Unless you have special requirements for handling versions of active evidence, typically you will find the Engine's support for succession sets the easiest option when later working with evidence in your eligibility/entitlement rules (see Active Succession Set Rule Objects.
The initial implementation of the "data" rule classes should concentrate on creating the rule attributes with the correct data types.
Later, you can add:
Product-specific calculations should be placed on product-specific "calculator" rule classes later.
Once the "data" rule classes have been written and their rule sets published, the developer can write and publish rule object propagator configurations so that rule objects are automatically created by the Engine. These configurations will need to be in place before the overall product can be tested online, but are not required while the eligibility/entitlement rules are hard-coded, and are not required in order to unit-test parts of implemented eligibility/entitlement rules.
Later you can revisit your data rule classes and check for commonality across your products. You can refactor your rule classes so that any which are common across products are normalized into a common rule set, and thus are propagated to only once (to improve performance and lower database storage requirements).