You might have a requirement for having different business processes for each of your different data combinations, supported by one data structure. The approaches that Prema and Danny suggest are for when the business process is the same for the different data combinations, and there are some instances when this isn't the correct approach.
Combinatorial testing is the practice of navigating down the combinations of data which are likely to cause errors. It is a QA discipline that needs careful configuration to make sure you are testing the likely error combinations. That, or a model-based testing approach.
CA Agile Requirements Designer makes combinatorial testing easy, by presenting users with the graphical model, highlighting the paths. Combine this with the data shredding and synthetic generation capabilities in CA TDM, and you have the ability to automatically generate messages for each of the different combinations of data. We have an integration between the output of this process and running virtual services, so SV fully supports this method of testing, and your virtual service will listen to every request, giving an exact match.
Each of the three products have good use cases. Combine them together, with the out-of-the-box integration points, and you get a compelling use case for the automation of this kind of advanced testing technique.