K.

Back

K.

Back

K.

Film Finder: Mood-Led Discovery

Helping Radio Times app users find their next favourite film by mood, not just by title.

Project details

Team

Kanza Leghari

Snr Product Designer

Max C.

RT App Editor

Software Engineers

Alex M.

Lead App Developer

Platforms & Softwares

UserTesting

User research

Loveable

Initial ideation

Claude Design

Prototyping

Summary

Radio Times has reviewed over 36,000 films, but users could only reach them through keyword search, which assumes they already know what they want. Research showed many people choose what to watch based on how they feel, turning to social media, AI tools or friends for inspiration.

Film Finder started life as an AI experiment by the Radio Times editorial team, who built a quick concept prototype in Lovable. We took that idea from prototype through user testing, native app design and development in a quick sprint. I led the design of the in-app experience, which now sits alongside search on the Discover screen and is powered by the same AI recommendation model behind For You.

Radio Times has reviewed over 36,000 films, but users could only reach them through keyword search, which assumes they already know what they want. Research showed many people choose what to watch based on how they feel, turning to social media, AI tools or friends for inspiration.

Film Finder started life as an AI experiment by the Radio Times editorial team, who built a quick concept prototype in Lovable. We took that idea from prototype through user testing, native app design and development in a quick sprint. I led the design of the in-app experience, which now sits alongside search on the Discover screen and is powered by the same AI recommendation model behind For You.

Most successful acquisition drive since app launch

Uplift in session durations

Continue reading for the detailed version.

Research &
Discovery

Old discovery experience

Discovery in the app relied on keyword search and a filter sheet of type, genre and streaming service. It worked for users who knew what they were looking for, but offered little help to someone who just wanted something good to watch tonight.

We retained the keyword and filter search experience in the new design but added in a layer of more exploratory queries with the film finder.

From editorial experiment to concept

The idea came out of AI experimentation within RT editorial, who used Lovable to quickly prototype a standalone Film Finder on the web. It explored three discovery modes: “I’m in the mood for…”, where users pick a vibe such as Feel-Good or Mind-Bending; “I want something like…”, where users enter a film they love; and Random Roulette, which picks a single film within the user’s filters. Each mode surfaced RT-reviewed films with star ratings, trailers and where to watch.

Rather than starting from scratch, we used this prototype as a ready-made test bed to validate the concept with real users before investing in native design.

User testing

We tested the prototype with around 20 participants through UserTesting, covering current habits, a task for each mode and overall sentiment.

Mood-based selection performed best. Most participants felt the results matched their expectations and trusted them, often because they recognised the films. “Something like” was more mixed: searching for Pride & Prejudice returned romantic comedies when participants expected period dramas, showing the model leaned too heavily on vibe over attributes like era and genre. Random Roulette split opinion, with some finding it playful and others frustrated by repeats and older titles.

Across all modes, participants felt six results was too few, wanted more recent films and broader streaming coverage, and questioned where star ratings came from. Most didn’t use filters, simply because there weren’t enough results to narrow down.

Design & Experimentation

Phase 2

Version 1: bringing Film Finder into the app

I translated the concept into a native flow focused on the strongest modes. Random Roulette was dropped, and the mood journey became a guided sequence: choose a mood, say whether you’re watching with family, and select your streaming services.

To address the genre accuracy issue, I added an optional “match the vibe” step where users tick films that fit their mood before seeing results. This gives the model a clearer signal while staying skippable for users who want to go straight to recommendations. Results moved from six cards to a full poster grid, with a Refine sheet for streaming service and a recency slider responding directly to feedback about older films.

Film Finder was introduced through a “Try our Film Finder” card on the homepage and a “Looking for a film?” section below the search bar on Discover.

Version 2: giving discovery its own space

In V1, Film Finder sat beneath search, competing for attention on a single screen. In V2, I separated them with a Search / Film Finder toggle at the top of Discover, so keyword search and mood-led discovery each have a clear, dedicated space.

The “something like” search gained year labels in autocomplete to help users pick the right title, and a short note explains that picks are based on RT reviews and star ratings enhanced by smart technology, responding to testers’ questions about trust and transparency.

A/B testing film finder

To measure its impact, we ran an A/B test where half of app users had access to Film Finder and half did not. We compared how each group engaged with the app and how often they returned. The Film Finder group came out ahead, showing higher engagement and stronger return visits, which gave us the confidence to roll it out to everyone.

Outcomes &
Learnings

Film Finder is now live for all Radio Times app users, and has become the centrepiece of the most successful acquisition campaign since the app launched, turning a discovery feature into a reason to download.

Building on editorial’s Lovable prototype let us test a real, working concept with users early, so we knew which modes resonated before committing to native development. Moving from experiment to development in a single sprint also showed how quickly AI-assisted prototyping can turn an editorial idea into a product feature.

Film Finder shares its underlying model with For You, which personalises the Radio Times homepage using behaviour and preferences. Together, they give users two routes to discovery: one shaped automatically by what they watch, and one they can steer by how they feel. You can read more in the For You case study.

Film Finder is now live for all Radio Times app users, and has become the centrepiece of the most successful acquisition campaign since the app launched, turning a discovery feature into a reason to download.

Building on editorial’s Lovable prototype let us test a real, working concept with users early, so we knew which modes resonated before committing to native development. Moving from experiment to development in a single sprint also showed how quickly AI-assisted prototyping can turn an editorial idea into a product feature.

Film Finder shares its underlying model with For You, which personalises the Radio Times homepage using behaviour and preferences. Together, they give users two routes to discovery: one shaped automatically by what they watch, and one they can steer by how they feel. You can read more in the For You case study.