We try to measure how successful work is, primarily, by the making or missing of team commitments, but also through retrospectives where we determine if the team has learned or grown through the experience. We try to always help teams be as “self organising” as we can.
Thus a team with a Snr / Jnr pairing would take on a user story they both agree on. Through a mentoring approach the story is completed in an acceptable time, the work is committed, tested and deployed. In retrospective(s) the team has a look to see if the team has grown through the pairing and if the jnr is catching up in an acceptable manner. Both of these need to occur for the outcome to be favorable.
If the code is completed super quick, but the jnr is not learning anything then something isn't right and we’re loosing growth of a resource (sounds bad to put it that way I know). In the same way if the jnr is growing happily, but the code takes forever to complete then we are loosing business value and again something is wrong.
The “mentoring approach” is the difficult part and it’s usually the make or break whether the pairing is working.