In any given week, there will be a limited number of songs. That's the idea behind format radio. But each week they rotate out some songs and rotate in some songs. If they add too many songs, and that means people have to wait too long to get to one of their favorites. That's what kills stations with big playlists. I've given examples in the KITS thread about the kinds of songs by Echo & The Bunnymen or other bands that demonstrate the song selection at this station. There is no one single station that will be able to cover all the possible choices available.
Maybe this is a good moment to repeat, mostly for the passionate or interested radio listeners, the facts on how a playlist is created.
Within a station's format a list is made of all the possible song that might "fit the format" and then, usually, a few "what if" songs to measure the outer limits of the format. Hooks
* of those songs are played for P1/P2 listeners.
Those selected listeners for a music test are picked within the strongest core of a station, not broad fringes. If a station is highly ranked 25-54, that is too broad. So we'd usually find the median age, and test +/- 8 to 10 years on either side... let's say "28 to 45" or something like that. Often we emphasize the younger side as that is where we have to "invite new listeners in". And often we don't go to the highest age, as we know we have that group already.
Results are tabulated and we look at gender (if a stations needs both men and women in a balance), age ranges, ethnicities and the like to see subsets. Often a song does well in parts of our target, but does horribly and negatively in another part; we can't play those songs. To be on the playlist, a song has to be at least a good "like" in every important subset while getting significant "love it" scores in other age/gender cells.
For example, an AC station that targets women might test men as well. In that case, some songs may be weak (scores from 50 to 70 on a scale of 1-100) with men, but really strong with women. We might play those songs, but make sure that no more than one per sweep or even one per hour play on the air.
We also may find some songs do better within our older core demo segment. So we make sure that they are separated in airplay from other songs that are not as strong in younger core demos. We may also find that a song is liked, but is overexposed or "burnt" so we make sure that it rotates slowly... slower than its score would justify... so we don't burn it to a crisp.
Despite what stations may say on the air about "playing what we want", everything is measured and balanced. The idea is that each sweep or set will contain some songs any individual listener totally loves, and other songs that they like and, at worst, maybe one song that is "take it or leave it... I don't care much about that one". Each sweep has to be as strong as every other one, and this is done with scheduling rules that force this kind of consistency and balance.
The idea that MusicMaster and Selector just do "iPod on shuffle" programming is as far from the truth as Beijing is from Minneapolis.
* "Hook" is about 8 seconds of the key part of a song. When testing is done online or in person with dials, station listeners generally score each hook within 4 to 5 seconds... in fact, often withing a second or two! After around 6 or 7 seconds, nobody changes their score.