Why Most n8n Automations Break by Month Three (and How to Build Ones That Don't)
Postat pe - Modificat ultima dată pe
<p>Every automation looks great on launch day. The demo runs, the client is happy, everyone moves on.</p>
<p>Then, somewhere around week eight, leads stop reaching the CRM. Nobody notices for days. When someone finally does, the question is always the same: how long has this been broken?</p>
<p>Why this keeps happening</p>
<p>After years of building and repairing n8n systems, I've seen the same six causes again and again. None of them are about n8n itself. They're about how the workflow was built.</p>
<p>1. No error handling</p>
<p>Most workflows are built for the happy path. The first time an API times out, returns a rate-limit error or changes a field name, the execution stops and the data is gone.</p>
<p>The fix: every workflow that matters gets an error path. In n8n, that means an Error Trigger workflow, retries with a wait between attempts on nodes that call external APIs, and a clear decision about what happens when the retries run out.</p>
<p>2. Silent failures</p>
<p>An automation that fails without telling anyone is worse than no automation, because people assume it's working.</p>
<p>The fix: alerts go where the team already looks, whether that's Slack, email or Telegram, with enough context to act: which workflow, which record, and what failed.</p>
<p>3. Duplicates on retry</p>
<p>Retries fix one problem and create another: the same lead created twice, the same invoice sent twice.</p>
<p>The fix: make critical steps idempotent. Check whether the record already exists before creating it, or use a unique key, so running the same step twice has the same result as running it once.</p>
<p>4. Expiring credentials</p>
<p>OAuth tokens and API keys expire. When they do, workflows fail all at once, usually on a Friday night.</p>
<p>The fix: know which credentials expire and when, refresh them automatically where the API allows it, and get alerted before they expire, not after.</p>
<p>5. Upstream changes nobody announced</p>
<p>Someone renames a field in the CRM, adds a required question to a form or changes a pipeline stage. The workflow keeps running and quietly writes bad data.</p>
<p>The fix: validate the input at the start of the workflow and fail loudly when the data doesn't match what's expected.</p>
<p>6. Nobody knows how it works</p>
<p>The person who built it is gone, and the workflow is a canvas of sixty unnamed nodes.</p>
<p>The fix: name every node for what it does, keep workflows under version control and hand over a short runbook: what it does, what can break and how to fix it.</p>
<p>A quick checklist before you call it done</p>
<p>- Does every external call have retries and an error path?- Will someone be alerted within minutes if it fails?- Can any step run twice without creating duplicates?- Do you know when each credential expires?- Is invalid input rejected instead of silently processed?- Could someone else fix it tomorrow using the documentation?</p>
<p>If you answered no to two or more, the automation will probably break. The good news is that every item on this list can be added to an existing workflow without rebuilding it.</p>
<p>If you have an automation that's already acting up, send me a message here on Freelancer with a short description of your setup, and I'll tell you where I'd start.</p>