Analytical models exist for some magnetic systems, however for these models solutions are only practical for simple cases. Experiments allow observations to be made of real systems, but we are limited to the detail which can be extracted from these measurements.
When computational resources are available, more complicated models can be used which provide a link between experiment and theory. The motivation for using computer simulations is two-fold: firstly, it is possible to interpret experimental results, and secondly new designs can be predicted and subsequently developed, reducing costs.