n8n Fundamentals
Every workflow you will ever build in n8n — a simple lead-notification automation or a multi-agent system with sub-workflows — is assembled from the same small set of primitives. Learn these once and the rest of this guide is just recombination.
Why fundamentals first
Most people learn n8n backwards: they copy a template, get something working, and never learn why it worked. That's fine until the template breaks or the client asks for one small change — and then there's nothing to fall back on. n8n's own "build with AI" prompt-to-workflow feature will get a workflow 50–70% built for you in seconds, but the last 20–30% — the part that actually makes it production-ready — is where the fundamentals below do all the work. Learn them once, deliberately, and every workflow after this becomes a matter of recombining the same handful of ideas.
Triggers vs. actions
Every workflow has exactly two kinds of nodes. A trigger is what starts the workflow — a form being submitted, an email arriving, a schedule firing, a webhook being hit. An action is everything that happens afterward — adding a row, sending an email, calling an API. A workflow needs at least one trigger and usually several actions strung together.
Instant triggers vs. polling triggers
Triggers come in two flavors, and the distinction matters for both speed and cost:
| Type | How it works | Example | Trade-off |
|---|---|---|---|
| Instant | The external app pushes data to n8n the moment something happens. | Webhook, chat message, Telegram message | Zero delay, but the app has to support pushing (a webhook). |
| Polling | n8n asks the app "anything new?" on a fixed interval. | Gmail trigger, Google Drive trigger | Delay equal to the poll interval — and every poll can count against your execution quota. |
A one-minute poll runs 1,440 times a day. On n8n's free/starter execution allowance that alone can burn through your monthly quota in under two days. As a rule of thumb: poll as infrequently as the use case tolerates — hourly is plenty for most inboxes — and reach for an instant trigger (webhook, chat, or a native "instant" trigger where the app offers one) whenever you can.
Nodes, credentials, and the canvas
A node is one step in a workflow — a single connection to an app, a single logic operation, or a single transformation. Click a node to open its configurator, which is always laid out the same way: Input on the left (data available from every node that ran before it), Parameters in the center (what you're telling this node to do), and Output on the right (what the node produced). Close a node with Escape, the "back to canvas" control, or by clicking anywhere outside it; reopen it by double-clicking.
Most app nodes (Google Sheets, Gmail, Slack…) reduce to four underlying operations no matter what the app calls them: create, read / search / get, update, and delete. Once you notice this pattern, unfamiliar nodes stop being intimidating — you're really just asking "am I creating, reading, updating, or deleting something in this app?"
A credential is n8n's word for a saved login — an API key, an OAuth connection, or a token. Create one the first time you connect an app; every workflow after that reuses it. Give credentials and workflows real names as you go (not "My workflow 3") — you will thank yourself in a month.
Mapping & expressions
Mapping is the single most important skill in n8n: moving data from one node into a field in a later node. Every node has access to the output of every node that ran before it in that branch — not just the one immediately before it. In practice, mapping means dragging a value from the Input panel onto a field, which writes an expression.
{{ $json.full_name }}
Anything inside {{ }} is live code (JavaScript under the hood, though you rarely write
it by hand) evaluated at run time. $json always refers to the current item; if you need
data from a node further back than the one immediately before it, reference that node by name —
{{ $('Telegram Trigger').item.json.chat.id }} — because $json alone only
ever reaches back one step.
Fixed value vs. expression
A field left as a plain string is a fixed value — it never changes between runs. A field toggled to expression is dynamic — it's re-evaluated per item. Use fixed values for constants (a subject line prefix, a fixed email address) and expressions for anything that should differ per lead, per email, per row.
+ characters from this phone
number"), and paste the answer back in. This is the normal way experienced n8n builders write
anything more complex than a straight field mapping.
The six data types
Everything that moves through n8n is JSON underneath, and JSON has exactly six data types. Learn to recognize each one on sight — n8n color-codes and icon-codes them for you — because most "weird" errors turn out to be two mismatched types being compared.
| Type | Looks like | Recognize it by | Typical use |
|---|---|---|---|
| String | "jane@acme.com" | Purple, wrapped in quotes, "T" icon | Names, emails, messages |
| Number | 1000 | Green, no quotes, "#" icon | Budgets, counts, IDs |
| Boolean | true / false | Green, no quotes, only two possible values | Flags — rejected, verified |
| Array | [ "apples", "bananas" ] | Square brackets — an ordered list | Multiple leads, multiple line items |
| Object | { "name": "Jane" } | Curly braces — key/value pairs | One lead, one row, one receipt |
| Binary | A file "chip" in the output panel | Downloadable/viewable box, not text | PDFs, images, audio, CSVs |
The two that trip people up are array and object — and they're almost always nested together: an array of objects. A Google Sheet is exactly this shape: the sheet is the array (a list of rows), and each row is an object (name, email, budget…). Once you can look at a spreadsheet and a JSON array side by side and see they're the same data, arrays and objects stop being abstract.
1000 and the other is the string "1000".
This single mismatch is responsible for more n8n errors than everything else combined.
Filters & conditionals
Three nodes handle almost every branching decision you'll need:
- Filter — keeps items that match a condition and silently discards the rest. Use it to drop leads below a budget threshold, or emails that don't have an attachment.
- IF — splits data into exactly two paths, true and false, based on one condition (does
serviceequal"ads"?). - Switch — splits into as many paths as you need, matching one field against a list of possible values (route an email to promotions, social, personal, sales, or misc).
For Switch, use "contains" rather than "equals" when matching AI-generated text — a model might return "category: promotions" instead of a bare "promotions", and an exact-match condition will silently fail to route it anywhere.
Merging paths back together
Once an IF or Switch splits data into multiple branches, the Merge node (in "Append" mode) brings them back into a single stream so the rest of the workflow doesn't need to be duplicated per branch. This is how, for example, both the "ads" and "SEO" reply paths in a lead workflow can converge back into one shared "update the spreadsheet" step instead of two identical copies of it.
Split Out & Aggregate
These two nodes are inverses of each other, and together they're how n8n handles "do this to every item in a list."
- Split Out takes one item containing an array (four leads in one list) and turns it into four separate items, each processed independently — filtered, mapped, and written to a sheet one at a time.
- Aggregate does the reverse: it waits for every item in a batch to finish, then bundles them back into a single list. Use it before sending a summary notification — without it, ten processed leads means ten separate emails instead of one digest.
When looping over API calls or database writes at volume, pair a Loop Over Items node with a short Wait step between iterations to stay under the target app's rate limit (Google Sheets, for example, throttles reads and writes per minute) — and set Continue on Fail in node settings so one bad item doesn't kill a 1,000-item batch.
Pinning data
Rebuilding a workflow by re-submitting the same form or re-sending the same message dozens of times while you build is a special kind of tedium. Pin Data freezes a node's last output so you can keep re-running everything downstream of it without re-triggering the source. Pin the trigger while you build, edit the pinned JSON directly to test edge cases (a budget of $1,200 to pass a filter, a service value of "SEO" to hit the other branch of an IF), and unpin before you go live — pinned data is a test fixture, not what runs in production.
Where to go next
With triggers, mapping, data types, and branching in hand, you have everything needed for the first real build: a complete instant lead-response system, end to end.
Sources
| Source | Used for |
|---|---|
| Jono Catliff (Automatable), "n8n 8 Hour Masterclass — Build & Sell AI Agents (Automate Anything)" | Primary source for this guide's concepts and terminology |
| n8n docs, "Work with nodes" | Trigger/action model, node basics |
| n8n docs, "Understand n8n's data structure" | The six data types, arrays & objects |
| n8n docs, "Expressions" | Mapping & expression syntax |