Exploration 01
EE TV, search that thinks
Concept
·
TV + mobile
·
Voice & natural language
The idea
Nobody knows what they want to watch. They know how they want to feel.
TV search is built around a fact that isn’t true: that people arrive knowing the title. Everything follows from that assumption: a text field, an alphabetical keyboard, and a grid of genre rails for when the field fails. So the person who wants something funny, or something like the thing they watched last week, has to translate that into a guess at a name and hope.
Underneath, the problem is that conventional search matches tags. A film is labelled horror, or comedy, or 2012, and the query has to land on one of those labels to return anything. Ask for something a tag doesn’t describe and you get nothing, not because the film isn’t there, but because nobody thought to write that particular label on it.
EE TV: voice search journey
On the TV: the customer speaks a starting point, answers a short run of questions to narrow it down, and lands on three films rather than a wall of them.
On mobile
The query a tag can’t answer
The mobile screens make the argument more sharply, because they show a question the old model simply cannot process: “find me the horror film with the main character from Harry Potter.”
There is no tag for that. Answering it means knowing who starred in Harry Potter, knowing which horror films that actor has since appeared in, and understanding that “main character” here means the performer rather than the role. Three separate pieces of reasoning, none of which live in a metadata field; and the result, The Woman in Black, is a film that was always in the catalogue and was simply unreachable by asking for it that way.
Mobile search: prompts instead of genres, a visible thinking state, and a niche natural-language query resolved to a single confident answer with play available immediately.
Key moves
01
Suggest prompts, not categories
The chips aren’t genres. Make me laugh. Give me a scare. Something like Shrek. Movies with a Hans Zimmer score. They’re moods, comparisons and credits, the actual shapes people’s preferences come in. They also do a quiet teaching job: they show the customer the kind of thing they’re now allowed to ask for, which matters when you’ve spent a decade training people that search means typing a title.
02
Narrow by asking, not by filtering
On the TV, the journey doesn’t dump results and offer filters. It asks a short run of questions and uses the answers to close in. That’s the right trade for a screen you’re operating from a sofa with a remote or your voice, filtering is a mouse interaction wearing a TV costume, whereas answering a question out loud is genuinely easy from ten feet away.
03
Three results, not three hundred
The TV journey ends on three films. Every instinct in a catalogue business says show more, but more is what causes the twenty minutes of scrolling that ends with nobody watching anything. Committing to a small, confident answer is a harder design position to hold and a much better experience, and on mobile the niche query narrows all the way to one, with Play now right there.
04
Show your working
“You’re looking for a horror film with the actor from Harry Potter; that’s The Woman in Black with Daniel Radcliffe.” The answer restates the reasoning before delivering the result. When a system makes a leap this large, showing the chain is what separates a confident answer from a lucky guess, and it gives the customer somewhere to correct it if the leap was wrong.
05
Say when you’re thinking
Reasoning takes longer than matching a tag, so the wait is real and needs naming rather than hiding. The Thinking… state (the same pattern used for voice in the parental controls work) makes the delay legible, and keeps a cancel within reach for anyone who’s changed their mind.
Get in touch
© Harry Thorpe · 2026
DesignED with FRAMER