Pick a responsible number of points on the feature to examine. Establish control limits as necessary in X,Y,& Z for each point location (or all point locations if you group them into one control study)
Note: Control limits have nothing to do with specification limits.

After a sufficient number of instances have been measured and it is determined that the process is "in-control" figure out what size tolerance zone is required to contain each point relative to its expected surface normal vector (each point will yield a different zone). Find the displacement of each point from the center of its specified tolerance boundaries. If the specification is equal bilateral subtract displacement from the basic profile, if unilateral or unequal bilateral compare each point from the center of its tolerance boundary. Double the deviation to reflect the size of the tolerance zone required to contain each point. Compare the mean and associated distribution of those zones to the upper specification limit. Always check the shape of the distribution and apply the normal or best-fitting non-normal model or transform the data to reflect a normal distribution but don't forget to transform the upper specification as well.

If the ratio between (Tolerance Consumed and Tolerance Specified) is typically low the process control parameter can be the consumed tolerance itself rather than scrutinizing the coordinates. If that ratio is high it may be better to monitor the coordinates of the points individually to identify the source of variation and improve if possible.