Make vs n8n for a small business

Make and n8n connect the tools a company uses. An example process: an enquiry arrives, its details go into a spreadsheet, a model prepares a summary, and a salesperson receives it with the enquiry in the CRM. The platform you choose affects how that process is built, what it costs to run and who can change it later.

I work with Make, n8n and custom code. Before choosing, check the specific integration you need and establish who will look after it once it is running.

Who will update the automation when the form changes

Suppose a new field is added to a form. Someone needs to include it in the workflow, map it to the right place in the CRM and check that the data is saved correctly. If an employee will do this, involve them in choosing the tool.

In Make, you build a process in a visual editor using modules and connections. Standard integrations can cover a lot without writing code. It is worth trying when a team wants to change fields, conditions and notification text themselves. The Make platform description outlines scenario building and the available features.

n8n also has a visual editor. You do not need to assume every change will require a developer, although more involved workflows can require expressions and work with data structures. Hosting is a separate choice: you can run n8n in the provider’s cloud or on your own infrastructure. The options are described in the n8n documentation.

A useful practical test is asking the future maintainer to make one change in a working example. Have them add a field or find the cause of an error. This shows what they can handle themselves and where they will need help.

How to compare costs

Make bills usage in credits. In many modules, one operation uses one credit, but some AI features follow different rules. With your own connection to a model provider, you may pay separately for Make credits and the provider’s tokens. The details depend on the module and connection type, as explained in Make’s credits documentation.

Paid n8n plans base workflow billing on executions of the whole workflow. One execution can include many steps. That is a different unit from a single module operation in Make, so comparing two matching numbers on the pricing pages would be misleading. The n8n pricing page defines an execution.

For a useful comparison, run the same process on the same sample data. Check consumption, retries and what happens when you exceed the allowance. Include model charges and any additional services. A long workflow run once a day can produce a different bill from a short task triggered by every email.

Your own server needs looking after

n8n offers a free Community edition for self-hosting, as well as paid editions with additional features. The available options are listed in n8n’s plan documentation.

With your own installation, you need to pay for infrastructure and arrange maintenance. Someone updates the software, makes backups and responds if the service stops working. Establish who can restore it and whether restoring from a backup has been tested.

If nobody inside the company or on the implementer’s side is responsible for that work, compare Make with n8n Cloud. Running your own server makes more sense when you need control over the installation and have someone responsible for it. The absence of a Community subscription charge is not enough to calculate the total cost.

Where the data goes

Running n8n on your own server lets you control that part of the system. Connections to other services still need separate checks. If a workflow sends an email’s contents to an external model, those contents leave the server. The same applies when saving data to a cloud CRM.

Map the route: form, automation platform, model, CRM, logs. At each point, establish what is stored, who can access it and how long it is kept. This can reveal, for example, that the full customer enquiry remains in execution history even though only the case number was needed for subsequent work.

Check the specific integration

Finding an application’s name on an integrations list is not enough. You need support for the action you want to perform. A module might be able to add a contact without exposing a field your CRM uses. In that case, check another connection or the possibility of using the API.

During the trial, also try disconnecting access or supplying incomplete data. Check whether you can locate the error and retry the task safely. If a retry creates a duplicate record, that needs fixing before regular use.

For a stable task, you can consider a script, for example in Python. It will need maintenance, testing and somewhere to run, but it may suit a process nobody needs to edit visually.

If you have not described the process yet, start by collecting examples as outlined in the post on your first AI implementation. They make it easier to establish what you need from the platform.