{"id":21090,"date":"2026-10-10T22:18:00","date_gmt":"2026-10-10T20:18:00","guid":{"rendered":"https:\/\/www.lukaswojcik.com\/blog\/?p=21090"},"modified":"2026-10-05T14:51:34","modified_gmt":"2026-10-05T12:51:34","slug":"saas-n8n-make-code-en","status":"publish","type":"post","link":"https:\/\/www.lukaswojcik.com\/blog\/en\/saas-internet-earning\/saas-n8n-make-code-en\/","title":{"rendered":"Micro-SaaS with n8n, Make or Custom Code: What Operation Really Costs"},"content":{"rendered":"

A workflow prototype becomes visible quickly. A sellable product also needs tenant separation, billing, failure handling and support. Comparing n8n, Make and custom code therefore requires the same customer task rather than a count of integrations.<\/p>\n

\"One
Conceptual comparison; no measured vendor values.<\/figcaption><\/figure>\n

One reference workflow<\/h2>\n

The assumed product accepts a product file, validates required fields, compares prices with a previous baseline and delivers a report. Every route uses identical sample files, expected results and size limits. Three completed implementations were not tested here; this is a design and evaluation protocol.<\/p>\n\n\n\n\n\n
Route<\/th>\nStrength<\/th>\nCheck<\/th>\n<\/tr>\n
n8n<\/td>\nVisible workflows, infrastructure control when self-hosted<\/td>\nLicensing, updates, credentials, isolation<\/td>\n<\/tr>\n
Make<\/td>\nManaged platform and quick integrations<\/td>\nCredits, data volume, limits, exit<\/td>\n<\/tr>\n
Custom code<\/td>\nPrecisely defined processing and tests<\/td>\nDevelopment, operation, dependencies, on-call work<\/td>\n<\/tr>\n<\/table>\n

Licensing before architecture<\/h2>\n

n8n’s help distinguishes hosting client workflows and credentials from embedding with customer access, naming Enterprise and Embed respectively. Consulting on client-owned instances is treated separately. This is not blanket permission for every SaaS backend. The actual data and access routes must fit the agreed licence. [1<\/a>]<\/p>\n

Consumption per customer task<\/h2>\n

Make bills through credits. Consumption depends on the modules and processing; AI and code steps can count differently. A statement such as \u201cten steps\u201d is not a cost forecast. [2<\/a>]<\/p>\n

Assuming eight billable actions per record across 500 records gives 4,000 actions before retries or branches. This is synthetic volume arithmetic, not a measured Make bill or a general rule for credit conversion.<\/p>\n

Failure handling defines the service<\/h2>\n

Evaluation needs an empty file, missing fields, a timeout after saving a result, a duplicated job and revoked access. A stable job identifier helps avoid duplicate processing. Repeated delivery must not silently duplicate billing. Logs need customer attribution without unnecessary full raw data.<\/p>\n

The protocol records development hours, successful jobs, retries, platform consumption, support and recovery time. Only a delivered complete report counts as service. Custom code also carries update, monitoring and on-call costs.<\/p>\n

Choosing a route<\/h2>\n

A platform can simplify a bounded integration pilot. Specific validation rules, large datasets or strict isolation justify comparing custom code. A hybrid can keep customer jobs and billing in the backend while using a platform for selected integrations. The common reference task, rather than tool preference, supplies the decision.<\/p>\n

\n

Related tools<\/h2>\n