Usually would be a projection difference, like local projection vs web mercator is very common with agol to cause differences there. The 206.375 is interesting that it matches (to 6 sf, compared to the first feature which is 4% different) , are there more like it?

Perhaps there are arcs or ellipses? Area the areas complex aggregates, suggesting that holes are measured differently?

This probably has a logical cause, but without data it is hard to say as it can have a lot of different causes. Can you share an FFS with a record where the values really differ?

I think the SHAPE_Area will change towards AREA_SQM when repair geometry has been run.

I have attached a sample of the data for you guys to have a look,

Unfortunately I have to do a manual Field Calculator in ArcGIS to match the areas after the GDB has been written...which isn’t very clever tbh!

The workflow has got an ‘AreaGapAndOverlapCleaner’ at the end of the process but it shouldn’t be the case of the discrepances...

Can you try and explain what is unexpected? 

If I use an AreaCalculator, my results are as good as identical to the values in the column SHAPE_Area, largest being 2.9e-11.

Thanks ​@nielsgerrits  for looking into this…

I think I have sorted now by adding an additional AreaCalculator rigth before the writer and seems to do the trick.

…maybe that’s what I had to do: recalculate the areas again (rookie mistake!)