- A builder often suits a short, standard workflow. Test the choice against your actual requirements.
- Compare total costs, destination results and safe recovery rather than the number of blocks.
- Self-hosting does not prevent external data transfers. A difficult component can be replaced separately.
n8n is a workflow automation builder: you connect applications into a sequence that carries out agreed actions without someone copying the data by hand. Make and Zapier address similar needs. A builder will often be enough to move a website enquiry into a task list and notify a colleague. When a process controls stock reservations, staff permissions or several related records, check its rules and recovery first. You may need one custom component, or a system built around the process.
The useful question behind “n8n vs Make vs custom development” is what your business promises, and what happens when a step fails. The examples below are hypothetical. They illustrate decisions rather than describe our clients or promise a service for setting up any builder.
What is n8n, in plain English?
Imagine an enquiry moving through these steps: receive the details, check the phone number, find the customer, create an enquiry and send a notification. A visual editor represents the actions as connected blocks. A connector is a ready-made connection to an application. A trigger is the event that starts the sequence, such as a new enquiry or a scheduled time. The complete sequence is called a workflow.
Applications exchange information through an API, an agreed way to send requests and receive responses. Seeing a service in a connector catalogue does not establish that the action you need is available. Check the specific operation: creating a contact, updating an order or fetching changed stock levels. Also check whether the connected service allows it on your subscription.
n8n, Make and Zapier all provide catalogues of application connections. These can save the work of writing a separate connection for a routine exchange. They do not define your business rules. A block that creates a customer still needs an answer to “How do we recognise a customer who already exists?”
When a workflow builder is enough
A good first use has a clear input and an outcome you can check. For example, a website enquiry enters a task list and the assigned employee receives a notification. Each enquiry has a reference, duplicates can be recognised, errors are visible and a short delay is acceptable. Building a separate system just to transfer those fields may add unnecessary work.
Another example is preparing an internal report from information that has already been checked. If preparation fails, an employee can see the failure and run it again. Ready-made connections can shorten the path to a useful trial. For a simple flow with modest usage, setup and subscription costs may be lower than separate development. That is something to calculate for the actual process.
A builder also helps when a team is still refining the sequence. People can check whether an extra notification or validation step is useful before committing to a larger system. Someone must still own the running workflow. A readable diagram does not supervise the result.
How to compare n8n and Make for your business
Try one representative process in both tools using test records. Check that the required actions exist, all necessary fields arrive and an employee can understand a failed run. This tells you more than the number of connectors in a catalogue. Use the same questions if Zapier is also under consideration.
n8n adds a hosting decision: use its cloud service or run it on your own infrastructure. That changes who looks after the server. If nobody in the company can maintain or restore it, the option to install it yourself does not establish a cost saving.
Look beyond the successful demonstration. Try a missing required field, the same enquiry arriving twice, an unavailable application and revoked access. A suitable tool should let the people responsible understand what happened and take the agreed next step. This is a decision about your process, rather than a general ranking of products.
What you pay for: executions, credits and tasks
The billing models below were checked against the official pricing pages on 20 October 2026. Check the current plan and counting rules before subscribing. No fixed prices are quoted here. One business enquiry does not necessarily consume one billing unit.
- n8n: paid plans use workflow executions as the main counting unit: a run of the complete workflow rather than each block. Plan limits and additional services still need checking.
- Make: credits are the billing unit. For ordinary non-AI actions, one operation generally uses one credit. Some features follow different rules, including usage-based AI features.
- Zapier: tasks generally represent successfully completed actions. Triggers do not consume tasks, and some actions have exceptions. Count the specific steps your flow will use.
Consider an illustrative calculation: 500 enquiries each month, with four billable actions per enquiry. Under a per-action model, that means 2,000 units before exceptions. A model that counts the complete workflow run starts from a different basis. This is arithmetic to explain the distinction, not a forecast of any platform's invoice.
Include repeated events, lists of records and peak days. Add the bills from connected services and the time spent investigating failures. Compare total monthly running costs, including support. Custom software also needs hosting and maintenance, so moving to it does not automatically remove recurring costs.
Where a builder starts needing more work
Branching and shared business rules
A large diagram alone is not a reason to rebuild. Difficulty increases when decisions depend on earlier changes: who approved an order, whether its price changed and whether its stock has already been reserved. If several workflows repeat the same rule, someone may update only one copy. The company then gets different answers to the same question.
A builder plus custom development can work when a difficult check can be separated into a small software component. The workflow sends information and receives a decision. A custom system becomes more relevant when staff need shared rules, decision history and screens for their work. Moving an unclear process into code will not clarify it by itself.
Volume and processing time
Copying one contact and synchronising a large catalogue involve different workloads. Establish the normal and peak number of records, file sizes and how often the external application permits requests. There is no universal order count beyond which n8n or Make becomes unsuitable.
Test with a workload resembling real use. Measure the time until the correct record appears in the destination, and watch the queue: records waiting to be processed. If the delay breaks a customer promise, the processing method needs attention. A batch transfer or one separate component may solve it without replacing everything.
Failures, retries and an unknown outcome
Suppose the flow creates an order but the response never arrives. Repeating the creation blindly could create a duplicate. Protection against this is called idempotency: receiving the same event again should not repeat an action that has already completed. It needs a stable event reference and a check of the actual outcome in the receiving system.
n8n provides error workflows and execution inspection. Having these tools does not establish that a particular workflow recovers correctly. Agree when a retry is safe, when processing must stop, who investigates an unknown outcome and which unfinished step can resume. Resolve this before using the process for reservations or payments.
Decision table: builder, custom component or custom system
These hypothetical examples suggest what to investigate first. The choice still depends on a tested process, data access and an acceptable delay.
| Approach | Example | Condition |
|---|---|---|
| Builder | Enquiry to a task list, with a staff notification | Required actions exist, duplicates are visible, delay is acceptable |
| Builder + custom component | A standard enquiry transfer with a separate contract check | The difficult rule can be isolated and tested separately |
| Custom development | A staff portal with roles, shared stock reservations and approval history | Related changes need consistent saving and access controls |
This is not a progression every business needs to follow. A stable notification can remain in a builder for years alongside a custom system. Rewriting needs a specific problem to solve. Owning more code is not a useful outcome on its own.
n8n Cloud or self-hosting: where does the data go?
According to n8n's pricing page, checked on 20 October 2026, hosted-plan data is stored in the EU on servers in Frankfurt, Germany. With self-hosting, you choose the infrastructure. This describes where n8n is hosted, rather than the complete journey of information through your workflow.
If your server sends text to a cloud model or writes it to another service, that information leaves the server. Map the source, fields sent, recipients and retention periods before connecting company data. For Make and Zapier, check the hosting and processing terms of the selected plan rather than assuming they match n8n.
Self-hosting adds responsibility for updates, backups and tested restoration. n8n offers a free Community edition for self-hosting; paid editions and applicable usage terms need separate checking. A free edition does not make operation free. Assign someone who can look after it, including arrangements for holidays and absences.
Give the workflow only the access it needs
Keep service accounts under company ownership. Grant access to the required records and actions rather than connecting a personal administrator account for a simple notification. Check whether execution history retains unnecessary personal information. Review workflow exports before sharing them too, because settings may contain sensitive values.
Who looks after the workflow when its author leaves?
Monitoring means checking that the promised outcome exists. A green run indicator does not prove that the right enquiry reached the right employee. Reconcile incoming events with destination records and make events without a result visible.
Before handover, ask for a description of the flow, its connections, retry rules and a shutdown procedure. The next person should be able to locate an unfinished event, check its destination and resume processing. Have them practise with test data without coaching from the author. Also agree who responds to failure alerts and during which hours.
Access can expire, an employee can change a password and a connected service can stop responding. None of these requires a change to the workflow itself to interrupt the business process. Support arrangements need to cover them. Our separate guide to business automation mistakes covers the organisational side of implementation.
What to check at handover
Start with a normal test enquiry. Record its reference, find the destination record and check the fields and notification recipient. Submit the same event again. Define the expected behaviour first, such as updating the existing enquiry without another notification. Check the receiving application as well as the diagram.
Then omit a required field. The workflow should show an understandable reason for stopping and avoid creating a partial record where your rules prohibit one. Temporarily block a test connection. Check who receives the alert, whether the unfinished event is visible and what happens after access returns. Keep real payments and customer messages out of these tests.
Also test silence: no results arrive, yet no explicit error appears. Events may have stopped reaching the workflow altogether. Agree how to distinguish a quiet day from broken intake, perhaps through a test enquiry checked at its destination. Preserve these examples as acceptance criteria and repeat the relevant checks when fields, permissions or connected services change.
How to move from a builder without throwing away the work
Do not start by switching off the old flow. Preserve its rules, sample inputs and expected outcomes, field mappings and record references. n8n can export workflows as JSON, a text format for structured data. Make can export a blueprint containing modules and settings. These are useful documentation, rather than the finished code of an independent application.
- Identify the part to replace: validation, storage or the complete process.
- Compare old and new results using test data. Include missing fields, repeated events and an external application refusing a request.
- Allow only one route to make real changes. Parallel checking must not create two orders or send two notifications.
- Agree the switch, treatment of waiting records and a way back. Returning to the old route must account for records already created by the new component.
Keep connections that still do a useful job. The earlier work retains value in tested rules and examples even if some blocks need replacing. An export does not promise a one-click migration or uninterrupted service. Those outcomes require their own plan and checks.
What to bring to Artbrain
Describe the starting event, required result, typical and peak volume, acceptable delay and consequences of a duplicate. Add a diagram of the current route without secret keys or personal information. Passwords are not needed in the first message. We can advise whether a builder is sufficient. When custom development is needed, we discuss the scope and prepare a written estimate.
We build our own systems and connect external services to them. Our catalogue of integrations for Artbrain systems describes that scope: the connections apply to systems we develop. Setting up or maintaining n8n or Make is not an offered service in this description.
If the task centres on company documents and answering customers, consider custom AI assistant development separately. Sources, access permissions and required actions are agreed before work begins. Adding an AI step does not, by itself, decide whether the rest of the process needs custom development.
FAQ
What is n8n in simple terms?
n8n is a workflow automation builder that connects actions across applications. A new enquiry might trigger a data check, a record being created and a staff notification. Validation and retry rules still need to be defined.
Should I choose n8n or Make?
Test the same process with sample records in both tools. Check the required actions, field transfers, running costs and recovery after failure. For n8n, also decide between cloud hosting and your own infrastructure. The choice depends on the process requirements.
Can I use n8n for free?
A free Community edition is available for self-hosting. Check its usage terms and the differences from paid editions separately. Hosting, updates, backups and maintenance still require resources.
How do n8n, Make and Zapier billing models differ?
As checked on 20 October 2026, paid n8n plans mainly count workflow executions, Make uses credits and Zapier counts successful actions as tasks. Exceptions and additional rules apply. Estimate usage for your actual workflow against the current plan.
When should a business choose custom development over a builder?
Consider it when the process needs shared complex rules, consistent updates across related records, staff roles and working screens missing from the current solution. Sometimes one custom component is sufficient. Artbrain can advise whether a builder will do and scope a custom system where needed.
