Question 1

In what shape does n8n pass data between nodes?

Accepted Answer

State the structure, then give the practical rule it implies — the second half is what makes the answer memorable. "All data is an array of objects. Each element is an item and can carry two keys: json for ordinary data and binary for file data. In the binary part the file is stored base64-encoded alongside metadata like mimeType and fileName. The most important consequence: one item = one run. Ten items into an HTTP Request node means ten requests; ten into a Gmail node means ten emails. Not knowing this leads to two classic mistakes: building loops by hand where none are needed, or sending 200 emails when you expected one. The item count also changes from node to node — Split Out raises it, Aggregate lowers it. The item count is the first thing I look at when debugging."

Question 2

An API returns a 500-element array inside one item, and you need to act on each element. What do you do?

Accepted Answer

This is a practical-task question — describe the solution together with its consequences. "I use the Split Out node to turn the array into separate items: 1 item in, 500 items out. From there the downstream nodes work over each element automatically — no hand-built loop needed. But I immediately account for two things. First, rate limits. 500 items into the next HTTP Request node means 500 requests. If the API has a limit, I batch with Loop Over Items and put a Wait between batches. Second, unneeded fields. If each of those 500 items has 40 fields and I need 4, I trim with Edit Fields straight away — for memory and for readability. If the result needs to be reassembled at the end, Aggregate does the reverse and collects everything back into one item."

Question 3

A workflow performs more operations than expected — 50 emails instead of one. What could be the cause?

Accepted Answer

Name the cause, then list three different fixes — that is what shows depth. "The cause is almost always the same: the node receives 50 items, not one, and runs once per item. So the problem is not in that node but in the step before it. I check that node's input item count first — it is shown in the node header on the canvas. Then I pick one of three fixes depending on what I actually want: Only the first item matters — turn on Execute Once in the node's Settings tab Everything should become one message — put an Aggregate node in front to collapse them into a single item Only some are needed — reduce with Filter or Limit This bug is dangerous because it hides during testing with a single item and surfaces in production when 50 arrive. So I always make sure my test data contains more than one item."

n8n's data structure: items

📚 Sources and documentation