In Making London's hidden film clubs discoverable, I didn't need to scrape new data. The information needed to identify a film club was already sitting in the collected data; in the title, the URL, or a performance note. I just needed to ask a different question of that same data — and film format turned out to be the same story.
For anyone who hasn't come across this: most screenings in cinemas are digital. But there are also many cinemas that run physical prints — 35mm, 70mm, occasionally 16mm — usually for classics or restorations (what I've called the format source). There's also other presentational aspects, such as IMAX, ScreenX, Dolby Cinema, etc. (what I've called the format presentation). Each setup has its own look and its own following, and for 35mm print some people will specifically go looking for that over a digital screening of the same film.
The Problem: 35mm Is Playing Somewhere, Good Luck Finding It
Say you want to know what's playing on 35mm this week. Maybe you're chasing a specific film, maybe you just prefer the way film grain looks over digital. Either way, you're stuck doing the same tour every cinema-format enthusiast in London already knows by heart: check the Prince Charles Cinema, check Picturehouse Central, check the BFI Southbank, check whichever smaller venues you think of (like The Cinema Museum). Some venues put the format in the listing title. Some bury it in a paragraph of copy under the showtime. Some add it to the mountain of tags sitting under the listing, indistinguishable from every other tag on the page.
Sometimes it's searchable, but often not. Sometimes you can filter for it, if you know the incantation to use. But no single site treats format as a consistent piece of structured data.
Leveraging the Data I Already Have
Cinemas already want you to know the format; a 70mm presentation of a classic is exactly the kind of thing a venue is proud to advertise. The information isn't hidden by intent, it's just unstructured and sometimes deprioritised. Clusterflick's pipeline already retrieves every venue's listing page to build the standard set of fields: title, date, time, director, year, etc. Format is one more piece of data worth pulling out of the same source.
The challenge is standardisation, not just discovery. One venue writes "35mm presentation," another writes "on 35mm film," a third just puts "[35MM]" in the title. The extraction has to normalise all of that down to a single consistent value — and avoid false positives — before it's worth anything as a filter. Once it is, it behaves exactly like date or genre; pick it from the filter bar and the film list narrows to match.
Dogfooding the Output
Owning the whole flow, from what's retrieved off a venue's page to what shows up on the Clusterflick films page, means I get to use the thing I built (and that's been the motivation all along!). Clusterflick's formats page now exists because format is structured data not just prose in performance notes. Generating a dedicated page for each format is nearly free, just the same underlying field presented as its own browsable view.
That's the value of not just collecting data but consuming it. If the format page felt awkward to use, or the filter missed obvious cases, I'd know immediately, because I'm the first user of my own output.
(And I did spot issues; the first pull missed a bunch of obvious IMAX performances at the BFI IMAX and Science Museum. Whoops!)
Other People Are Building on This Too
The data pipeline's output isn't just for the Clusterflick site. It's published openly (I've an upcoming post on why Clusterflick is open source), and a small but growing set of independent projects already build on it (see Built with Clusterflick for who's currently using the feed).
Structured format data isn't just useful to me; it's useful to anyone building something on top of the same open London cinema data. If you're looking for what's being shown in London, take a look at the showing data that's freely available (and updated daily) on https://clusterflick.com/about/
No Safety Net This Time
Adding format data wasn't the first time I applied this pattern of extracting more value from the same data. I first did it with accessibility data — but that's for the next post, where I can dedicate a proper dive into cinema accessibility. For accessibility data, I was lucky there's a second dataset out there, the UKCA's "comprehensive and searchable listings database for accessible screenings in UK cinemas". Using it as a comparison has caught real bugs (in both directions) and also increased confidence in how accessibility information is detected and tagged.
Format has nothing equivalent. No dedicated site tracks 35mm and 70mm screenings across London cinemas the way Clusterflick now does, which means there's no external source to validate against. That means not knowing what I'm missing or mistagging, beyond spot-checking venues by hand (please let me know if there's a data set out there I can use!). That's a bit disappointing, but I think it could be a good sign. If nobody else has built this, it's either because nobody wanted it, or because pulling clean structured format data out of scattered cinema venue listings is a genuinely hard problem.
I'm hoping it's not the former. Right now the format source pages track 142 films across 12 venues (135 on 35mm, 5 on 70mm, 2 on 16mm) spread across cinemas that range from the Prince Charles and the BFI to the Castle Cinema. I don't think that's a niche nobody cares about, maybe it's just a niche nobody had bothered to structure yet.
Top comments (1)
How did you handle inconsistent formatting in the existing data, and are there any plans to integrate natural language search? I'd love to swap ideas on this, following for more content on searchable film formats.