There wasn't a ton of market research, iteration or testing, at least not up front. Executives would green-light a project and budget based on their personal experience and/or biases and opinions about what will work and what won't..
Not sure if that's still the case (this was ~10 year ago), but that rarely works in tech.
- Movies begin as risk assessments and cash flow analysis in the finance dept based on historical data.
- The revenue graph for a movie is a function of release month, the number of celebrity actors in the film, the box office receipts, followed by DVD sales, followed by TV release and a long tail of syndication (airlines, cable etc).
- Studios know, for example, that a romantic comedy in Feb staring Jennifer Aniston and Adam Sandler will make a guaranteed $35 million at the box office regardless of what the movie is. A summer blockbuster with Will Smith will have a guaranteed $100 million in box office and DVD sales spread out over time. Sequels to popular movies will bring in a certain minimum income guaranteed. Movies that minimize dialog and maximize action do well in the global market. Include China or a Chinese celebrity actor in the plotline and you are guaranteed millions of additional revenue. This provides the bean counters information on how much to spend on movie production to minimize risk and ensure a steady revenue stream throughout the year.
- So the finance dept will green light a February rom-com with a production budget of $10 million, a summer blockbuster where there can splurge $75-100 million on production etc. They monitor other studio release dates to avoid going head-to-head with their releases. They keep a stable of celebrity actors on studio contract and force directors to include plot lines that maximize a global audience.
- It is only after the finance dept gives the green light and budget that they go and look at possible movies to make and pick a director. Story ? - that's just about the last thing to work out (Pixar is the exception).
While there‘s a limited subset of AI like GANs that I‘d actually call creative, the usual tabular models are none of these.
So what do you do to maximise chances for a box office success?
Do a sequel to (or remake of) a well known movie, hire a star, use the standard playbook (never skip the „all is lost“ moment for dramatic effect) and add a love scene.
If you ever wondered why 2020 Hollywood feels like the same boring stuff in 8k UHD over and over again.
I think Netflix did pretty well with just hiring some known creative people and telling them „whatever you do, just use your gut feeling and we‘ll buy it“.
But I get your point, either extreme can potentially produce less than great results.
That's been going on since the first wagons rolled into Hollywood.
I'm pretty sure even Shakespeare was doing a 16th Century, theatrical equivalent.
That's your toolkit. Build from there. And that' been true since we're telling stories, more or less.
Those Netflix decisions about who to hire are extremely data-driven - much moreso than Hollywood. House of Cards was their proof of concept in 2013:
> Sarandos says their wealth of data on user viewing habits proved there's a large audience for Fincher, Spacey and political thrillers.
https://web.archive.org/web/20130126061225/http://www.huffin...
Really initial production is only half your production costs, and is a tricky process where what's greenlit often doesn't bear much relation to what comes out the other end. So the market research and metrics work better when you know a bit more about what the final product is.