Most issues I’ve encountered with Excel and CSV input data are due to those formats not storing data as having specific types such as dates, numbers, text, etc.   When importing from those sources, ArcGIS may assume data types, which it does by sampling and examining the first X records.   If later data in that file violates the assumed type, such as text in a what initially appeared to be a numeric field (like a zip code, where American zips tend to be numeric while foreign zips tend to be alphanumeric), you get a load error which can be hard to track down.   Plus, other users that edit Excel files are not restricted in any way in what they type or how they format columns.  Also, Excel will in some cases alter data to suit its own assumptions; for example, all our zip codes here begin with a leading zero, which Excel assumes is not needed and so drops it without confirmation.  As noted previously, while Excel can export to CSV it doesn’t always format the output correctly if say commas are embedded within a cell’s data.

Where applicable I prefer database sources like Access and SQL Server, since they require data to be typed nearly identically to those types found in file geodatabase format and SDE databases.  Depending on the situation, I may first import data from other formats into an Access or SQL Server staging database, and do non-spatial manipulations there before importing into ArcGIS.  Those offer more tools, speed, and functionality to handle formatting, scrubbing, exceptions, and advanced logic than ArcGIS.