pick decide what will appear on the bill.
When you run a trial, include a rejected enquiry, a second owner and an outage of the destination CRM. Check which operations the platform recorded before the outage and what a retry does. Send test notifications to controlled destinations and use non-production credentials. A successful replay should finish the handoff without a second acknowledgement or a second sales record. Record the cost of that case rather than assuming it looks like a successful first pass.
How should you compare the prices?
Open the vendor's pricing page with your estimated monthly allowance in hand. Use the same billing term across options, and distinguish a monthly price equivalent from what you actually pay upfront. Check which plan has the connector, admin roles, execution history and support you need. A smaller allowance with a missing connector is not a cheaper version of the same system.
Then add the costs that sit outside the automation bill. A WhatsApp provider, paid CRM account or AI service can charge separately. If you pay in rupees for a plan listed in another currency, use the amount payable for your term and note any taxes or currency conversion costs. Use the live plan selectors rather than a screenshot from someone else's account. The broader marketing automation pricing guide covers budgets for setup and support; the workflow estimate here helps you price one specific job inside that budget.
Ask what happens when you reach the allowance. Does the workflow pause, stay in a queue or keep going with extra charges? Which user can see that state and make a decision? Find the alert settings in the actual account because plan rules can differ. A temporary pause on a weekly report might be fine. Losing the only copy of a customer's enquiry while a quota is exhausted is a different problem. Keep the enquiry in a system you control before letting downstream automations act on it. That gives you something to reconcile after the service recovers.
Test a reporting boundary as well as a busy day. A marketing team may call a spreadsheet import “one job”, while the production configuration starts a separate run for every row. Agree whether the estimate covers the whole workspace or just the new workflow. Existing automations may draw from the same allowance. Give each flow a name the operations team can find in the usage report and record which team owns it. When a bill increases, you want to know whether more customers arrived, an existing workflow changed, or a retry kept running after a destination was unavailable.
Does self-hosting n8n remove the cost?
It changes who pays for the work. If you run your own instance, budget for hosting, updates, data recovery, credentials and the time to investigate failed runs. The n8n Sustainable Use License guidance distinguishes consulting and workflow building from hosting n8n and charging people to access it. If the offer involves selling access or embedding it in a customer-facing product, check the exact arrangement with the vendor before assuming the Community edition fits.
One practical way to compare is to list what happens after a connector stops working. With a hosted tool, someone still has to check authentication, review partial runs and recover missed records. With your own server, that same person may also have to restore the service and its configuration. Test whether the person responsible has the access and notes they need. A lower software bill does not settle that question.
When is custom code worth considering?
Stay with a workflow tool if it already does the job, someone on the team can maintain it and failed runs are visible and recoverable. A simple routing flow has little to gain from a rewrite. Before replacing it, remove unnecessary polling, decide where duplicate checks belong and see whether two notifications can use the same saved event.
Consider a custom integration when the part that keeps failing needs stronger control over state. For instance, if a CRM record is created but the owner alert fails, your process should remember that the record already exists and retry only the alert. A platform can implement that design too, but it may require a backing store or enough separate workflows that a small integration becomes easier to operate. Compare the estimate for just that part with the cost of maintaining it in the platform.
Price custom code with real ownership attached: who fixes it when an API changes, who reviews failures, and how the team can safely replay work. Use a worksheet with one column for the platform option and another for code. Include implementation, regular operating costs and support for each. Review the worksheet after a busy month. A saving that exists only under a forecast, before you add maintenance, is a weak reason to move.
What should you take to a quote call?
Bring the workflow sketch, a normal and busy month of event counts, and the outcome you cannot afford to miss. If you already run the automation, bring a redacted execution summary and identify where the team does manual repairs. Ask for a quote that separates the tool allowance from building and maintaining the flow. If you want a second opinion on the count, share the workflow with us. One real example is enough to find the first question worth answering.
Bharat Vasireddy
Founder & Digital Marketing Strategist8+ years of experience in performance marketing and web development. I help businesses scale through data-driven SEO, Google Ads, and high-performance websites.