User Experience Design Portfolio of Martina Pérez
gif14.gif

Hybrid Chat Assistant

Increasing the number of users who complete the Setup

gif01.gif

Overview

I was working along with Fran Riádigos, founder of ChattyLabs, on a multi-country and multi-language mobile app that allows users to listen & reply to their instant messages by voice.

 
gif09.gif

Problem

The app was already published on Google Play Store and we’d been tracking some events on Google Analytics, so we got into the logs and realised that 65% of users abandoned the app during the Setup process.

The challenge here was to find out a way to increase the number of users who complete the Setup.

 
gif03.gif

Approach

We had multiple brainstorming sessions. Finally, we decided to design a Chat assistant. We knew that it needed to be very easy to use, with a friendly tone of voice, able to handle mistakes and help users recovering from errors.

To that end, we thought of a Hybrid assistant, using both text and voice. On that way, users would not be only setting up the app, but learning how to interact with it.

 
gif04.gif

Defining User flows

I worked on the v1.0 of the user flow with the aim to cover every single path users could follow. The developer told me it was not feasible because of the limited capacity.

I designed a v2.0 of the user flow much more simplified. We also added new options as to ask the user to select the main use of the app.

On the v3.0 of the user flow we went deeper on Replying to messages and the response of the system in every case.

 
gif11.gif

Design

Designing the chat assistant was one of the most difficult parts. The goal was to design a clear UI as to help the user to focus on the content.

The second piece of work was to work on the tone of voice and narrative of the Chat Assistant with the aim to explain every single step in the simplest possible way.

We added a last question to measure the user satisfaction with the new assistant.

 
gif12.gif

Tracking events

We had to decide which steps were worthy to track to avoid getting messy data.

I defined a bunch of 26 events to track helping me with an Excel sheet. To be sure about the usefulness of the data, I wrote down the purpose of each of the events and we ended up reducing the number to 14.

At last, I created the funnels on Firebase in order to see at once where users dropped off.

 
gif08.gif

Outcome

We made a first release to our group of Beta Testers who sent us their comments. They were very pleased with the new assistant, but gave feedback about some confusing messages. After updating the app based on their feedback, we did an A/B Testing where we showed the new assistant to 20% of users. The results turned out to be surprising.

We increased the number of users completing the Setup process to 70%. Besides, 98% of users were satisfied with the new assistant.

 

Takeaways

  • To measure the effect of UX is crucial to define clear metrics at the beginning of the project.

  • At the time of testing an app which involves multiple processes and tools, and also has multiple contexts of use (car, home…) with a complex variety of errors and experiences depending on the Android device and connectivity, it’s useful to build the feature, release it and test it in context, instead of building a rough prototype that would show an experience very far from reality.

  • Collaborating with a developer on designing the user flows and layouts is not only rewarding, but key to the success of the project, especially when creating the concept of an app that involves multiple technical processes.