Portfolio selection, not surprisingly, depends on the risk and returns characteristics sought by the candidate.
Perhaps the easiest example is the prototypical "hard core" candidate. Let's suppose she has 3.5 or so from a major university, a GMAT in the low 700s, about 4 years at a major MC/IB/or MNC with good progression, strong references, etc. In her case we might structure a portfolio with a few ultra elite MBA programs, a few elites, and perhaps a trans elite.
Now suppose a candidate has a similar profile to the student supra but is seeking some form of scholarship or merit aid. In this case, we would "tilt" the cluster allocation to reach schools that are more likely to give merit aid.
Now consider a prototypical young applicant- strong academic factors, strong but somewhat nontraditional references, but little conventional work experience. This is a more difficult problem to optimize since many of the schools that are most receptive to younger applicants are at opposite ends of the clustering spectrum. Some young applicants choose the high risk/high return route- apply only to the top clusters, if the are not admitted they simply apply again in the future. Other young applicants use the "balanced" approach- apply throughout the cluster spectrum with the intent of the school in the highest cluster that accepts them.
Does this help?
Hjort