
Which movie genres share the same fans?
I analysed 32 million film ratings to see which genres attract the same people. The result is a map of audience taste.
Last week, I posted my research into the problem caused by the growing popularity of horror. Specifically, the word “horror” now encompasses such a broad and diverse range of films that it has begun to lose its intended utility.
I argued that we need horror subgenres to achieve what the word “horror” once did in the 20th century, i.e. to give audiences a clear idea of what to expect.
That piece prompted several readers to get in touch with questions about genres and the audiences they serve. Today, I want to use data to explore one of the key aspects people asked about - the audience. Or more accurately, the audiences.
To do this, I turned to the MovieLens database, which contains over 32 million film ratings. It gave me a way to measure how similar or different genres really are, based on the people who actually watch and rate those films.
Measuring fans’ taste overlap
To understand how similar viewers perceive genres, I employed a technique called Principal Component Analysis (PCA). This compresses complex, high-dimensional data into a two-axis plot, keeping as much of the underlying variation as possible.
This technique is often used in genetics, image analysis and recommendation systems. Here, it reveals which genres get similar ratings from the same people.
The chart below is the key output, and several key insights can be gleaned from it.
Each point in the chart represents a genre, and genres that are close together are typically liked by the same viewers. Those far apart have less audience overlap.
To make it easier to interpret, I’ve labelled the axes in plain English. These are my words to explain what we’re seeing, not technical terms from the dataset.
Horizontal axis = Challenging to comforting Viewing
Genres on the right are the ones people turn to when they want something familiar, warm or safe. Musicals, animations and family films tend to land here. Genres on the left sit outside that comfort zone. They are darker, heavier or more intense. Horror, crime and psychological thrillers cluster in this space.Vertical axis = Adrenaline to heart
Genres at the top focus on emotion, character and reflection. They invite viewers to feel something lasting. Genres lower down rely more on movement, tension or energy. They keep you alert, push the pace and often aim to thrill more than to stir.
There are several key lessons to be learned from these calculations:
1. Horror sits on its own (in the corner)
Horror is far to the left and deep into the lower half. This means it is both challenging and fast-paced. Viewers who enjoy horror tend to rate other genres very differently.
That makes horror an outlier in terms of taste, as it attracts a distinct group.
2. Family (films) stick together
Children, Animation and Musical sit close together on the far right. These genres are comforting and emotional. Viewers who like one tend to like the others.
They rarely rate darker genres highly as this group favours warmth, safety and joy.
3. The ‘thinking genres’ ponder together
Drama, Romance, Documentary and Mystery are tightly packed in the upper centre. These are the genres for viewers who value emotional depth and moral complexity, and tend to be slower, more thoughtful and character-led.
4. Excitement, without the fear
Sci-Fi, Action and Thriller all appear in the lower half of the chart. They share a taste for pace and adrenaline. These genres are more challenging than comforting, but not as extreme as horror.
5. War crimes and crime wars
War films are found high up and far to the left. It is emotionally serious but not comforting. This may reflect its historical significance or perceived importance. Crime is nearby, though with less intensity.
6. Comedy and adventure are middle of the road
Comedy and Adventure hover near the centre. They are lighter than horror but more dynamic than drama. This position suggests broad appeal and a mix of pace and warmth.
What this tells us
This type of research is not intended to instruct filmmakers or film professionals on how to view genres. Instead, it highlights how audiences already view them.
It gives us clues about which genres are compatible. If your film sits at the intersection of Action and Thriller, you’re likely to reach a large overlapping audience. If you’re blending Documentary and Horror, the task may be more challenging.
By examining genre through various lenses, we can move beyond the notion that genres are fixed, rigid categories. We can start to see them as clusters of taste.
Notes
Today’s data comes from MovieLens, a long-running film recommendation site created by GroupLens at the University of Minnesota. The dataset I used includes 32 million ratings by 200,000 anonymised users across 87,000 films. I focused on users who had rated at least 20 films. I wanted to look at which genres share the same fans, so I focused on ratings of at least four out of a possible five.
Each user’s ratings were aggregated into a genre profile, based on the share of their high ratings across genres. From these, I calculated Pearson correlation coefficients between every pair of genres. The result is a symmetric correlation matrix showing how strongly the same people like two genres. Positive values mean overlap in taste, negative values suggest divergence.
I then applied Principal Component Analysis (PCA) to this matrix. PCA reduces the dimensionality by finding the two directions (components) that explain the most variation in the data.
Each genre becomes a point in this two-dimensional space. Genres close together have overlapping fans, while those far apart don’t. The final axis labels are mine, based on how genres are grouped in the space. The maths finds the patterns, and I added the words afterwards to help explain it.
It’s worth noting what this chart doesn’t capture:
It doesn’t distinguish between mainstream and niche titles.
It doesn’t account for subgenres, like political thriller versus slasher horror.
It also averages out each user. Someone might enjoy both Sci-Fi and Romance, but rate them very differently depending on their mood, style, or subject.
Still, as a snapshot of how genre preferences cluster, this map is helpful. It’s based on what people actually rate highly, not what they say they like. That makes it a grounded way to look at how genres group in the minds of audiences.
When one of the Global Streaming Platforms feels there are 350+ different audience profiles, and another uses only 15 -- and they all try to provide a regular cadence of supply to each of these groups, it makes me wonder what type of films hit the most multiple audience profiles consistently? Of course that also requires knowing what all the audience profiles are...
I'm not asking for it, to be clear, but I'd love to play with an interactive version of this where you could click on a genre and see the same distribution for subgenres. It's interesting to see how far action and war are from each other - I wonder which subdivisions of those would be closer through this lens.