n8n Compare Datasets nodeConfigure Compare Datasets in n8n.

This is where the comparison starts. You name one field per input, and n8n treats items sharing that value as the same record before it looks at anything else. Select Add Fields to Match to compare on several keys at once.

• Input A Field: the field name taken from input stream A, typed as text, for example id.
• Input B Field: the matching field on input stream B, which can carry a different name when the two systems disagree on naming.

Once two items match on the key but disagree on the rest, this selector decides what the Different branch actually contains.

• Use Input A Version: treats input stream A as the source of truth.
• Use Input B Version: hands that role to input stream B.
• Use a Mix of Versions: output uses different inputs for different fields. It opens Prefer, where you pick the main source, and For Everything Except, the field names pulled from the other side, comma separated.
• Include Both Versions: output contains all data, with a more complex structure.

Turned on, the comparison tolerates small type differences between fields. The number 3 and the string '3' are then treated as the same, where a strict comparison would send the pair straight to the Different branch. Leave it off when a type change is exactly what you are hunting for.

• Fuzzy Compare: a switch on the main panel, off by default, applied to the whole comparison rather than to one field.

Some fields move on their own and say nothing about the record. This option lists the ones that shouldn't be included when checking whether two items are the same, as text, separated by commas.

• Fields to Skip Comparing: field names to ignore, for example updated_at, updated_by. In the documented example, two datasets matched on person.language come back as different until person.name is added here, after which n8n returns them as matching.

The same type tolerance is also reachable from the node Options collection, which you open with Add option. Every option you add shows up as one more field in the panel, and the ones you leave out keep their default value, so adding it here is how you make the setting explicit next to the other comparison tweaks.

• Fuzzy Compare: whether to tolerate small type differences when comparing fields, the number 3 and the string '3' being treated as the same.

By default a field name written parent.child reaches into a nested object. Turn this option on and that stops: the name is read literally instead. Off, which is the default, keeps the nested reference working.

• Disable Dot Notation: whether to disallow referencing child fields using parent.child in the field name.

Duplicate data is the normal case, not the exception, and this selector says what to do with it. The default is Include All Matches.

• Include All Matches: output multiple items if there are multiple matches, which on the documented fruit example gives three items in the Same Branch tab.
• Include First Match Only: only ever output a single item per match. Two items come back instead, the repeated apple counting once.