Really great analysis. I have a few additional questions that I think might add some illumination.
1. Did the debut film make money? You have the variable of the debuts that made 2.5x increases is 1.64, but I'm curious if there's any additional details for any film that makes money.
AND then
2. I'd pair this with the market value of the genre. Are romances higher because they make more money? (More people see them? The mainstream likes them?) I don't know.
I found it interesting that the more people variable was more important than the good film variable (at least how I read your piece), that breaks my heart in a way, but I'm not surprised. A film that sells is more "valuable" because it makes money, not necessarily art.
Which points to the heart of all my questions: Capitalism. What gets made is what makes money. If you want to make art, you're probably not in it for the money.
Great analysis. I appreciate how you broke down the takeaways so that for all of us preparing to make our first feature, we can make decisions to increase our chances of making this a career (though I feel like the biggest thing we can do is ignore the odds and do the work)
Going forward from 2018, it seems more factors would need to be in the mix: social media presence, followers count, and youtube subscribers. Since filmmakers are bringing their online audiences to the multiplex, it would make sense that those audiences would give online creators more chances to direct movies by supporting their work.
Hey man, I really love your research and essays but I can't read the ones that are as AI written as this one is. The small fragments in other essays are jarring, but they're only around for a few sentences, but this one is chock full. I'm sad, because the piece is probably really cool! This is a fun question and the figures show some neat investigative methods. But right from the get-go, we see stuff like "and it's the measure running through all of today's research".
(Sorry if this comment is frowned upon; mobile Substack doesn't show me anything saying otherwise and it felt ~informationally rude to click off the post without saying why)
Hey Alex. I assure you it's not AI-written. I use a lot of AI to find, crunch, and analyse data (and to make the charts), but then I structure and write the articles. I then use spell and grammar checks via Grammarly, which may be what you're picking up? I'm quite dyslexic, so I assure you you would not prefer the raw, unchecked drafts!
Generative AI is definitely writing phrases that end up in these articles. If this really is a thing you are unaware of, some step in your writing process is injecting large amounts of AI text into your articles.
This one flags as 49% AI on Substack's pretty-accurate detector (the other 51% is human human).
Your articles range from "mostly you" to "about half generative AI" when I hit the detector button. I typically only press the button when I'm finding an article unusually odd to read. I don't know where the AI-ness comes from, but the heavy AI articles I typically give up on.
I'd be happy to use Pangram.com to pick 2,000 words to check against (this could be useful because it highlights the text with its guess at the author) if that would be helpful. I have seen it be the case before where an author is unaware that their editing process is replacing large swathes of words with AI output.
To clarify: I don't think it's a great crime. I just really like some of your articles and really dislike others, and the percentages that Substack makes available (only *after* I get this feeling) provide a really strong correlative case. I don't mean for this to be a witchhunt, I just really have seen authors get taken by surprise on this before. Maybe using the Substack AI percentage indicator and comparing some articles to drafts could clear it up? I don't mean this as homework, just help.
No worries, I appreciate the helpful spirit you're engaging with it. I would have thought that the length of the Notes section would correlate a lot with AI detection, as that gets much less love from me. The purpose of the Notes section is accuracy over readability or voice.
I write the way I would personally express it, and then Grammarly et al. come in with suggestions. I also use Claude to play 'red team' for my articles, as it has access to the raw data and can find out where I've gone wrong, either in the maths or the writing.
I tend to defer to them, unless the meaning has shifted. (For example, in the preceding pargraph I wrote "if it's causing consternation", but the suggestion was "if it's causing confusion", which is an entirely different meaning).
I don't see AI writing as any sin, but it is something I don't really want in my work. This site is highly personal, in that there is no-one else. I do the admin, research, writing, images etc. So the idea that it loses my voice is sad.
I was thinking about this a couple of weeks ago, when I did my second-ever guest post, sharing an article from Daniel Parris (and he did the same back to me). His article is great, but of course it's written in a different voice. I received a wide range of responses, some of which you can see in the comments section. Having an equally talented film data stats person write an article was enough of a heresy for some to speak up, both in comments and emails to me. (For the record, the supportive replies I got for it were twice the volume of the negative).
I am learning that AI is seaping in through every hole of our lives. Egh
As someone who has made two features this is really really interesting. I always told myself after the first that if I could just make another I’d put myself ahead of many others in my position. It’s cool seeing numbers to back that feeling I had.
Seems for me the positive factors in my favor were that I’m young (21 when I directed the first), I waited only 3 years between projects, and that I’m a man.
That said I had no real success with the first outside of the personal pride I felt at what we accomplished. Didn’t play any festival, is still sub 1k views on YouTube. To my knowledge nobody has ever watched it that I haven’t directly talked to about it. I had no major attachments or connections.
I always told myself with certainty that I’d make another regardless. The truth is that I got very lucky, at least in terms of being able to go again just three years later. I found a job that gave me tons of flexibility and enabled me to give myself to the project fully. I was fired mid way through production but at that point it was too late lol.
I guess my advice, keeping this data in mind, is to just make the next. The second you find yourself in a spot where you’re mentally and physically ready, don’t hesitate. I’m so grateful for the chance to make another film. It changed my life in so many ways and further allowed me to see myself as a filmmaker. Thanks for compiling this and good luck to anybody out there who’s between the first and second! Keep going.
Fascinating. One thought about documentary directors having the lowest rate of returning to make a second feature: I imagine they'd need a huge emotional motivation to invest years of their life creating a documentary about a subject they care deeply about, and I wonder if their low return rate is due to a trifecta of: a presumably low financial reward, the high level of bureaucracy and inherent legal risk of documentary filmmaking, and the low likelihood of finding another subject they feel equally passionate about.
Fabulous analysis! Especially since for many filmmakers, it is the second film or even third film that helps them to breakout (off the top of my head, I'm thinking of Spike Lee's Do The Right Thing, which was his third film). We fret so much on getting that first film done (which, fair, is always challenging) but it's usually the follow-up(s) that can make the difference between a long-tail career and a one-hit wonder.
This is a great read, thanks Stephen. I would be interested too, particularly looking at filmmakers from the UK, what success came to directors who had previously been directing in Advertising (as opposed to just Shorts and TV). Unlike many other countries which like to keep their types of directors separate, the UK has a bit of a tradition of directors coming up through making commercial films - namely the Adland 5 - and it looks to be the route to take these days for starting-out directors.
So, I the lesson here is to make two features instead of one debut. I'm mostly kidding. It feels like its both harder and easier to get seen now if you are able to put the time and resources into getting in front of audiences. And if your debut being seen is the biggest indicator, then there's something we can do to impact that outcome.
Really great analysis. I have a few additional questions that I think might add some illumination.
1. Did the debut film make money? You have the variable of the debuts that made 2.5x increases is 1.64, but I'm curious if there's any additional details for any film that makes money.
AND then
2. I'd pair this with the market value of the genre. Are romances higher because they make more money? (More people see them? The mainstream likes them?) I don't know.
I found it interesting that the more people variable was more important than the good film variable (at least how I read your piece), that breaks my heart in a way, but I'm not surprised. A film that sells is more "valuable" because it makes money, not necessarily art.
Which points to the heart of all my questions: Capitalism. What gets made is what makes money. If you want to make art, you're probably not in it for the money.
Nice work. Thanks.
Great analysis. I appreciate how you broke down the takeaways so that for all of us preparing to make our first feature, we can make decisions to increase our chances of making this a career (though I feel like the biggest thing we can do is ignore the odds and do the work)
Going forward from 2018, it seems more factors would need to be in the mix: social media presence, followers count, and youtube subscribers. Since filmmakers are bringing their online audiences to the multiplex, it would make sense that those audiences would give online creators more chances to direct movies by supporting their work.
Hey man, I really love your research and essays but I can't read the ones that are as AI written as this one is. The small fragments in other essays are jarring, but they're only around for a few sentences, but this one is chock full. I'm sad, because the piece is probably really cool! This is a fun question and the figures show some neat investigative methods. But right from the get-go, we see stuff like "and it's the measure running through all of today's research".
(Sorry if this comment is frowned upon; mobile Substack doesn't show me anything saying otherwise and it felt ~informationally rude to click off the post without saying why)
Hey Alex. I assure you it's not AI-written. I use a lot of AI to find, crunch, and analyse data (and to make the charts), but then I structure and write the articles. I then use spell and grammar checks via Grammarly, which may be what you're picking up? I'm quite dyslexic, so I assure you you would not prefer the raw, unchecked drafts!
Generative AI is definitely writing phrases that end up in these articles. If this really is a thing you are unaware of, some step in your writing process is injecting large amounts of AI text into your articles.
This one flags as 49% AI on Substack's pretty-accurate detector (the other 51% is human human).
Your articles range from "mostly you" to "about half generative AI" when I hit the detector button. I typically only press the button when I'm finding an article unusually odd to read. I don't know where the AI-ness comes from, but the heavy AI articles I typically give up on.
I'd be happy to use Pangram.com to pick 2,000 words to check against (this could be useful because it highlights the text with its guess at the author) if that would be helpful. I have seen it be the case before where an author is unaware that their editing process is replacing large swathes of words with AI output.
To clarify: I don't think it's a great crime. I just really like some of your articles and really dislike others, and the percentages that Substack makes available (only *after* I get this feeling) provide a really strong correlative case. I don't mean for this to be a witchhunt, I just really have seen authors get taken by surprise on this before. Maybe using the Substack AI percentage indicator and comparing some articles to drafts could clear it up? I don't mean this as homework, just help.
No worries, I appreciate the helpful spirit you're engaging with it. I would have thought that the length of the Notes section would correlate a lot with AI detection, as that gets much less love from me. The purpose of the Notes section is accuracy over readability or voice.
Here's a good example of what happens in the writing process, which I may need to be more careful about in future, if it's causing consternation. https://www.dropbox.com/scl/fi/uavvg3rk0hartm098zbw6/Screenshot-2026-10-05-at-21.33.12.png?rlkey=8yqzbx2ltokkan3nraab9c8z8&st=vorn8xsk&dl=0
I write the way I would personally express it, and then Grammarly et al. come in with suggestions. I also use Claude to play 'red team' for my articles, as it has access to the raw data and can find out where I've gone wrong, either in the maths or the writing.
I tend to defer to them, unless the meaning has shifted. (For example, in the preceding pargraph I wrote "if it's causing consternation", but the suggestion was "if it's causing confusion", which is an entirely different meaning).
I don't see AI writing as any sin, but it is something I don't really want in my work. This site is highly personal, in that there is no-one else. I do the admin, research, writing, images etc. So the idea that it loses my voice is sad.
I was thinking about this a couple of weeks ago, when I did my second-ever guest post, sharing an article from Daniel Parris (and he did the same back to me). His article is great, but of course it's written in a different voice. I received a wide range of responses, some of which you can see in the comments section. Having an equally talented film data stats person write an article was enough of a heresy for some to speak up, both in comments and emails to me. (For the record, the supportive replies I got for it were twice the volume of the negative).
I am learning that AI is seaping in through every hole of our lives. Egh
As someone who has made two features this is really really interesting. I always told myself after the first that if I could just make another I’d put myself ahead of many others in my position. It’s cool seeing numbers to back that feeling I had.
Seems for me the positive factors in my favor were that I’m young (21 when I directed the first), I waited only 3 years between projects, and that I’m a man.
That said I had no real success with the first outside of the personal pride I felt at what we accomplished. Didn’t play any festival, is still sub 1k views on YouTube. To my knowledge nobody has ever watched it that I haven’t directly talked to about it. I had no major attachments or connections.
I always told myself with certainty that I’d make another regardless. The truth is that I got very lucky, at least in terms of being able to go again just three years later. I found a job that gave me tons of flexibility and enabled me to give myself to the project fully. I was fired mid way through production but at that point it was too late lol.
I guess my advice, keeping this data in mind, is to just make the next. The second you find yourself in a spot where you’re mentally and physically ready, don’t hesitate. I’m so grateful for the chance to make another film. It changed my life in so many ways and further allowed me to see myself as a filmmaker. Thanks for compiling this and good luck to anybody out there who’s between the first and second! Keep going.
Fascinating. One thought about documentary directors having the lowest rate of returning to make a second feature: I imagine they'd need a huge emotional motivation to invest years of their life creating a documentary about a subject they care deeply about, and I wonder if their low return rate is due to a trifecta of: a presumably low financial reward, the high level of bureaucracy and inherent legal risk of documentary filmmaking, and the low likelihood of finding another subject they feel equally passionate about.
Fascinating. Great post!
this is amazing. that -0.83% comeback rate for female directors 😭
Fabulous analysis! Especially since for many filmmakers, it is the second film or even third film that helps them to breakout (off the top of my head, I'm thinking of Spike Lee's Do The Right Thing, which was his third film). We fret so much on getting that first film done (which, fair, is always challenging) but it's usually the follow-up(s) that can make the difference between a long-tail career and a one-hit wonder.
This is a great read, thanks Stephen. I would be interested too, particularly looking at filmmakers from the UK, what success came to directors who had previously been directing in Advertising (as opposed to just Shorts and TV). Unlike many other countries which like to keep their types of directors separate, the UK has a bit of a tradition of directors coming up through making commercial films - namely the Adland 5 - and it looks to be the route to take these days for starting-out directors.
So, I the lesson here is to make two features instead of one debut. I'm mostly kidding. It feels like its both harder and easier to get seen now if you are able to put the time and resources into getting in front of audiences. And if your debut being seen is the biggest indicator, then there's something we can do to impact that outcome.