Test correlation in structural dynamics usually involve modal comparisons between an experimental modal analysis and your FEM modes, or comparison of frequency response functions.
For more specific applications where the representation of a given field is of particular interest, your experimental setup must provide an information closely related to it. You did not provide any details on your application, so the answer can only be very general.
I agree with Mr. Guillaume Vermot Des Roches. It depends on the applications.
In any case you must check the FE model at different level of observation before delivery it. The first one is the global behaviour (displacement of a point, 1st natural frequencies and so on). Than you can pass to the local observation using puntal sensors like strain gauge, accelerometer, and so on. If the material is not homogeneous or anisotropic check accurately the corrispondence between the numerical points and phisical ones.
It could be better if you specify your test bench.