Will music recommendation tools ever account for the diverse ways people discover and embrace music? In a post at Rhapsody's blog, Rhapsody VP of music programming Tim Quirk cites concrete metrics from BigChampagne to show how unlimited access to music can reveal the wide range of a person's tastes.

As Quirk explained, BigChampagne has created "tracks per fan" (TPF) and "artist correlation" (AC) metrics to show the depth and diversity of consumers' digital music collections. TPF counts how many songs by a unique artist are in an average person's collection. A TPF of 6.0 means a person has six tracks by that artist. AC starts with an artist and measures what other artists are in a person's collection. So in relation to Band A, an AC score of 60% for Band B means 60% of people with Band B in his or her collection also has Band B.

The TPF scores given in the post show the average person's digital music library is not built on the album format. Only the No. 1 ranked TPF artist -- Lil Wayne -- has a TPF over 10. At 17.26, Lil Wayne is far above the second through fourth artists -- Gucci Mane, Aventura and the Beatles -- who range from 8.44 to 8.22. So for every person with a full album or two, there are many more people with just a track or two.

The list gets interesting around No. 500: '70s Welsh prog rock band Man has at TPF of 1.38 and '80s supergroup Asia is 1.37.

But it's the AC metric, which reveals the diversity of a consumer's music collection, which is most revealing. Quirk offers Guns N' Roses as an example. The most highly correlated artist is Eminem at 67.08%. That means 67% of people with Guns N' Roses in their digital music libraries also have Eminem tracks. The AC list includes many hard rock, heavy metal and classic rock bands. But it also has country stars Tim McGraw, Garth Brooks and Kenny Chesney. The Black Eyed Peas, which is hardly a rock or country act, is No. 2 with a 58.9% AC score. In other words, popular music is popular across much of a populace.

Those AC scores show that the contents of a person's library are at odds not only with terrestrial radio, but also with personalized radio and online recommendation engines. Clearly there are reasons other than genre or scene that a person likes a particular artist. Yet recommendation engines don't account for this cross-genre interconnectivity. They treat people as being far more predictable than the Guns N' Roses-Tim McGraw relationship would attest.

As a result, consumers' diverse tastes may not be addressed by many of today's digital technologies. If you type "Guns N' Roses" into Pandora, will it stream Black Eyed Peas or Tim McGraw? After reading Quirk's post, I started a Guns N' Roses station at Pandora. The band's song "November Rain" was followed by Bon Jovi's "You Give Love A Bad Name," AC/DC's "Back in Black" and Metallica's "Turn the Page." The latter three bands have the fifth, sixth and seventh-best AC scores for Guns N' Roses. Yet Eminem and Black Eyed Peas, neither of which is likely to be played on that channel, scored higher than all three.

Quirk offers a few more examples of how radio can fail to capture fans' diverse interests: 38% of Beck fans listen to Johnny Cash and Eric Clapton; 73% of people with R.E.M. on their device also have at least one T.I. track, and 27% have at least one by Phil Collins; and 31% of Barry Manilow fans and 30% of Neil Diamond fans have R.E.M. songs.

Without taking these ranges of interest into account, personalized radio will struggle to match iTunes on shuffle.