A matrix is best and fastest if you just have numbers everywhere. If you have tabular data where different columns have different data types, like col 1 is scalar numbers, col 2 is strings, and col 3 is vectors, etc. then you should use a table. Tables can have different data types in every column - double matrices cannot. But all elements of a particular column mut be the same, so can't have a double in row 1 and a string in row 2.
If you need maximum flexibility, at the cost of speed and overhead, then you can use a cell array, which has no requirements that data types match up within rows or columns. Basically with a cell array, anything can be anywhere. See the FAQ: FAQ: What is a cell array?
Structures are pretty similar to tables but are probably less efficient than a table. They are more flexible than a table though, for example one structure might have only some of the fields that another structure in the same structure array has - but it's pretty rare that you'd have that situation.
Bottom line: I recommend you become better acquainted with tables. (I assume you know everything about regular numerical matrices of course.)
