There are generally two types of models that can be built using data: Surrogate models and System models.
Surrogate models usually involve a target (or dependent) variable and one or more predictors (or independent) variables. The mathematical relationship between the target and predictors is effectively termed the model. This math relationship is abstracted from available data.
System models do not usually have a single defined target. But there could be several targets called "outputs". The "inputs" are usually provided by users. A system model maps the relationship between inputs and outputs using well-established physical or economic laws. More often than not, some of these relationships are also based on reasonable assumptions.
In either case, models are needed to drive simulations. But a big caveat about modeling and simulation: "All models are wrong, some are useful". We agree with George Box and emphasize that models must be handled with care!