The Algorithm vs The Audience
16/April/2026
There’s a creeping sense within independent film circles, one that feels both inevitable and unsettling, that we’ve ceded something fundamental. It isn’t just the way films are financed or distributed anymore. It’s something more subtle: how the very choices about what gets made are being shaped by invisible forces. As I’ve written before about the binge economy and how streaming alters our consumption patterns, we’re now seeing a deeper shift from human to data-driven decision-making. This shift begs a big question: who actually decides what gets made, now the audience, or the algorithm?
In the past, creative commissioning often involved a blend of instinct, taste, and risk. You pitched a story to a producer, an editor, or a commissioning exec. They might love your idea, hate it, or be somewhere in the messy middle. They talked to other humans, watched dailies, argued, laughed, and argued some more. There was inefficiency, sure, but there was also experience in the room.
Today, platforms like Netflix, Amazon, and Disney+ increasingly make commissioning decisions based on data patterns and predictive models. They look at what viewers have watched, when they stopped watching it, how quickly they shared it, and whether they watched multiple episodes back-to-back. They then feed all of this into algorithms designed to anticipate what will perform well. Rather than commissioning based on serendipity, gut instinct, or curatorial vision, many greenlights now come from patterns buried deep in code.
This is not entirely new. Hollywood has always chased “audience demand” and studios have always tried to make films that appeal to the largest possible demographic. But data-driven commissioning raises the stakes. Platforms aren’t just guessing what might work: they’re crunching millions of data points and using machine learning to decide before a film is even made whether it will succeed or fail. That’s powerful and deeply reductive.
At industry panels and in trade press, you’ll hear a familiar refrain: “We listen to the audience.” “Our decisions are audience-led.” But here’s the rub: the kind of data platforms use is not the same thing as audience preference in the human sense. It’s behavioural, not emotional because algorithms track behaviour, not aspiration; patterns, not passion.
Think about it this way: if your film is about a quiet corner of British life, something rooted in place, class, or nuance, its audience might be relatively small. But by no means does that mean it lacks cultural value. Algorithms, however, see a limited expected reach as a commercial risk. So they deprioritise films that don’t fit a pattern of high projected engagement. This isn’t the audience rejecting those stories, it’s an algorithm saying those stories are unlikely to generate eyeballs, and that becomes justification for avoiding them in the first place.
This is the myth of “viewer demand.” It’s not that audiences don’t want nuanced storytelling, it’s that platforms can’t easily quantify that desire from the behaviour they measure. Someone might adore an arthouse film, discuss it with friends, recommend it to their local community, and return to it years later, none of which neatly translates into data points like “number of completions” or “seconds watched in first 7 days.” Yet those metrics are now the primary currencies of commissioning.
So what gets made is mostly what fits a pattern, and what looks safe are films that past data suggests will be watched. These could be a formulaic love story, a thriller with predictable beats or a high-concept adaptation. All of which means that truly distinctive voices risk being sidelined in favour of films that often feel unoriginal.
One of the most paradoxical aspects of data-driven commissioning is this: it promises efficiency and insight, but it actually incentivises homogeneity. When platforms prioritise predicted viewing behaviour, they signal that risk, at least creative risk, is something to be minimised. That’s not inherently illogical from a business perspective; it’s just a different set of values. But from the standpoint of art, disruption, and cultural conversation, it’s concerning.
Take a look at successful films that defy expectation, stories that take time to breathe, characters that don’t resolve neatly, or narratives that resist easy categorisation. These films often generate devoted audiences precisely because they don’t conform to the usual patterns. But for this very reason of being unconventional, fitting them into an engagement model is much harder.
In a streamlined commissioning environment, films that are more likely to attract engagement receive priority. Films that are difficult to predict, which often include the boldest, most original work, end up on the margins.
That’s not to say nothing daring gets made. But increasingly, the path to commissioning now favours algorithms over curators; this is completely wrong! It favours what a predictive model can measure over what a human can imagine.
This might all sound like a critique of streaming platforms or technology broadly, but that would be too simplistic. Because audiences do matter. We choose what we watch, how long we watch it, and where we return. Viewer behaviour still influences trends, and even algorithms depend on human action.
But here’s the difference: audiences have agency, whereas algorithms have authority.
Audiences decide what they want next week, next month, and next year. They talk about films, recommend them, debate them, and let them simmer in cultural consciousness - they create momentum only humans can. Audiences have nuance and contradiction; one person can love a slow, meditative drama while another adores a high-octane thriller. They are messy, unpredictable, and wonderfully human.
Algorithms, by contrast, are statistical. They smooth out messiness in search of patterns. They extrapolate from the known to predict the unknown, and in doing so, they compress nuance into measurable categories. They do not account for mystery, depth, or surprise; those are the very things that make art meaningful.
So when a platform says a decision is “audience-led,” what it really means is that it is algorithm-predicted to suit the largest measured audience behaviour. Not actual human desire in all its richness.
So, what does this mean for Independent Filmmakers? For emerging filmmakers, especially those outside the commissioning mainstream, this shift feels personal. Independent cinema has always lived in the spaces that algorithms struggle to quantify. If filmmakers such as Derek Jarman, David Lynch and Jean Cocteau were around today, relying on their films to be financed by algorithms, the world might have missed out on many masterpieces.
In earlier eras, festivals and sales agents played a crucial role in amplifying films that didn’t fit predictable profiles. A strong festival run could spark critical buzz, attract distributors, and secure deals territory by territory. But when commissioning happens earlier, and when platforms don’t rely on festival performance to guide acquisition decisions, that whole ecosystem changes.
Suddenly, the metrics that matter most are not the ones that celebrate nuance and artistry. They are instead the ones who reward scale, pace, and predictable engagement.
For British independent films rooted in specific landscapes, cultures, and social realities, this poses a real challenge. These films often thrive on subtlety, texture, and voice. They resonate deeply with those who seek them out, but they rarely produce the kind of data that signals mass engagement to an algorithm.
That’s not to say independent films can’t succeed; they can, and they do! But the pathways to visibility are sadly narrowing. Instead of festivals sparking distribution momentum, filmmakers increasingly find themselves pitching to platforms earlier or tailoring ideas to known engagement patterns.
Some might argue that algorithms are simply the latest form of curation. After all, they do reflect human preferences; they just analyse them differently.
But an algorithm is not a curator in the human sense. Curators make choices, sure, but ones rooted in context, conversation, cultural framing and from a truly human perspective. They champion risk, provoke debate, and sometimes celebrate films that defy commercial logic precisely because they push discourse forward.
Algorithms, on the other hand, optimise for retention and efficiency. They do not elevate films based on artistic ambition; they elevate them based on the predictability of engagement. That’s not inherently negative, but it is limited and very black-and-white, and those limitations matter because they shape not just what gets watched, but what gets made.
So how do we navigate this? How can filmmakers, critics, and audiences assert human agency in a landscape increasingly dominated by metrics?
First, we have to recognise that our choices matter even if algorithms interpret them in reductive ways. When we choose what to watch, what to recommend, what to talk about, we are signalling values that no algorithm can fully grasp.
Second, we need to advocate for commissioning models that balance data insight with human judgment. The platforms that publicly celebrate risk, that invest in curators with deep cultural expertise, and that measure success beyond raw engagement metrics will help keep the creative ecosystem healthy.
Finally, as filmmakers and creatives, we should continue to make films that challenge, surprise, and unsettle. Not because they fit a pattern, but because they matter. Audiences will find them perhaps not always at first glance on a homepage, but in the conversations they spark, the communities they build, and the reasons we return to cinema in the first place.
The future of film is not binary because the tension between algorithm and audience isn’t a battle we will “win”, it’s a dynamic we will live within. But if we treat data as one tool among many rather than the final arbiter of worth, we can preserve space for stories that defy prediction.
Because the best films, the ones that endure, that change us and stay with us, are rarely the ones that were easiest to quantify.
