Designing a chatbot — A UX case study

André Nascimento
6 min readMay 14, 2020

--

This is an article about my time at TERA’s UX Design Bootcamp and my learning experience through our attempts at improving consumer behavior. I worked for two months with a team of incredible people, whose mission was to solve a particular problem for a well-known company. It was a fascinating journey that helped me improve my skills in product design.

My role

  • Project management 🧭
  • User research 🔍
  • Information Architecture 👨‍💻
  • User testing 🔬

Context

Reclame Aqui was the company that provided us with this challenge. It is an online platform that connects businesses and customers by allowing consumers to use their platform to make complaints about a service or product they have had a poor experience with. The companies, then, have the possibility to reply to the complaints and ultimately come to an agreement with the customer. The users at Reclame Aqui rate products and companies according to the number of complaints and the company’s effectiveness at addressing the issues.

First steps

Reclame Aqui’s product and marketing team gave us a report based on a quantitative research, containing data showing the consumers’ flow before buying any product.

  • A majority of the surveyed consumers visited the site and read the complaints before purchasing a product or hiring a service.
  • Almost all of the consumers only found the site after searching for the product on Google.

The Reclame Aqui team wanted to improve the experience for those users. Before the start of our project, we framed the challenge in these terms:

“How might we improve our user’s buying decision through the Reclame Aqui platform?”

Empathizing and defining

We started off by creating three rows of post-its and filling them with all the information we had about Reclame Aqui and its users. We divided these rows into “certainties”, “suppositions”, and “doubts”.

The “suppositions” and “doubts” rows set our purposes for the research.

We interviewed eight consumers and surveyed 35 others, with the goal of understanding their behavior and connection to the platform. The outcome of the interviews granted valuable insights and learnings; the main patterns identified were:

  • Consumers seek online validation before buying a product;
  • Reclame Aqui was essential in the decision of NOT buying something;
  • Even users who were familiar with the site didn’t search directly on the platform;
  • The majority of the users didn’t understand how Reclame Aqui works (i.e., how to post a complaint or who would respond to it); however, they trust the information the platform provides.

We used all the research material to create an empathy map and two personas. These tools helped us to focus on our users and to make strategic design decisions.

Empathy map

Ideating

We were a multidisciplinary group, each one of us with unique backgrounds and skill sets. At that point, we had a billion ideas on how to improve our users’ experience. Unfortunately, we spent so much time ideating and debating all the pros and cons that we almost delayed the project. We needed to move forward and narrow down the possible solutions, so everyone chose one idea each to discuss one last time. Then, we applied the Dot-voting method to reach a consensus.

The winner idea was a Facebook Messenger chatbot. Our hypothesis was:

  • It will offer a direct channel for users to find precisely the info they need;
  • It will benefit from being a popular app amongst users;
  • It will provide a new experience without the complexity and costs of building a new app.

Prototype

Prototyping a conversational interface was something new for all of us, as there wasn’t much of an interface to design. For this task, we needed the proper methodology and some guidance. After a few days of searching and talking to our instructors, we defined the following steps:

Benchmarking: We put together similar products to analyze their strengths, weaknesses, and new opportunities.

Product canvas: We created an artifact to document our goals, value propositions, success factors, metrics, and to centralize all we had discovered about users and similar products. The product canvas helped us to define the requirements of the chatbot.

Bot personality: Based on all of the research so far, we determined the bot character, establishing how the bot would respond to interactions and the right tone of voice. We adopted the face of Reclame Aqui’s mascot. and applied a persona’s template to summarize the characteristics of Reclaminho, the mascot’s name.

Chatbot userflow: We used Miro(❤️) to map all of the possible interactions the user could have with the bot, and then we started writing the script for the bot’s responses, defining the bot’s replies to every interaction. We had to be very careful at this stage, couldn’t leave any dead ends, otherwise, the whole experience would be frustrating.

We used different labels for every interaction. Yellow: buttons; white: bot responses; red: error responses; green: carousel with options; blue: user interactions

We used Botsociety to simulate a few scenarios and have a feel of how the bot would look like.

User Testing

How do we test a Facebook Messenger chatbot without developing it?

One of my teammates changed his Messenger profile picture to our bot’s cartoon and initiated an in-app chat with our users. After briefing the users, we followed the script defined at the userflow stage.

His reaction was the best 😂

The main testing results were:

  • There were a few gaps in the lines;
  • One of the branches of the userflow was a dead end;
  • We needed to improve the farewell and welcoming lines.

It was time to improve the flow and iterate more!

Next steps

Unfortunately, the Bootcamp was over. Although the job wasn’t!

We introduced the product at the demo day and collected all the feedback from the Reclame Aqui team.

For the next steps:

  • Build the MVP;
  • Keep usability tests to improve the script and flow;
  • Define what are the metrics of success of the product.

Beyond all the methods and tools we used, the most valuable thing I learned was the ability to prioritize and converge ideas to achieve a goal. It was amazing to see how abstract approaches turned into a well-elaborated strategy to help the users’ buying decision through the Reclame Aqui platform.

--

--

André Nascimento
André Nascimento

No responses yet