Looker models should be as clean and obvious as possible to our end users. We want to make sure that there aren’t redundancies between explores. We want the maximum amount of data reachable in the fewest amount of explores.
This brings us back to the original question, which view should we use for our explores? Our general rule of thumb is that once every potential entity is joined to the original FROM statement, the FROM with the most records is most likely our base. For example, suppose we have a User and Orders table. Not all Users have orders, so we will have more records starting from Users than Orders.
Unfortunately, not all data is conveniently set up with a super set containing all possible elements. For example, we could have Orders without any Users(maybe our store sells items anonymously). When this is the case, we may have to expose both explores and train our users how to pick which one to answer different questions. We could also stick to a single explore by UNIONing both Orders and Users into the same table and then joining both the entities on both sides. This method allows us to reduce the number of explores, but will also require training to deal with the complexity of these entities.
Also if your data is quite complex and these rules do not clearly match your schema, feel free to reach out to us at Looker! Chances are that we have already encountered your use case or have some insights that can help you determine the best view.