This isn’t a proper blog post – by which, of course I mean it’s not littered with jokes and pun with a strong narrative through-line and a rewarding yet poignant sign off. This is, instead, about a thing I made in Claude in 17 minutes.
The Problem.
It’s really difficult to be programming and also connect with past data – if accessing the data becomes clunky and less agile, it can very quickly turn into “gut feeling”. Working out how many screenings of any film to programme has always been a finger in the air job, let alone knowing the best days and times for those screenings to be on. So I decided to try and build something that helped with that – a scorecard that gave data-based guidance to help a cinema programmer.
The Data.
What I easily have to hand is sales performance data – so film title, date, time, numbers sold and box office income, event ID and instance ID from multiple years. It feels like a massive list, but actually, on the scale of things, it’s a pretty tiny data set. So I got Claude to ingest the sales data and then append information around certificates and genre – that kind of thing. It then worked those into groups of films and looked at whether they tend to do better or worse than other groups. I also asked it to take into consideration that sales now, in 2026, are worse than pre-pandemic and to not estimate sales from there, simply to help guide genre popularity.
How I developed it.
I’ve dabbled in Claude – so I started with a simple concept and then layered it up slowly. Starting with judging by how well films do by genre. Then adding the certification impact. Then the difference between films and Live Screenings. Also that sometimes there are occasional zeitgeist films (Hamnet!) that break all the rules… Then adding rules around previous sales, significant days (e.g. easter, valentines). Then adding seasonal variance – so we sell more tickets on average in December than August. Then finally cross layering that with day of the week and time. All to tell me how many screenings are sensible and on what days and time.
Nice and simple.
So what does it do.
So say you’re thinking of programming Top Gun 3 in July.
July is a poor month for films so the score will start low. That score will get worse because action films do badly at our cinema. The film however is going to be inexplicably certificate 12A which usually do better. The film will also have Rhys Ifans in (this film is fictional, you know that right?), which adds a local and community pull. It’s also been acclaimed as the greatest film ever made so we get to push our “big hit of the year” button.
The net impact? It tells me for initial onsale I should put in 5 screenings, aim for 425 total audience and avoid matinees.
Fun right?
What next?
Want an external eye to help separate the wood from the trees? I’m pragmatic, to the point, practical and offer marketing advice to make you smile. Why not get in touch, tell me you arts marketing problems and see if I can help.
If you like this then please drop me a message, share this post if you think it’s good and feel free to join my mailing list!