‘Stranger Things: El Bot’ by Tangent
The people at Tangent were underwhelmed by their experiences of chatbots, which were often incoherent exchanges littered with "sorry I don’t understand"s. So they had a brainwave for a fun internal project – to use the nation’s excitement for Stranger Things
Before they explored the technical potential of AI, they needed to get under the skin of Stranger Things
By combining demographic and psychographic data, Tangent profiled the show’s biggest fans, focusing on those discussing the programme on Twitter, to uncover their behaviours and preferences. Naturally TV came out on top, but there was also a love for gaming, comics, alternative music and the news.
This helped Tangent decide how some of Stranger Things
The bot automatically customised the experience and even ascertained how far through the series viewers were, to avoid spoilers. By routing through a custom RunKit API, Tangent linked Chatfuel and Google’s Dialogflow to improve the bots’ natural language processing capabilities. This meant the bots could deliver relevant responses rather than error messages, providing a coherent, linear discussion.
The project started with an extremely modest media boost. Less than two months later, El Bot had handled more than 22.7k messages and achieved an 89% open rate for sent broadcasts. There was a 35% return to El Bot more than once, against an industry average retention rate of just 4%.