La Bringue was founded a bit over a year ago, with the purpose of exchanging party recommandations among electro lovers. It first started as a Facebook group between close friends, but quickly reached 15 000 members through word-of-mouth. Through this growth, the team confirmed that there was a real need for guidance in the parisian night life, especially to break the routine and discover new places.
Quickly, the team decided to address that need through an app. However, they discovered the power of bots when Facebook introduced them at F8 2016. That’s when they chose to create a bot instead of an app. Why is that? We talked with the founding team, and here are their insights on bot building.
Why do you think a bot fits the needs of your users better than an app?
“As we looked into bots after their announcement at F8, we realized three things make them great for our audience :
- It’s natural: the type of information we provide doesn’t require a complex graphical interface. It needs natural conversation.
- It’s quick: who wants, when looking for an after party in the middle of the night, to use an app, browse through dozens of suggestions and figure out which option fits better? No one. A bot provides quick answers for a rushed user.
- It’s friendly: the fact you’re having a conversation with an AI gives you the possibility of creating personalized answers, to the point of creating a true personae for your bot. And because it is among your closest contacts on Messenger, it has to be tailored to your needs.”
Did you notice a change in usage of your product after you launched L2-B2?
“Today, 6000 people use L2-B2 per week. Users come to the bot around twice a month, which is the average amount of nights out for our target audience.
L2-B2 brings an equilibrium to La Bringue. Our Facebook group brings a sense of community where people want to interact with others, share information or get feedback on a new club. It recreates this feeling of belonging you find when you go out with friends.
The type of information we provide doesn’t require a complex graphical interface. It needs conversation.— Raphael Doulonne- Co-founder of La Bringue
Our blog brings curated information to users, in a one-way manner, much like a media.
L2-B2, our bot, provides instant answers and advices to an immediate need.
It’s the complementarity of these three services that creates a true satisfaction for our users, day and night.”
How did you create your bot?
“L2-B2 is developed in NodeJS, a language we didn’t know before diving into bot building. It’s simple and powerful. We discovered it when building our first bot on Wit.AI, by using their template bot.
It was important to us that the bot would be as adaptable to human speech as possible. We wanted to be able to talk to the bot like you talk to a friend, we wanted to create a conversational space with emotions, feelings, a sense of friendship even. To achieve that, we had to implement Natural Language Processing (NLP), or we would’ve ended up with a 100% button bot.
We wanted to create a conversational space with emotions, feelings, a sense of friendship even.— Axel Delafosse - Founder at La Bringue
We’re now using Recast.AI‘s NLP, and it allows us to truly understand what our users want. That’s the beauty of conversation! In an app, we’d have to be restricted to pre-set fields, whereas in a bot, the user expresses his needs directly, through no intermediary. It’s then essential to leverage the power of NLP to gather all information and provide an excellent service, not through predefined flows but through the liberty of human speech.
However, to create a smooth interface for our users, we redeveloped our own version of context management based on Random Access Navigation (RAN). We needed to step out from a basic survey flow to adapt to whatever input we got. Today, no matter in what order a user gives key information such as music style, location or date to the bot, L2-B2 is capable to understand which information he got, which information it still needs, and ask the relevant questions. If all key information are provided in one sentence, he directly presents suggestions. That is something we developed from scratch.”
What are the next steps for L2-B2?
“Currently, we’re working on improving two things:
- Search and discover: to present suggestions to the user when we don’t have exactly what he is asking for.
- Analytics: to better understand what our users are asking or which cards are clicked the most.”
A few best practices for fellow bot builders?
“There are three things we really want to put an emphasis on.
First, don’t create a questionnaire bot where the user has to answer 3 or 4 questions in a row to get what he wants. Strive to provide the information as quick as possible, and leverage the power of text to do so.
Second, focus on the core of your bot. It’s tempting to create a bot that knows and answers everything, but rather choose to address one use-case perfectly.
Finally, find the right balance between simplicity and identity. Create a user experience that is smooth and efficient but also conveys a personae that fits the values of your brand.”