What Should I Watch
An early attempt at conversational content discovery.
My RoleDesign Lead
CompanyAmazon
PlatformAPL
ScopeNew Feature
Background
One of Fire TV’s hero features has always been voice search. However, voice assistants (like Alexa) operating without the support of LLM were often limited in many ways. One of the more egregious limits was the inability for Alexa to keep context, or in other words, remember a conversation.
But humans aren’t perfect, especially at the end of a long day. Not everyone knows the right title of a movie (most of the time I don’t even know what genre I want) and the right subscription service to watch that movie. In situations like this, we turn to the person next to us and ask:
“Hey what’s that movies where Tom Cruise is a fight jet pilot?”
“Top Gun? There is also a sequel”
“Ah ok, which one is free to watch?”
What Should I Watch was an experiment to look at a CX where Alexa is capable to have such conversation and surface the right recommendations.
Approach
The standard Fire TV search is not built to help you narrow stuff down, but rather re-rank results so the more relevant titles appears on top. With What Should I Watch, I proposed the team to change our thinking and take on a funnel like approach. After the initial query, subsequent modifications moves customer up or down the funnel, giving customer a clear sense of control of their search result.
While customer chats with Alexa, customer intentions are translated into filters. Customer can see in real time what filters are applied or removed based on where the conversation is going.
I also proposed a grid based results page to further enhance the funnel metaphor: it is much easier to see results get narrowed down in a grid.