Here is one request where user wanted to replace all commas, but not to replace if comma is present within quotes. Example and solution are as follows:

input:

Name,Age,"City, Country",Profession
Mitch,27,"Melbourne,Australia",Student
John,31,"Boston,USA",Chef

Expected output:

Name      Age      "City,Country"      Profession
Mitch       27        "Melbourne,Australia"      Student
John        31       "Boston,USA",      Chef

Solution:

$ sed -r 's/,/\t/g;s/("\w+)\t/\1,/g' test.txt 

Name    Age "City, Country" Profession
Mitch   27  "Melbourne,Australia"   Student
John    31  "Boston,USA"    Chef

here is the explanation of the code

sed -r -use extended regular expression.
s/,/\t/g - Replace all commas with tab (first sed command)
; - starts next sed command
s/("\w+)\t/\1,/g - This has two parts: store a pattern, replace pattern followed by tab with pattern it self and comma. Here the pattern is ", one or more word characters. This way of recalling a pattern is called back-referencing.
1 is order of capture pattern. If there are two patterns, they would be called after order of appearance.