Textie: A Case Study

designing a better way to text

The Problem

As our team brainstormed potential behaviors to change, we thought about aspects of our lives that we wanted to improve, one of which being meaningful relationships. With the huge shift to online communication due to both the pandemic and development of technology, we’ve all felt the stress over accumulating unread messages and inability to respond thoughtfully to all of them. We wanted to create something to address this problem and hopefully help us and the people around us.

Problem: we all want meaningful, high quality relationships. But it’s hard to keep these over geographical distances, and texting is hard.

Finding our target persona

We started off by conducting a baseline diary study, in which we tracked the texting habits of six different students over the course of a week. We observed their existing habits around text messaging, their goals around managing relationships online, and how their stress levels changed throughout the day with respect to the number of new or unread messages they received. We decided to design for the persona of Arjay Amagarwal, who best represented the type of person who could benefit from using a product with the mission that we had.

A rough intervention

Next, we wanted to experiment with tactics that could help people change their habits around texting and make it a more manageable activity. After some brainstorming, we narrowed in on the idea of helping users change their habits by visually organizing their text messages in front of them. To test our assumptions on the efficacy of a method like this, we asked participants to, over the course of a week, write down each of their incoming or unread text messages on a separate post-it note containing information about the text message and lay it out in front of them to see, organizing however they’d like.

This intervention did not actually go as expected and we ended up sending constant reminders to users to complete the tasks and stay on top of it. Many users completely missed doing the tasks for days at a time and found it difficult to create a habit around doing this, see any use or help in this intervention, or find the motivation to continue. While we were able to gather insight on how user’s organized their text messages from this intervention study, the biggest takeaway was that doing an additional activity actually added a lot of friction and increased user stress. From creating a system map, we realized that an intervention like this added to an already overwhelming experience.

Brainstorming a better solution

Taking our insights from the intervention study, we conducted another iteration of brainstorming for our eventual prototype. Themes we saw in our mass brainstorm included de-stressing mechanisms, accountability, automating the experience, and enhancements to the visual interface. From there, each of us sketched out an idea for an interface and we narrowed it down to three top ideas. We decided to further explore the third idea in the images below, a story based daily digest of unread messages because we thought it would be interesting to reinvent the interface of texting, and it provided us with the ability to incorporate other ideas and concepts we learned about behavior change (anchoring, B=MAT, etc.)

We wanted our product to serve as a way to destress at the end of the day, which is illustrated in the product-as-hero storyboard below, with the added bonus of getting something done and building relationships.

Task Flows and Sketchy Screens

From choosing the concept we were to move forward with, in order to explore more of the screens and interactions in our app, we next mapped out all of the user tasks or actions they needed to take in order to achieve the goal of our functionality. We grouped our core interactions into four main flows:

• Onboarding. A series of pages where users can set their preferences around the habit of using our app.
• Home screen. A high level overview of everything our app has to offer.
• Intermediary screen. An overlay to the actual chat screen, which contains information about sender and summary of text messages.
• In-story screen. A detailed chat screen which features reply suggestions and remind-me-later features.

Refining our design and moving to higher fidelity

Visual Design

Next, we defined our visual design and brand and ultimately settled on a color scheme, typography, and iconography that featured elements of coolness, roundness, and simplicity to convey a sense of calmness and playfulness to the user.

Mockups

Our team worked to refine the designs we made earlier into higher fidelity mockups using the branding guidelines we settled on earlier. We focused on screens that would provide the most behavior change potential, whether that be making texting more actionable through AI generated responses, making reading texts more frictionless through NLP generated summaries, or through anchoring with notifications that comes up after an existing habit you have.

We then linked these pages up into a clickable prototype that could ultimately be used for usability testing.

Usability Testing

One of the most important pieces to this process was making sure that users could actually use our app and to get a proof of concept of sorts through a usability test.

We created a set of tasks that encompass the key interactions of our product:

• Setting up a Textie time. We wanted to see if it was intuitive for users to set up a designated time each day to use Textie to respond to their texts and if anchoring made sense for a habit like texting.
• Responding to an individual message when you don’t know the right words. We wanted to see if people would make use of the AI-generated responses, if they would find those helpful, and/or if those were easy to find.
• Deferring response to a busy group chat. We wanted to see if people would make use of the “remind me later” function and how intuitive that concept was, despite having it be a pattern we saw in our comparator research.

We conducted this usability test on two students and instructed them to complete these tasks with general goals in mind. Overall, the students were able to successfully carry out the tasks. However, from observation and a post-experiment short interview, we got some good feedback on parts of our app that were unclear or could use improvement. Here are the top 5 issues we saw, and how we would fix them in the future.

1. Transparency around read receipts. Users seemed unsure of what was going to happen and thus hesitant to click from the intermediary screen into the in-story chat screen. There was a lot of uncertainty and confusion around whether clicking into the intermediary screen would remove the chat from the story queue or send a read receipt. To mitigate this, in the future we would be very clear in onboarding how our app would be communicating with the native messaging apps, by including a demo or a series of instruction pages. We would also like to add clear settings where users can turn on or off read receipts based on how they interact with the app.
2. Clarity of affordances and actions. What happens when you click the X? Or when you skip to the next story? Or when you don’t respond? Does the story stay in the story queue until you respond or do one of the above actions remove the story from the queue? In an improved design in the future, we would be more clear around what each action does and present this information in the onboarding demo.
3. Authenticity of AI generated responses. When the only response to turning down someone in our clickable prototype was “Sorry I’m busy!”, users were hesitant to select this option as it did not sound authentic or like something they would say. This raises the question of how personalized the AI generated responses can be and prompted us to not only consider smarter AI responses tailored to each individual’s texting language, but also a design choice in which selecting a generated response populates the text field so that the users can modify it instead of just sending, offering more flexibility.
4. Mental models of the “remind me later” button. One user mentioned that when clicking on the clock icon which represented deferring response to that chat at a later time, they expected something like “remind me in 5 minutes” or “remind me in 1 hour” versus the “remind me after breakfast” type options we had. They were used to these mental models of a remind me later button from apps like Slack. Because of this, we might consider changing the format of the remind later button to be more like a snooze button rather than setting another Textie time.
5. General confusion around icons. We had a number of icons including the information icon and the progress bar which caused some confusion such as what the information icon was used for and whether that was unique to each chat or persisted across all conversations.

Overall, this was a wonderful experience to exercise the design thinking skills we learned and also apply concepts we learned about behavior change to a topic that interested us.