This is expected n8n behavior. A few things cause differences between manual and scheduled runs:
-
Schedule Trigger data - When you run manually, the Schedule Trigger produces mock/static data (just the trigger timestamp). In a scheduled run it fires at the actual cron time. If any downstream node uses
$('Schedule Trigger').item.jsonvalues like the trigger time, results will differ. -
Gmail
newer_than:3d- This is time-relative. In manual mode you might run it at a different time than when the schedule fires, so the Gmail query returns a different window of emails. -
Execution context - Manual runs sometimes execute with cached/pinned data if you have pinned data on any node. Check that no nodes have pinned data active.
To debug: add a āSetā node right after the Schedule Trigger and log {{ $now }} - compare the value between manual and scheduled runs to see if timing is causing the difference. Also confirm no nodes have pinned data (the pushpin icon).
Hi @Richard8, welcome!
Iād start by logging item counts after each critical stage (merge, filter and match, find items) and remove executeOnce temporarily during debugging, because that is often where scheduled runs diverge from editor runs.
Thank you. I will attempt this. The error is totally consistent, and only with one set of data. āPrice please to steam clean three seater settee and two armchairs, thanksā the time trigger result in 1 x armchair. the manual execution results in 2 x armchair. n8n AI and gemini have tried to help for the last week but I still cant get the code to correct this error
thank you. im trying this
Hi Tamy, Although Ive spent a year learning to use n8n I am at a loss how to solve the difference in manual execution and time trigger for some workflows. It give quotes. When an item has a word number it defaults to 1x. eg "Price please to steam clean three seater settee and two armchairs, thanks
" becomes ā1x three seater settee (Ā£40), 1x armchairs (Ā£35.5)ā BUT only on time trigger. n8n AI and Gemini cant solve it. If I shared the workflow would you be interested to take a look? The AI agent verion work perfectly, I jut wanted one without AI. Regard
Richard
hi @Richard8, of course
go ahead
have you pinned any data by any chance? This will make manual executions be different from production runs
Hi. Thank you for replying. Gemini also said this and I checked carefully and there is no pinned data. The items and prices are taken from google sheets. If I add items both manual and time trigger flows use the exact same data, but on prices (usually where there are words number, eg a two seater sofa,) the time trigger converts to a single item, whereas the manual trigger works. Im baffled as to why its different. Ive been using the n8n AI and gemmini for ages and canāt solve it..
can you let me know your email so I can share it with you please
You can send it through the messages here on the platform, select my profile and send it via āmessageā