23 May 2007

Google Trends - Part 2 - American Idol

OK, I've been tangentially aware of American Idol for the past few years and have seen it a few times so I'm not interested in getting on a soapbox and critique "reality TV" or offer commentary on the competitors. What I am interested in, however, is how the recent rise in popularity in Google's Trends tracking, could act as a predictive analysis tool, aimed at larger...things....

So...American Idol has already aired (presumably concluded) for the night here on the east coast of the US so I'll fill-in a bit of background for the uninitiated (like myself, really): after some recurring period of group talent contests, two contestants are left in competition, after some number have been voted off of the show by popular opinion, one week at a time. The winner of the contest will be decided by popular opinion as well.

And...the winner is...huh, nothing posted to the official American Idol site. I guess I'll post more about Google Trends and predictive analysis as soon as I have some sort of official results to go with for verification.

Shame...I was getting interested but now tired as the night progresses. Sigh....


OK, Blake Lewis and Jordin Sparks were the final two contestants on the show and Ms. Sparks won by popular vote. What happens if we plug both of their names into Google Trends? Well, currently, the results are disappointing as Google Trends is still a beta/lab product and there isn't any current, searchable trend data for yesterday. Let's see if there is anything to be gleaned from the 100 hottest searches for 23 May 2007.

OK, a search for "jordin sparks" was the 12th most popular search on the day of the show's season finale with search activity peaking around 2100 EDT. "blake lewis" is nowhere to be found in the top 100. Does this mean anything as far as being a potentially useful prediction tool? Is it more or less significant that guest performers on the show - Smokey Robinson (#1), Gladys Knight (#5), Tony Bennett (#6), Bette Midler (#7) and Joe Perry (#10) - were more popular searches, between 2000-2100 EDT, than Jordin Sparks?

Perhaps Ms. Sparks became the clear popular favorite during the day. Let's see what the previous day had for searches related to Ms. Sparks and Mr. Lewis.


The closest search was for "american idol winner" (#9 on 22 May and #2 on 23 May). So, maybe search trends aren't terribly useful for predictive analysis. Maybe there isn't enough data available at the moment. Let's see who the finalists were last season and plug them into Google Trends...

Taylor Hicks won the season 5 contest and Katherine McPhee was runner-up on the 24 May 2006 season finale. The Google Trends results appear to clearly support Mr. Hicks the days leading into the competition:

"taylor hicks" "katherine mcphee"

Let's look at season 4 for the fun of it. Carrie Underwood won the season 4 contest and Bo Bice was runner-up on the 25 May 2005 season finale. Shaking the Google Magic 8 Ball shows...nothing useful:

"carrie underwood" "bo bice"

They're tied until the next day when everyone is trying to find out more about the winner. Having access to the vote tallies could describe this if the popular count goes nearly 50%-50% for the contestants.

Might as well look at the 2004 season since Google has the data to play with. On the 26 May 2005 season finale, Fantasia Barrino won over Diana DeGarmo. Reading the Google tea leaves shows...more nothing.

"fantasia barrino" "diana degarmo"

If anything, Ms. DeGarmo may have had a slight popular lead entering the vote.

Have we learned anything? No, probably not; at least not from a predictive analysis point of view. If I think of it in the coming months I'll come back and see if Google has search/trend data available to cover this most recent season.

No comments: