I had similar wondering and this is what I found on AWS:

• Strings are Unicode with UTF-8 binary encoding. The size of a string is (number of UTF-8-encoded bytes of attribute name) + (number of UTF-8-encoded bytes).

• Numbers are variable length, with up to 38 significant digits. Leading and trailing zeroes are trimmed. The size of a number is approximately (number of UTF-8-encoded bytes of attribute name) + (1 byte per two significant digits) + (1 byte).

So Names of the attribute can play significant role it seems.