We reach this conclusion because we believe in causality: certain physical processes are observed to happen always in a certain order, never otherwise. For example, I have never eaten boiled eggs in my existence on this planet without my having to cook them first.
Before I go further, the correct rendering of your statement is that no causal link from event $A$ to event $B$ can lie outside $A$'s future lightcone, often stated "no signal can propagate faster than $c$". See my answer heremy answer here where I show that certain processes can travel at faster than $c$ if there is no causal relationship between the events within the process. The travel of a material object you refer to is just a special case of a process whose events are causally linked.
Now to the rest of my answer: we postulate that the order between my cooking my eggs and my eating them is not changed when one observes these processes from relatively moving frames: there is no inertial observer moving relatively to me whose motion causes them to see me eating my eggs before I cook them. This is a crucial point once one derives the Lorentz transformation, because the Lorentz transformation shows that there is a relativity of simulteneity. Different, relatively moving observers do indeed observe different before/after time relationships between certain events. So how do we save our notions of causality?
We postulate that causal links cannot travel faster than $c$. For example, the link between my cooking my eggs and my eating them is (1) directed forwards in my time (pointing to my future) and (2) lies inside my future light cone. As long as this inside-my-future-lightcone condition is fulfilled, the basic properties of the Lorentz group guarantee that a boost cannot change the order of events linked by such relationships, even though it does change the time and space co-ordinates of those events.
There are other ways to motivate no-faster-than-light travel: for example, that the energy required to accelerate something with nonzero restmass to $c$ would be infinite. However, the upholding of causality is the overarching and most forceful motivation: there is no way to uphold the observed orders of events within a process if any pair of events on that process's worldline lie outside each others' future/past lightcone. If we believe in causality, no faster than light signalling is an inescapable conclusion.
See my answer heremy answer here and herehere where I show how the basic order-preservation properties of a subluminal boost follow. Also this answerthis answer discusses causality in more detail.