The hierarchy used throughout this site is: conservation first, restoration second, repair last. Conservation stabilises the object as found; restoration returns it toward an earlier state, accepting some loss of later evidence; repair makes it work, accepting the most change. The right level depends on the object's significance, and the default for anything historically important is the most conservative option that achieves the goal.
Reversibility is the working test. An intervention that can be undone without trace — a removable support, a documented lubricant — is preferable to one that cannot. Irreversible work (polishing a case, refinishing a dial, replacing a wheel with a modern cutting) should require a positive justification, not merely be the easiest route.
Original material has evidential value beyond its function. A scored pivot records how the watch was used and serviced; a refinished dial erases the factory's hand. This is why 'as found' condition often outranks 'restored to new' in both scholarship and the serious market: the object is a document, not just a machine.
Documentation is part of the treatment. Photograph before, during and after; record every part removed, replaced or altered; keep replaced parts with the object. A future researcher should be able to reconstruct exactly what was done — an undocumented restoration is indistinguishable from an unacknowledged one.
Finally, know the boundary of competence. Complex striking work, chronometers, enamel dials and anything of museum significance belong with specialists. The most skilled act of conservation an owner can perform is often deciding not to act, and choosing the right professional instead.