Feeder length approximation as part of Arc Flash Data Collection can sometimes be a headache, typically we measure the approximate distance from panel to panel, and then add the distance to the roof/ground and return, to get an approximate length for cables, I mostly use a laser distance measurer which is quite convenient and fairly accurate.

Does anyone do this differently? e.g. contacting cable supplier for feeder schedule etc...

Sounds as if you have a good method in place. Another option which I have used is AutoCAD.

If the client can supply a layout of it's plant. Else, it's pretty neat and shiny, but it's a lot of effort for the result.

I think your method is good plus if you can get (not usually unless a new install) cable pull sheets.

The other thing to do is if you are unsure of conductor lengths to input various length to see where the hazard category changes. You may be out to several hundred or even a thousand feet before a change. Voltage drop may become more of an issue then and you can reasonably assume a length but this should be stated in the report.

If to scale AutoCAD drawings are available then all we'd need to do is to use the "Measure Distance"option from panel to panel, quite convenient, and perhaps add the distance to the roof/floor and return

Cable pull sheets are also good if available, thanks for the feedbacks I will keep it in mind

If the site is new, and labelling is correct and you can get access to both ends, Teck cable has meter marks on the outside. Otherwise sounds like you are definately getting it close.