You have many options for seeding of variables. They may be created and seeded in the base language used for your protocol, they may be created using lr_save_*() API references or, they may be fed from files, or they may be fed from web pages returned to the user. This last item is termed 'correlation' is considered a foundation skill for those developing scripts with LoadRunner.
You may refer to your LoadRunner training material for a review of three different methods for identification and collection of information from websites. You may also refer to the standard documentation for LoadRunner, correlation functions related to your virtual user type
Please refer to the questions located here for further assistance.