Blink the Bee 

Messenger Chatbot

Role: 

Product Designer

 

Duration: 

Dec 2017 - Sept 2019

 

Tools:

Sketch, Invision, Chatfuel

 

Tasks: 

1. UX design for onboarding

2. UX strategy for improving retention

3. Conversation design

4. To-do list management design

5. Conducted users testing

Teaching small talks to a chatbot

Building a conversation system for a productivity chatbot.

Blink is a productivity chatbot that preformed on Facebook Messanger. It helps users to remember daily todos and remind them when there are important events. Blink featured on Facebook Messenger as one of the top chatbots with over 1M users. As the product starts expanding, we decided to develop a small talk system to help our users focus on adding todos, and avoid the frustration when unexpected events happened.

 

After the solution launched, we obtained feedback that users are more engaged with Blink and feel less frustration. It also improved product retention rate to 19% and Converted 22.9% of novice users into engaged users.

Sadly, Facebook dis-continue some features on the messenger platform, and now Blink is no longer in service.

What is Blink 

Blink is a productivity chatbot on Facebook Messenger that helps users remember their todos. Text Blink your todos so Blink can add them to your list, and remind you at the right time. It's mimicking the simple jot down interaction in our daily life, but instead, I am jotting down these todos to a chatbot. 
For example, Text "Feed cat every day at 8 pm", and Blink will add it on the list and send you a notification at 8pm.

Adding list item

Reminder

Delete list item

Virtual Assistant on Messanger 

Blink designed to be a virtual assistant on Messanger, helping users improve daily productivity, forming good habits, and remembering special events. To engage with our users Blink should have his/her personality. And this is how we describe Blink: Bright, peppy, reliable, organized, proud, tolerant, social-bee. In the early stage, Blink can greet, complete a simple conversation, and even tell bee-puns jokes. And based on the user feedback, they love this little bee!

Situation ​

The characteristics Blink carried make it approachable and as a consequence this causes distractions. Users attempted to chat with Blink and requested for more bee-punchline. Although we loved to see the engagement happening but with the early stage of the product, Blink has not equipped the ability to handle the various "small talks". Blink could not distinguish some todo items and small talks, and it ended up confused users and added everything on to the list, which is not an expected interaction for users. And the repeated conversations make Blink sounds like a can message robot and we want to avoid that.

Define the problem ​

I researched the feedback channel and trying to pin the pain point for Blink users, and here is the conclusion:

  • Instructions are scattered throughout the chat flow.

 ​      We found out that users are more familiar with the traditional UI when it comes to learning the instruction. In   

       the current design, to retrieve instructions users need to text particular keywords which is not as easy to adopt

       for novel users and we want to help them.  

  •  Low error tolerance in the current UX design:

        The current UX flow is designed in a linear model, when mistakes happened users cannot undo, instead,

        they have to remove the mistake and add the correct one. The UX flow needs flexibility for preventing the 

        error from happening.

  •  Unexpected interaction feedback. 

        The thing that frustrated users the most is Blink adding their messages to the list.

Our goals: 

  1.  Help users focus on finish the task, which is adding/editing todos on the list. 

  2.  Avoid frustration.

  3. Chat flow system that is flexible for managing content.

  4.  Increase novel users engaging rate and user retention rate.

Solution:

With the goal in mind, we decided to refine the chat flow for Blink, and build a small talk system for Blink, which will help Blink deal with various conversations and reply with proper answers. 

Design Process ​

1. Data Analysis: 

Our data analyzing engineer dived into the conversation keywords pool,

sorted out the most common messages, and breakdown them into

several major categories. ​

  • Good intentions: Greeting, Joke request, emoji

  • Function requests

  • List items

  • The how-to questions

  • Others intention

2. Technology: 

Due to the vast user base, we know that Blink needs to equip an Ai system to sort out the massive daily conversation, so we look into Watson by IBM and believe that it can provide excellent assistance for Blink. 

3. UX Design for chatbot: 

I designed the corresponded answer for each category, and here are the logistics for the response flow. 

The how-to questions

For the "how-to" categories, users usually have the following needs. Users want to know: How to edit the todo list? How to change the reminder time? How to stop the reminder? And How to quit Blink? For each common questions, we design the flow for each to guide users to achieve what they want. 

 

Greeting, Jokes request, and Curse word categories

I created some good intent responses and product-related responses to guide users adding items on the list.

 

List items

We tried to improve our conversation database for Blink, but there are always new words so we need to crowdsource the data. Initially, we will pull some list items from the users and use machine learning to train Blink to recognize similar todo list items or reminders. We also tried to distinguish a list item by analyzing the composition of the sentence, if the format aligns with the reminder format: verb/objects +(time) + (frequency), We will add it to the list. 

 

Others

If the sentence did not apply to any of the categories above, we throw it to the Others bucket. At this bucket, Blink will double-check with users to make sure if it is a list item or not. If the answer is yes, Blink will add it to the list, and Watson will collect the data for future recognition.

User Chat Flow

4. Conversation Design

Blink's personality is what our users love, so I tried to keep and enhance charismatics by bringing in more natural dialects. Meanwhile, I also some UI format, like graphic and button to facilitate the novel users, help them understand Blink's value at easy.

Testing:

After the system is implemented, we started internal testing. Before we soft-launch the new chat flow to a small portion of users I want to make sure the flow works correctly, all the language is properly delivered.

 

•  Internal test: Fix the bugs, initial Watson Ai training and refine conversation.

•  Beta testing: Small group users training Watson Ai.

Outcome: ​

After we launched the small talk system. We received more positive feedbacks.

•   The app rating on FB messenger improved from 2.4 to 3.6.

•   Users are more engaged with Blink and feel less frustration.

•   The new Blink improved product retention rate to 19% and

•   Converted 22.9% of novice users into engaged users.

Final Thoughts: From a robot to a virtual assistant

It's interesting to see people interact with a productivity chatbot like Blink. I've seen some users ask Blink for advice; people greet Blink as their friend and even curse Blink when they get annoyed. Giving Blink the ability to small talk is making the product much approachable. Meanwhile, we want the users to focus on the value that Blink provides, which is improving productivity and forming habits. The other good thing about the new small talk system is the structure became flexible, which is convenient for designers to swap out the content or expand in the future. As a user I love Blink and I often use it to remember my todos. As a designer, I saw a small chatbot transforming from a robot to a virtual assistant. Furthermore, Blink is able to create bonding with humans and create impacts on their life. It is amazing to see that happened. 

Like what you see?

Shoot me an email @ chu88ryu@gmail.com

Wenhsin Lin Portfolio | Product Designer | 2020