Tales from the Jugular
Can a Computer Program Tell You a Hit?
By: Frank Hill
Published: Wednesday, November 12, 2003
Programming Hit Songs
A company called Polyphonic HMI has finally done what has been traditionally left up to
the ear of trained marketing executives and trend-seeking club watchers. They've developed
a program to tell you if a song will be a hit or not.
It was really only a matter of time before it was created and I'm surprised it took this
long because music, like a lot of human behavior, is fairly pattern-oriented. Their
system, as far as I can tell, just looks at all the similarities of thousands of hit songs
and notes all the successful patterns. A new song can then be mathematically compared to
recent hits to see if it has the characteristics of an emerging trend or if it matches
past hits.
Using the content of over 250,000 CDs:
"The characteristics we measure have been identified in user
testing to be the ones that produced the strongest reaction in testers. Our HSS (Hit Song
Science) technology takes the analyzed data and overlays extra parameters relating to the
commercial success of the music...such as total sales, highest chart position, date of
release and others."
A lot of musicians do this subconsciously just from listening to so much music, but the
Polyphonic's Hit Song Science (HSS) supposedly turns the subjective into objective data
that can be used in reports.
Labels will be able to use the reporting information by Polyphonic HMI as a tool to
apparently cut the risk on new song releases. It's an identification system that can be
used musicians to target audiences.
Individuals will be able to use "similar song" links and we could all wear our
musical profiles on our sleeves.
How would you like it if a record company suit told your band that they liked you guys
personally, but the database report said that your songs stink and won't sell? How would
you like it if your songs were critiqued by a machine and potentially never heard by a
human ear?
It does raise some other questions:
--How effective are the parameters they've chosen for hit songs?
--How much of our musical interests are really trend based?
--How many trends were manufactured by companies and how many were generated from
grassroots music lovers?
--How much music is really sold based on the charisma or dress of the performer?
--How much great music fits well in a full album format but not in a singles format?
--How often does a great song fail for reasons that are within or without trends?
--What about bands that do well live, but rarely have charted hits?
--How many hit songs are hits only because they were played ad nausea till we purchased
them like lemmings only to wonder a month later why the hell we bought the damn thing in
the first place.
--How long will it be before the machines are programmed to make the hit music themselves?
It may not just be the stuff of anime.
--They do note that executives will only use HSS as a tool to go with the educated ear,
but if the tool is so good, how many will loose their jobs?
Supposedly they predicted Norah Jones and are now putting their money on a band named
Elcodrive. I wanted to get the opinions of our metal staff with the questions:
Do you think metal labels could use this system effectively?
Do you think this technology could help metal bands create better hits?
Links:
http://www.polyphonichmi.com/index.html
http://www.polyphonichmi.com/hssfaq.html
*******
Frank Hill:
I'd say that this program could work well in the pop market where a lot less is expected
as far as innovation and songs are similar. I mean, it's "popular" music which
is supposed to appeal to a mass audience, right? Take it or leave it; it is what it is.
If you run your company from a money standpoint, why not try it? All you're looking for are the set of
common denominators that will sell to the masses. If you look at it with an artistic eye, though,
some hits do stand the test of time as more than just 4 minutes to pass the time. I would think that the
company that shoots for "Revolution" would do better in the long run than the ones that look for versions of
"Oops! I Did It Again".
Some hard rock singles are structured like pop singles, but you have edgier guitars and
vocals, so it may work a bit with those. A lot of metal songs are way different though.
Take songs like "Master of Puppets" or "Hallowed Be Thy Name" which
are of the best the genre has produced. Anybody who makes a song fairly close to those
gets torn up by the fans. On the other hand, one could argue that those are upper echelon
songs and that metal mostly moves in popular trends also--Power, NWOBHM, Thrash, Death,
Nu--that become as played out as any pop movement.
Of note to the programing, I'd say most of the best metal songs never saw the light of any
popular chart. Metal singles rarely are released and don't seem to chart well, but metal
albums will hit #1 on the charts.
They do say that trends can be predicted, but not knowing what the programming is for the
program, I would assume that they would have to use an initial hit song first and then
gauge the reactions to it. If that's the case, then isn't it really just playing Follow
the Leader.
If any metal companies are using this system now, drop me a message, please.
MM:
They might be able to, but I can't remember the words "metal" and
"hit" ever going together, save for "Nu Metal". Korn is about the
closest real metal has come to producing a hit. Linkin Park, Staind, limpbizkit, et al.
are more hard rock, I would think. Some metal labels have already generated hits; look at
Roadrunner and Nickelback (on a more modest scale, they also have Slipknot). Interscope
has done well with limpbizki, also.
They might be able to use the system pretty well, but would artists as driven and
dedicated to their music want to rely on a machine to tell them if their record's going to
sell? Would they care? Granted, they should if they want to make any sort of living at
doing what they love for a living, but to many bands, making hits and being stars are not
as important to them as it is to the labels. I would think it would be a matter of getting
the artist to agree to the process, and then the program could probably be used well to
make even heavy metal a viable commercial product.
Leaving it to a computer to predict the tastes of an entire consumer populace is a risky
venture. People do tend to go in for manufactured trends (many and various "boy
bands", American Idol, Britney Spears, etc.), but people also get tired of having
their tastes dictated to them. This is why many "one-hit wonders" become one-hit
wonders - the first single is a smash, the label, MTV, and the radio outlets go into
overdrive playing and promoting the record, and people get tired of it.
They're ready for something and someone new. Tastes are fickle, and people are choosy as
to what they spend their disposable income on. Many big names in music are no longer big
names because the public grew tired of them. And if that was of their own accord, you
could imagine what having a machine crank out much of the same product would get very
tiring to people after a while.
And many things become hits that obviously were not meant to. Pink Floyd's "Dark Side
Of The Moon" stayed in the Billboard top 200 charts for over 20 years despite the
fact that there was not one commercially viable song on the whole album. The general
public may be seen as suckers in the eyes of many in the music industry, but I don't think
they're as stupid as many executives think they are sometimes.
[Other Maximum Metal Columns]
<< back >>