Songs Everybody Likes: Feedback

In a blog posted in April ("What Do You Do With A Song That Everybody Likes") I related the story of how one station handled a song that test with no negatives, but which got few (5%) favorites.

This station felt that favorites equal passion, and they felt that the song could not be a primary song for the station. In other words, they decided that "FAVORITE" as a listener opinion about a song was worth far more than a simple "LIKE" opinion.

I asked readers of the blog to tell me how much more important FAVORITE is compared to LIKE, in their opinion. I thought I'd summarize the responses I got.

Tie Breaker

Most programmers seemed to treat FAVORITE as a tie breaker. If all other factors are the same, the the edge goes to the song with more FAVORITE entries as opposed to LIKE.

What is interesting is that nobody seemed to feel that FAVORITE is the primary criteria. Positive feelings (LIKE + FAVORITE) seems to be what programmers focus on.

Burn

One programmer offered the opinion that:

"..focusing on "favorites" ultimately will lead to burn, a great way to make a favorite less so."

Frankly, that seems like a valid concern to me.

Industry Standards

The standard in our industry, when creating an overall score for a song is to give a FAVORITE score 50% extra credit compared to LIKE. If anything, that may be slightly more aggressive than the consensus of the people who responded to my question. That is the amount of "extra credit" for FAVORITE that I use in my own music research analysis, and after testing various ideas since 1974, I've become comfortable with it.

Other researchers, particularly those using a 1-7 number scale, will be found doing one of two other things:

1. Combine the top 2 scores (6+7), and simply use that. I'm no fan of that, because I think it literally takes the 4 and 5 answers out of the equation altogether. And we don't really know what 6 and 7 mean. My guess is that this is almost like counting only FAVORITE. Worse, a "6" score might, for some people, simply mean LIKE IT. We have no way of knowing. Gathering opinions from people that you simply don't understand the meaning of seems unnecessarily primitive. We can do better.

2. Combine the scores according to their 1-7 value. So a 7 score gets 7/6 more value than a 6 does. If we think of a "5" as the center point of LIKE and "6.5" as the center of FAVORITE, then the ratio is 6.5/4.5 or 13/9. That is very close (slightly less) to the ratio normally used in a semantic scale like mine (anchored by words rather than a simple number rating). So that seems reasonable, if you use a number scale.

Importance to Music Test Analysis

This issue is important, when we seek to do accurate music research analysis. Even music "fit" analysis requires a smart decision about the importance of FAVORITE responses. In deciding how similar two songs are, we have to compare responses. If one person rates the song a "FAVORITE" and another person rates it as "LIKE", are they the same opinion? No. How different are the opinions of those two people? Whether ranking songs on popularity or on compatibility, our decisions about what opinions actually mean are what guide the computers that crunch the data.