After reading up on cardinality, another couple of questions occurred to me:
- Does the density/distribution of salesperson-region relationships have high cardinality (salespeople strongly associated with one or two regions), or low (salespeople working across many regions)?
- Does the density of those relationships better match the old optimizer assumption, or the new one (exponential)?
In that case, an alternative might be to plot three densities: your empirical density, the previously-assumed density (uniform?), and the new density (exponential?). There are a few examples here, but tons more on the web: https://heuristically.wordpress.com/2012/06/13/comparing-continuous-distributions-with-r/
After that, there are a whole slew of goodness-of-fit tests for checking whether your data fit a given distribution or not; off the top of my head, you could probably start fitdistr() from the MASS package.
No clue how you'd go about generating all of those densities, though!