K-Earth Holding Its Own For September....

[oldies76]

and if that means playing the Macarena every hour, then so be it.

Now you really want to turn off the audience.....hehe!

 

[oldies76]

Despite the tight playlist, anything above 6.0 indicates the station is doing really well. Let's see how long KRTH can keep it up!
A 6.5 rating is pretty big compared to the Coffey days when the station was in the 3.0's, maybe a 3.3 or 3.2.

Don't even remind me of those days. What was that.....a select Motown song every 15 minutes, 3 Beatles tunes an hour, few 70's and a playlist so tight, Jhani Kaye had to rescue them. And that song from the movie Ghost, played to the moon!

Jay Coffey was a fine, personal and knowledgeable DJ back when they were known as "Classic Rock & Roll", circa 1987, back when you had Lisa Moree, Mr. Rock & Roll, Steve Scott, Bernadette and Jonathan Doll. Great times!!

 

[oldies76]

That is so far from how a music test is implemented that it is not even amusing.

You missed the obvious humor....

 

[davideduardo]

You missed the obvious humor....

Coming from you, it was more than likely to be deadly serious.

 

[ChannelFlipper]

Cluster or Factor analysis is a process that looks for commonalities among non-standard groups. That means using math to see if all your listeners can be divided by factors other than just age and gender. What this is like is perhaps best compared with a museum of art; some will favor modern art, some the impressionists and other the classic styles. They all like museums (your station) but they have a preference for certain subsets. That does not mean that someone who like Picasso does not like Monet or Michelangelo, it just means that they favor one subset.

If you use a math process to find those subsets, then you can better balance your hours and sweeps. In fact, you can find subsets of subsets... liking impressionist landscapes but not portraits. Or liking most impressionists but not pointillists. Knowing this about your music allows for different methods of flow and protections from "two like songs" back to back when the difference is not style or gender but appeal to a specific "cluster" or group.

This is tedious, but just think about "songs" when you read the first paragraph or so. A music test has data on each participant and mathematically we can find similar patterns among other participants. That is a cluster. It can be big (worth programming to) or a little subset (to be ignored). And it mathematically explains that there are similar groups of people who have very different taste.

It can get very complicated... in alternative rock, the clusters are very polarized to the extent that you say, "can I get them all to listen to the same station?" and that is the challenge. Cluster analysis makes sure you appeal to the strongest subsets and allows balancing so you don't piss off one of them in every sweep!

Now you are bringing the meat! I always suspected they did something like "clustering" but would not have known that was the term they used or how extensively they used it. I was always certain that if there are 350 songs in the rotation, whenever they play song #301-350 (or maybe it is something like 251-350) then they will always play a top tier song on either side of it, and for sure those bottom songs can never be represented more than once in a set, probably never more than twice an hour. Doing the cross-sections just adds even more nuance. I use data analytics in my profession too, but it is a tad more boring.

Every hour, most music stations have somewhere between 10 and 14 songs to fill the programming time. Each day has between 16-18 sell-able hours (maybe less on weekends). How that time is used, specifically, how those songs are picked is one of the most fascinating parts of the industry.

 

[davideduardo]

Now you are bringing the meat! I always suspected they did something like "clustering" but would not have known that was the term they used or how extensively they used it. I was always certain that if there are 350 songs in the rotation, whenever they play song #301-350 (or maybe it is something like 251-350) then they will always play a top tier song on either side of it, and for sure those bottom songs can never be represented more than once in a set, probably never more than twice an hour. Doing the cross-sections just adds even more nuance. I use data analytics in my profession too, but it is a tad more boring.

Every hour, most music stations have somewhere between 10 and 14 songs to fill the programming time. Each day has between 16-18 sell-able hours (maybe less on weekends). How that time is used, specifically, how those songs are picked is one of the most fascinating parts of the industry.

Almost all buys are now 6 AM to 7 PM, with the weekends at about 50% of the weekday sell. Most of what you hear filling stopsets at night and weekends is value added stuff or part of a package where, really, the nights and weekends are free.

Usually songs are divided by tiers as there is a significant margin of error in song scores... if you test on a rainy night you get a different average score than on a nice weather evening (of course, online testing is in some ways oblivious to this but also has hidden risks when you do not see the person face-to-face). So you put the songs in batches by age and by test scores. You might have several gold categories, one with overall stronger scores and others with weaker averages or defects among one sub-group. For those, you use codes to keep each kind separate: weak on younger women, weak on older women, weak on younger men, weak on older men, weak on heaviest listeners, etc. All those codes let the stronger songs pick their partners for fit before and after. The strong songs are scheduled first, and then the others are done to best fill in the other clock positions. And, of course, at the end the PD will manually massage the log if something just looks like a bad flow or segue.

 

[ChannelFlipper]

Almost all buys are now 6 AM to 7 PM, with the weekends at about 50% of the weekday sell. Most of what you hear filling stopsets at night and weekends is value added stuff or part of a package where, really, the nights and weekends are free.

Usually songs are divided by tiers as there is a significant margin of error in song scores... if you test on a rainy night you get a different average score than on a nice weather evening (of course, online testing is in some ways oblivious to this but also has hidden risks when you do not see the person face-to-face). So you put the songs in batches by age and by test scores. You might have several gold categories, one with overall stronger scores and others with weaker averages or defects among one sub-group. For those, you use codes to keep each kind separate: weak on younger women, weak on older women, weak on younger men, weak on older men, weak on heaviest listeners, etc. All those codes let the stronger songs pick their partners for fit before and after. The strong songs are scheduled first, and then the others are done to best fill in the other clock positions. And, of course, at the end the PD will manually massage the log if something just looks like a bad flow or segue.

It is ironic that obscure hit "Falling" by LeBlanc and Carr has had a lot of comments on this site lately , because back in the day I used it for a radio geek's "test" if you will. KOST used to have several AC competitors come and go who always had wider playlists and not as much focus (which is why I always liked them better, but I digress). It was around 1989 or so, and one of their competitors - I think it was KXEZ on 98.7, but I could be mistaken - was clearly going to be flipping formats. Nothing had been announced, but I could see it coming because the commercial inventory was down and they had opened up the playlist even more than usual and one day, the mid day DJ subtlely encouraged listeners to request a song. I hopped on the line and requested "Falling" knowing that it was not on even their expanded playlist. I was surprised it was the actual DJ who picked up the phone and when I made the request he paused, not knowing quite what to do. After a few seconds, and without putting me on hold or himself on mute so I can hear everything, he asked the program manager if he could play it. Program manager also paused a few seconds and says, something like, "ah what the hell, play it". Within 20 minutes it was on the air. I couldn't believe it.

And sure enough, about two weeks later, they were off the air.