“Demons in horror movies can target people or be summoned,” Mr. Bruzzese said in a gravelly voice, by way of example. “If it’s a targeting demon, you are likely to have much higher opening-weekend sales than if it’s summoned. So get rid of that Ouija Board scene.”Oh, if only THE EXORCIST, HELLRAISER and HELLBOY had involved targeting demons rather than summoned demons.
But let's take the idiocies one at a time.
First of all, the company tells you how similar your movie is to successful movies, and assumes that's a good prediction of how successful your movie is. The problem is, the more similar your movie is to IRON MAN, the more people are going to think you ripped off IRON MAN. Unless you are bringing a new twist, I tend to think that the closer you are to recent successful movies, the less successful you'll be.
Second, as fans of Nate Silver know, there's such a thing as data overfitting. If you analyze a data set with enough factors, you can come up with all sorts of correlations, of the nature of "candidates from cities with winning football teams never win the presidency" or "always win the presidency." Unless you can explain exactly why a targeting demon is better for box office than a summoned demon, I'm going to assume you have the box office equivalent of a cancer cluster: a meaningless correlation.
Third, the data set is not really big enough or precise enough to draw conclusions of. How many movies have targeting demons, anyway? Probably not that many. In a small data set, one big flop or one big hit can change everything. Prior to RETURN OF THE JEDI, you might have guessed that having little teddy bear creatures fight gigantic killing machines with stone age weapons would not be indicative of a successful movie. Now, of course, it's a guarantee of >$400M box office.
This isn't Big Data. It's Small Data.
And finally, you can write a stupid movie with all the right elements and have it flop. And you can write a ground-breaking movie that proves the conventional wisdom wrong. All the models in the world will not predict THE FULL MONTY or THE BLAIR WITCH PROJECT or, for that matter, STAR WARS. Remember, at the time A NEW HOPE came out, there hadn't been a hit science fiction movie in years, and the studio was convinced they had a flop. Built on hindsight, Mr. Bruzzese's analysis would have told them that STAR WARS was a losing proposition.
And SNOW WHITE. When that came out, there had never been a successful full length animated movie. So, obviously, SNOW WHITE was a terrible idea.
And TOY STORY. Big animated movies were done for, right?
What this really is all about is Cover Your Ass behavior. Studio execs don't like to be responsible for big, expensive flops, because it gets them canned. If they can spend $20,000 of the studio's money on a former perfessor's analysis that says it's a great script, then when the movie flops (as movies often do), they can say, "Hey, how was I to know it was gonna flop? The perfessor said it was a sure thing!"
This is similar to why studio execs hire overpriced stars. Many analysts have run the numbers and movies without big stars tend to make more profits -- because big stars get gross participation and it's hard for the studio to break even. But studio execs keep hiring Tom Cruise. Why? Because you can always say, "How was I supposed to know OBLIVION would flop? I got you Tom Cruise!" And then, maybe, they don't get fired for poor judgment.
(I have no idea if Oblivion flopped, or made back its money overseas, or what. Replace OBLIVION with WATERWORLD, if you like.)
Let's not confuse Mr. Bruzzese with actual metrics and data crunching. I am all for putting 20 civilians in a room and hooking them up to video game style sensors, showing them the movie, and building heat maps, and determining when they're excited and when they're bored from their skin galvanic response. That's Real Data. Like any tool it can be used well or stupidly, but it has the potential of being used well.