Fixing Chatbot Usability Problems
Using FBM 1.4 updates: menu tiers, menu localization, deep linking and FB Messenger rooms.
For the past few months, we have been working on localizing a WeChat-based, English-language-training bot for Facebook Messenger. The prompt/response nature of messaging interfaces aligns nicely with modern language teaching methodologies (such as the communicative approach) which stress active, student interaction.
We launched a proof-of-concept Real Deal English (https://m.me/realdealenglish) last month primarily targeting Traditional Chinese speaking users with an interest in learning English. We immediately hit a couple issues.
Problem #1: What does this thing do?
I am a fan of Adit Jain’s concept that users are always in one of three states while engaging with a bot: on-boarding, exploration or action.
The on-boarding and exploration stages are critical for giving users some context of what they can do with the bot. While we were able to localize welcome messages sent to the students, the fact that the persistent menu could only be in English presented an additional barrier to non-English-speakers wanting to explore what could be done with the bot.
Localization of the persistent menu has fixed this:
Beyond linguistic understanding there is functional understanding. Because we were limited by what we could include in the persistent menu we created a welcome carousel to help provide some context to what the student could expect with a ‘Home’ button in the persistent menu to return back to that carousel of options.
There were a lot of steps to potentially lose users. A 3-tier, persistent menu makes this more streamlined and accessible:
Problem #2: Why should I come back?
“A lot of the thinking around bots is naively utilitarian and lacks the social intelligence to recognize the range of human needs being met by person-to-person interaction.”
— Bradford Cross blog post
This quote has been stuck in my head ever since I read it. Service provider economics are clear, but do people want to talk to a bot? Interactions are not always just about communicating information they are about satisfying emotional needs for the people involved. I am increasingly thinking this is a fundamental problem with autonomous agents and a big reason why bot retention is awful. This is borne out by our own initial numbers, which show a lot of curiosity from students, but few repeat engagements:
Retention was never such a striking problem with our WeChat trials. I believe this was because we never completely handed the student off to the bot. Instead we used WeChat groups as study groups where a teacher facilitator would warm up the group of students, share a link to the bot lesson with the group and then invite them back to ask questions or demonstrate what they learned within the study group. In short, we augmented the bot experience with a social layer.
FB Messenger rooms and deep linking now lets us try to recreate this same experience in Messenger:
Our ideal use case then becomes:
- use FB ads to drive bot traffic
- get students to join a study group as soon as possible
- assume the student will play around with the lesson in the bot for a while, but then get distracted and leave
- later, teacher facilitators share a deep link to a specific bot lesson to the study group
- students tap on the link to return to the bot and review the lesson
- afterwards, the students returns to the study group to ask questions, discuss and demonstrate their mastery of the language by performing some task
This is what the flow looks like:
Time will tell what impact these changes have on student retention, but a big thanks to the FBM team for all these improvements. I look forward to seeing what is next on the product roadmap!
Hank
Reach out to me on FB Messenger. I am big believer in the ability of automated agents to help make training more convenient, effective and affordable.
