http://www.planetary.org/blogs/emily-lakdawalla/2015/03121716-ganymede-ocean.html It's interesting how they infer ice layer from rotation rate and outer shape and density. And then solve some equations to see if (within the ice layer) there could be a liquid ocean of a certain depth below surface, a certain salinity etc. Heating by tidal friction matters, of course. Ceres is estimated to have an icy mantel something like 100 km thick (under a thin rubble crust) but no layer of ocean. It orbits the Sun by itself and isn't subject to tidal heating. The ice mantle inference goes like this. Determine the size by inspection and the mass by observing interaction with other objects. If the density is substantially less than rock and it looks compact then there's water in its composition---calculate how much. Now the question is, is it ice mixed with rock etc, or is it DIFFERENTIATED into layers. That question is answered by comparing the FIGURE or shape (how flattened or oblate the ball is) with how fast it is rotating. The less differentiated, the more mass is farther out from the axis, experiencing more centrifugal force. The faster the spin, and the less differentiated, the more oblate, i.e. the more bulge at the equator. So if it is spinning relatively fast and still not terribly oblate, it must be differentiated.