Context and Challenge
Supportiv is an online peer-to-peer mental health support service founded in 2018. It matches people with similar mental health struggles in online chatrooms. The chatrooms are moderated by trained human moderators to keep them troll free. The matching is done by a smart system that uses natural language processing. The participants type what is their struggle in a conversational interface and the technology matches them. The technology is patented by Supportiv. During the chats the moderators use the AI again to recommend customized self-help resources to the participants while they chat. Supportiv’s key value propositions are its seamless onboarding, no ads and no data selling policy and guaranteed anonymity.
The challenge was to identify improvement areas for Supportiv and come up with a service redesign concept.
I started with the following:
- 1-month-long auto-ethnography.
- Semi-structured in-depth interviews with 27 Supportiv users and 7 moderators.
- Digital listening of online reviews and mentions.
- Web audits of 40 applications in mental health category, mental health forums and facebook groups.
- Literature review of key mental health trends, online peer support specifics and the applications of machine learning in mental health.
In this process I created user and moderator empathy maps, journey maps, a competitive matrix and synthesized all in a SWOT diagram. Based on the insights, I ideated with sketching exercises and prioritized the ideas based on strategically chosen criteria. At the end, I tested the redesign concept with contextual inquiries.
Business Model Canvas
As a starting point, to understand Supportiv’s current service and system better, I constructed the startup’s business model.
The business model canvas showed that:
- Supportiv has two distinct customer sectors. The first includes the end-user customers that are digitally savvy anglophone people that face extensive daily struggles, psychological tension and seek for help. The second includes the business customers that are US-based companies that offer Supportiv as part of their employee benefit packages.
- The startup uses a mix of usage fee and subscription with daily and monthly options for end-users and custom pricing for business clients.
- Its cost structure is more value-driven than cost-driven. The fixed costs are the cloud storage and tech development costs, the variable demand-dependant costs are the moderator salaries.
- It uses different channels for each marketing phase. For awareness it uses social media, search engine optimization and referral marketing. For evaluation, the key touchpoint is the website with reviews. The purchase can be done with paypal, amazon go and credit cards. The service is available through mobile and web apps that offer identical experience. The after-sales service includes weekly newsletter emails.
I created Supportiv’s stakeholder map to visualize the actors of the Supportiv system and understand their relationships. The map illustrated that in line with the main actors, such as the employees and the users, as a service with social component that is heavily reliant on partnerships, Supportiv also has several secondary and tertiary stakeholders that could potentially be involved in the growth of the service.
Through desk research, I identified 4 major mental health trends: blended care, e-therapies, such as cognitive behavioral therapies, self-help and online info resources and alternative interventions, such as creative and art therapies. In all of these trending ares, the use of smart digital technologies seemed to be increasing. The use of analytics-based technologies, virtual reality (VR) and augmented reality (AR) was particularly notable.
User And Moderator Empathy Maps
I transcribed, analyzed and synthesized the semi-structured in-depth user and moderator interviews using the Empathy Map framework.
The user empathy map showed that people start using Supportiv when they are stressed, depressed, lonely, or misunderstood. Their goal is to be heard and understood, but often they discover that Supportiv’s matching is not very precise — people with different problems can be matched in a chat. Participants leave the chatrooms and new participants join any time during the conversation. This is distracting and there is no room for meaningful conversations. Moderators are passive and sound unnatural. The content recommendations are usually relevant and useful but impractical as they are sent during the conversations and there is no save option. Also, the chatroom experiences vary too much — sometimes they are very useful, other times not useful at all.
The moderator empathy map showed that typical moderators are psychology students, or recent graduates working part-time. They are empathetic people, but at Supportiv they have too strict interaction guidelines and are not able to improvise and react properly to specific user problems. They can’t control the chat flows and panic sometimes because of awkward silences or conversations. Moderators also don’t have much interaction with each other to share insights about chat handling and they have unpredictable schedules with periods of idle time and hectic periods.
Current State Journey Map
Based on the auto-ethnography, I constructed a current state user journey map. This enhanced the user experience insights uncovered through the interviews and helped spot more issues.
The current state journey map showed that:
- One inconsistency with the service promise and actual experience happens in purchase phase. The service advertises itself as registration-free, but during the payment the subscribers add an email and set password to use it later for subscription access.
- There is no free trial.
- During the usage, the number of chat room participants can grow to 8 and in this case some people don’t get a chance to express themselves.
- The users are asked to evaluate the moderators only 5 minutes after joining the chat. This makes the feedback meaningless.
- The voice of the moderators is too robotic.
- In the after-sales, there is absolutely no customization of the emails
After conducting audits, I did a competitive analysis based on 7 criteria and constructed a matrix to visualize Supportiv’s competitive advantages. It showed that Supportiv has at least 3 direct competitors. The closest alternative offers an almost identical experience but for free. Supportiv‘s ’ indirect competitors are social media support groups, online forums, online counsellors, meditation apps and other self help apps. Supportiv’s competitive advantage in all cases turned to be its patented technology and moderators. While, Supportiv emphasizes anonymity in its communication, the evaluation revealed that it was more of a must-have rather than a differentiating point.
To synthesize the insights, I used SWOT analysis and visualized in the diagram below.
The diagram clearly showed that Supportiv is underutilizing its two key competitive advantages: AI technology and human moderators. The challenge was to find new ways for combining these two resources in a way that would take the unpredictability and stress from the moderators, give more consistency to the user experience and change the users’ perception of not receiving enough value for money.
Using the opportunities section of the User Journey Map and the SWOT Diagram produced in the research phase, I sketched 6 quick ideas. From the 6 ideas 3 were combined and developed a bit. In the final stage, I prioritized ideas by ranking them according to the following criteria: impact on the users, impact on the moderators, impact on the profitability, feasibility. The idea that got the highest combined score was chosen.
I developed the concept of Supportiv Power, the unbundled and enhanced version of the current Supportiv. It would offer two distinct sub-services: ML based self-care content and exercise recommendations delivered through its personified chatbot Amie and peer-to-peer support chat-rooms guided by moderators with fun alternative therapy exercises. In addition, there would be additional in-app features, such as a personalized library of self-help content, a mood-tracker and personalized notification. During the chats, the tone of voice and the role of the moderators would be redefined. The video bellow shows the concept and its additional features.
Future State Journey Maps
With the redesigned service user’s journey for Amie Bot user sector would not vary too much for new and existing users as there would not be a registration. The high-level interaction flow would be as the following “To-Be” journey map suggests.
The user journey for the peer support option would vary with each encounter, however with the redesigned option the structure for the onboarding would be consistent regardless of the further flow. I created a zoomed journey map for this phase.
The key benefits of the redesign concept would be:
- New unique selling proposition and competitive advantage
- More stress-free moderator experience.
- Employee cost reduction thanks to the addition of the chatbot.
- Long-term customer relationship thanks to the personalized notifications.
- Continuous algorithmic performance enhancement thanks to the data collected through the newly added art-therapeutic diagnostic activities.
- New customer sectors.
I tested the concept with 5 participants of the user interviews conducted back in the research phase. I showed the participants some of the screens and asked to express their opinion about the idea, its limitations and some improvement areas. Overall, 5 out of 5 participants confirmed that the enhanced product would interest them and that they would try it out. One of the participants proposed to include an option of having one-to-one chat outs. Another participant suggested to add a one year subscription option. I included the annual subscription proposal in the final service redesign features. As for the idea of one-to-one chat option, I included it in the future enhancements suggestion agenda. In addition, I did a quick exercise to designing the conversation flow of the chatbot. I asked the test participants questions that were going to be asked by Amie bot for recommendations. I used their answers to define the pre-programmed answer choices used in line with the assumed NLP-based answers in the interactive high-fidelity scenario-based prototype that I developed.
Challenges and Learnings
- When dealing with sensitive topics, in user research, it’s particularly crucial to show empathy towards the users and give them room to open up by being completely non-judgmental.
- When too much data has been collected in the research phase, it’s important to not be in a rush and spend as much time as it requires to carefully synthesize the findings, visualize them and connect the dots. Otherwise, powerful insights might get lost in a pile.
- When advanced technology is involved in the system, it’s important to develop an in-depth understanding of its mechanics before trying to ideate on enhancement solutions. Otherwise, opportunities might be lost, or the feasibility threatened.
I did the project for my master work at Poli.Design. The concept I developed does not reflect the views or vision of Supportiv’s founders. The process took 6 months. For more details, feel free to contact me.