This metric says something about how well was tested during the development phase. Any defects from production indicate that Software Testing missed something.
Obviously not to lay the blame on Software Testing if something was found in production, but to find out how we can prevent this next time. This could be in the basis of the process (e.g. unclear request by Product Management).
How to measure
Divide the number of defects found in production (e.g. with a specific tag) for a given release with the total number of defects found for that release. Excluded are the non defects.
Clearly we strive for 0%, but on the other hand if nothing comes back from the field; how good is our software used then?
Learn from production errors to increase coverage (e.g. extend test in certain areas)
Gain a larger coverage by using other parts of the company for testing as acceptance testers
Limit the number of combinations possible (e.g. only support win 7 and specific video cards)