Auronix - Challenge

lessons learned

Job Application Challenge

Company: Auronix

Role: Conversation Designer

Date: Abril 2021

Description: A travel bot chatbot to demonstrate understanding of chatbot functionalities and capabilities.

This project is the result of three days of working.
*The bot was designed in Spanish.

Challenge

Create a Chatbot with the theme of your choice and in the program/software of your choice. We are looking for someone capable of taking on challenges and making them a reality. We want a person who understands how a chatbot can work and all it can do.

You must have your chatbot on a public link so we can test it. You can create a folder on Drive or send it to us via Dropbox.

We will evaluate:

  • Originality and Creativity
  • User Experience (easy to understand)
  • Spelling and Writing

Bonus points:

Earn extra points by mentioning what data we can extract from this bot, e.g., number of users who completed the conversation, number of users who answered X question.

Case: Gurú de Viaje (Travel Gurú)

Guru de Viaje is a Mexican flight search engine; they do not sell tickets. They work with various airline websites and travel agencies to share offers to destinations within Mexico and around the world.

About

Site information:

"We hunt for deals on the web.
We are DEAL-HUNTERS for flights, hotels, experiences, and everything travel-related. Every day, we publish the best deals we find on various sites and tell you how and where to buy them, so you can take advantage of them."

"What We Do Sell:
We offer a variety of products such as tours and packages to different parts of the world, which are sold directly by Gurú de Viaje, just like any traditional travel agency."

According to its creators:

It is a travel demand generator through offers. Customers primarily come from the website, but there is greater retention through the mobile app. In 2019, the two platforms together accumulated a total of 18 million users.

The Problem

Gurú de Viaje has two platforms: the mobile app and the website. However, only the website has a Facebook Messenger Bot implemented, which is only visible in the "About Us" section and does not contain any initial dialogue.

Additionally, it offers the possibility to receive information through WhatsApp groups, which are specialized in the pre-selection of a route of interest.

In the case of creating personalized alerts, both in the mobile app and on the website, on the website, after registering, you create the alert by choosing destinations presented alphabetically by city name. Once the destination is selected, it asks for the departure point and the month. In the mobile app, before the cities appear, you can pre-filter based on whether the user knows their destination or not.

Pains

  • The user would have to join many WhatsApp groups if they wanted multiple destinations.
  • The groups are reported as full.
  • The personalized alerts do not have a search filter, so the user has to review the entire list or look for the specific letter of the destination they want to add.

Proposal

1. Define the Chatbot's Objectives

  • Assist users in purchasing tours and travel packages.
  • Help users create personalized travel alerts.

2. Choose a Platform

  • Integrate the chatbot on both the mobile app and the website.
  • Consider platforms like Facebook Messenger, WhatsApp, and in-app chat for wider accessibility.

3. Set Up Chatbot Framework

  • Use a chatbot framework such as Dialogflow, Microsoft Bot Framework, or Amazon Lex.

4. Design the Chatbot Flow

  • Greeting and Onboarding
  • Purchase Assistance Flow
  • Personalized Alerts Creation Flow

5. Implement Search and Filter Functionality

  • Ensure the chatbot can search and filter destinations.
  • Allow users to search by city name or by the first letter.

6. Integrate with External Systems

  • Connect with travel APIs for real-time information on flights, hotels, and packages.
  • Ensure secure handling of user data and payment information.

7. Testing

  • Conduct thorough testing to ensure smooth user experience.
  • Use beta testers to gather feedback and make improvements.

8. Launch and Monitor

  • Deploy the chatbot on both platforms.
  • Monitor its performance and user interactions.
  • Continuously update and improve the chatbot based on user feedback and new requirements.

Process

UX

  • User Persona
  • Journey Map actual en sitio web
  • Bot Persona
  • Journey Map con chatbot
  • Dialog

Miro board can be seen here.

Implementation

The video showcases the chatbot in action, with Dialogflow integration displayed on the right side. While the model is available for testing, please note that it's specifically trained to respond to the inputs provided within the video.

Metrics

  • Number of users who start
  • Number of users who finish
  • Time it takes for the user to complete their search
  • Moment in the flow where the user abandons
  • Time of day when most queries are made
  • Most frequent destinations, dates, and departure locations for purchases
  • Most frequent destinations, dates, and departure locations for alerts

Next Steps

  • Creation of the conversational flow for alerts
  • Consume API from Dialogflow for destination and departure information, offers, etc.
  • Implement webhooks
  • Implement Firebase
  • Further train the model

Tech Stack

FirebaseReactNode.jsGCPGCP