Software is the fast part. These take months, depend on people saying yes, and cannot be hurried by working harder. Nothing below has started.
Every reading needs review text, and we want it licensed, one agreement at a time, from outlets and archives who know we have it. The slower road on purpose. Scraping around a paywall would poison the thing at the root.
A roster of working critics willing to confirm what they actually wrote, so that calling a review bought is something we can prove rather than infer.
One public reading, every discard visible, held up against everyone who disagrees with it. If the method cannot take that in the open, it is not a method.
First revenue, from someone who needs to know whether the wave that hit their film was real. One customer is the whole difference between a project and a business.
The one that takes longest and cannot be bought. It happens after the readings are good enough that ignoring them costs more than answering the phone.
Nobody yet, and we are not going to pretend otherwise. When the readings are worth something, two groups can pay: publishers and curators who want the filtered feed rather than a scraper, and independent distributors who need to know whether a wave was real.
Not advertising. An ad-funded score has an advertiser, and an advertiser eventually has an opinion about the score.
A studio can never pay us anything about a film we are scoring. Not for placement, not for early access, not for a report. That one is permanent.