Imagine you have an answer set program that models a complicated set of rules. You want to know under what fact scenarios a given outcome can be derived according to those rules. You run a generic query, without modelling specific inputs in the program, just modelling the schema those inputs must follow. Because it needs to generate and consider a huge number of possible inputs, it takes a long time to run this query. When it finishes, it turns out there is only one model. You look at what values the inputs have in that model, and it turns out that only three inputs are relevant, and only when they have specific values set. Later, you want to check whether this property holds in a specific fact scenario. So instead of adding a model of that specific fact scenario to the program and running the query again, which will be faster, but still slow, you can write a new program that only checks the values of those three relevant inputs, which will be blistering fast.