n8n Merge Node Patterns: Modes, Join Types and Why Items Disappear
By Abhishek Patoliya · 2026-08-15
Most "half my items disappeared" reports in n8n trace back to the Merge node. It isn't broken. Its modes behave very differently, and the default for matching (keep only matches) drops anything without a partner, with no error. This lesson covers every mode, the five output types for matching, the options that fix type mismatches and duplicates, and how Merge interacts with item linking.
Before you start: read paired items; Merge is one of the nodes that affects them. Options below follow n8n's Merge node documentation as of 9 November 2026.
The four modes
| Mode | What it does | Use it when |
| Append | Outputs the items of each input one after another | Branches produce independent lists you just want in one stream |
| Combine | Joins items across inputs: by Matching Fields, by Position, or All Possible Combinations | Items on one branch relate to items on the other |
| SQL Query | Treats each input as a table (input1, input2…) and runs a SQL query over them | The join is too complex for the Combine options |
| Choose Branch | Waits for both inputs, then outputs Input 1's data, Input 2's data, or a single empty item | You only need one side's data, but must wait for both branches to finish |
Since n8n 1.49.0, Merge can take more than two inputs (set Number of Inputs), and SQL Query mode is available. Older versions support only two.
Combine by Matching Fields: the five output types
This is where items go missing. You choose the field to match on each input (for example customerId), then an Output Type:
| Output type | Keeps | SQL equivalent |
| Keep Matches (default) | Only items with a match on both sides, merged | Inner join |
| Keep Non-Matches | Only items without a match | Anti-join |
| Keep Everything | Matches merged, plus every unmatched item from both sides | Full outer join |
| Enrich Input 1 | Every item from Input 1, with matching data from Input 2 added | Left join |
| Enrich Input 2 | Every item from Input 2, with matching data from Input 1 added | Right join |
Because the default is Keep Matches, a customer with no orders disappears from a "customers + orders" merge. If one branch is your authoritative list and the other is optional extra data, you almost always want Enrich Input 1 (with the authoritative list on Input 1).
Worked example: enrich without losing anyone
Input 1: CRM customers [{ customerId: "101", name: "Acme" }, { customerId: "102", name: "Birch" }]
Input 2: order totals (API) [{ customerId: 101, total: 540 }]
Merge → Combine → Matching Fields
Input 1 field: customerId Input 2 field: customerId
Output Type: Enrich Input 1
Options → Fuzzy Compare: on ("101" and 101 count as equal)
Output: [{ customerId: "101", name: "Acme", total: 540 },
{ customerId: "102", name: "Birch" }] ← kept, no order data
Without Fuzzy Compare (off by default), the string "101" and the number 101 don't match, and with Keep Matches you would get zero items. That looks exactly like a data problem, but it's a type mismatch. Alternatively, convert the field's type with Edit Fields before merging.
Combine options worth knowing
- Clash Handling: when both items have a field with the same name, which one wins (by default Input 2), or add the input number to field names; and whether nested objects are deep- or shallow-merged.
- Multiple Matches: Include All Matches outputs one item per match (a customer with 3 orders becomes 3 items); Include First Match Only keeps one. Unexpected item counts after a merge are usually this setting.
- Fuzzy Compare: tolerate type differences when comparing (see above).
- Disable Dot Notation: treat
a.b as a literal field name rather than a nested path.
- Include Any Unpaired Items (Combine by Position): by default, items without a partner at the same position are left out.
Combine by Position: only for lists that stay aligned
Position pairs item 1 with item 1, item 2 with item 2, and so on. It's safe only when both inputs come from the same source with the same count and order, for example one list split into two branches that each add a field without filtering or sorting. If either branch has a Filter, IF, Sort or an API call that can skip items, positions drift and wrong items are paired without any error. Add an ID and use Matching Fields instead.
Timing: Merge waits for its inputs
n8n's docs say Merge combines data once all connected input streams are available (Append "waits for the execution of all connected inputs"). In workflows created on n8n 1.0 or later, branches run one at a time, top to bottom by canvas position, so Merge runs after the last of its input branches.
The docs don't spell out what a Combine mode outputs when one input receives no items at all (for example an IF branch nothing went down). Test that case with pinned data in your n8n version rather than assuming. If a branch is genuinely optional, Enrich Input 1 with the always-present data on Input 1 is the most predictable design.
Merge and item linking
- After an Append merge, each item links back to only one branch. An expression using
$('Node on the other branch').item resolves to null, without failing the node ("is on another branch" in the expression preview).
- After a Combine, one item can come from several earlier items, which can cause "Multiple matching items". Use
.first() or .all()[i], or carry the fields you need onto the items before merging. Both cases are covered in paired items.
Debugging missing items: a checklist
- Mode: Append never drops items; Combine can.
- Output type: Keep Matches (the default) drops unmatched items; did you mean Enrich Input 1 or Keep Everything?
- Field values: same type on both sides? Turn on Fuzzy Compare or convert first. Watch for whitespace and letter case too.
- Count higher than expected: Multiple Matches is set to include all.
- By Position: did a branch filter or reorder items?
Key takeaways
- Four modes: Append, Combine (matching fields, position, all combinations), SQL Query, Choose Branch.
- Matching defaults to Keep Matches (inner join); use Enrich Input 1 to keep every item from your main list.
- Fuzzy Compare fixes string-vs-number mismatches; Multiple Matches explains unexpected counts.
- Combine by Position only for inputs guaranteed to stay aligned.
- Label deliberate exclusions (rename the node, e.g. "Merge — only customers with orders").
Exercise: build the worked example with two Edit Fields nodes as inputs (one with customerId as strings, one as numbers). Run with Keep Matches and Fuzzy Compare off (0 items), then turn Fuzzy Compare on (1 item), then switch to Enrich Input 1 (2 items). Next lesson: the Aggregate node.
References (checked 9 November 2026)
Written by Abhishek Patoliya, Senior n8n Developer & Automation Architect, Surat, India. Verified n8n Creator.