The orthogonal variability model defines variability in a separate model to other domain artefacts.
As such it can be linked to all types of artefacts. This can be contrasted with other methods of variability management, such as FODA, where variability is mixed into the feature model. Benefits of the OVM approach are that variability does not get spread across models, and already complex models do not get overloaded with variability concepts.
Basically, you model the variability outside of all your other models, and link it into them via artefact dependencies. I like this idea, as you can use your one variability model with all your different models.
The core concepts are variation points and variants, variability dependencies, variability constraints, and artefact dependencies.