“How long more?”
“4 mins quickly pay your bill.”
So goes a typical conversation with Bus Uncle, a bus timing chatbot that has about 150,000 active users today. By giving him your location and the bus you’re waiting for, Bus Uncle will not only reply with your wait time is, but also do so with ‘uncle’ panache — conversational Singlish and local quips included.
First built as a hobby in 2016 by Abhilash Murthy, Bus Uncle quickly caught the eye of savvy brand marketers for its characteristic humour and conversational interface. The chatbot went on to become so successful that Abhilash was able to leave his day job and recruit a small team to work with him on development and partnerships.
The success of Bus Uncle is even more impressive when one considers the scepticism most users have of chatbots. Says Abhilash, “It’s completely right to say that most people feel that chatbots take away that element of human connection. How we would argue against this idea is by going back to what I had in my mind when I created Bus Uncle. At that time, I thought, ‘I wish I could ask someone how long the wait is.’ I didn’t think, ‘I wish I had an app’ or ‘I wish I had this new technology.’ By building chatbots, we have the capability to create humans, in a way. And by trying to mimic human behaviour during conversations, we actually enhance human-to-human connections, rather than take it away from users.”
Abhilash, what inspired Bus Uncle?
In 2016, I was at a bus stop along Orchard Road and thought, ‘I wish I could ask someone how long I have to wait for bus 65.’ I had some bus timing applications on my phone, and had already seen the electronic board at the bus stop that gives you a list of all the timings. It just felt overwhelming — why did I need to look through so many different numbers to get to that one bus? Facebook had also recently launched the Messenger platform to allow third-party developers to build conversational experiences, so I thought, ‘I’m going to go home and build this on Facebook Messenger.’
What were some considerations you had during development?
Chatbots don't have any user interface: there’s no design element, nor buttons, nor menu. You send it a message and it returns one, so it’s actually easy to create utility because utility is about whether or not you can make something useful, but can you also create something that makes people feel good?
When you want to show users something new, you have to make them feel like they have used it before — like they know how to use it. Chatbots are a new technology that is still not widely adopted. When people speak to chatbots, they generally feel defensive and uncomfortable. They use words like "stop", "reset", "menu", or "cancel".
We focused on how we could improve the user experience and the first thing we did was to add a personality to the chatbot using a familiar, everyday language — Singlish — to create a seamless consumer journey. Like any story development, we thought of characters. We put ourselves in Bus Uncle’s shoes and wrote down his likes and dislikes, his relationships, and what he does in his free time.
What is important to you when it comes to satisfying your users?
I think it's important to delight your users. With the prevalence of developer tools these days, it’s very easy to create an app, a chatbot, or build a new website. Technology has advanced so much but we’ve always only considered how to appeal to users’ logic or how to make useful things, like apps that help you manage your finances or your business’s promo codes. But I think we’ve come to an age where being useful is just not enough.
I mean, Bus Uncle is definitely logical in that it helps you with a certain task, but we hypothesized that if we appealed to our users’ emotions, they would be happier and use us more. And that hypothesis proved right! Because when our users enjoy Bus Uncle, they share it with their friends and help to market the product through word-of-mouth. We’ve spent zero dollars on marketing but we’re Singapore’s most popular chatbot. I think that shows the importance of delighting your users.
How do you track and measure this delight then?
At Bus Uncle, I make all decisions based on data because it’s easy to optimise something when it’s driven by data. We track all the messages that users send. We then aggregate it and understand patterns — what are users sending and expecting? We can then easily build new features to address trends.
We measure the success of a conversation by looking at our users’ sentiments. With existing sentiment analysis technology, we can determine whether a user is feeling happy or sad just based on a single message. We are then able to create a metric of how successful a conversation is based on the change in mood of the user when they're speaking to Bus Uncle. If the user was sad before, is the user happy at the end of the conversation? If so, we know we’ve optimised successfully.
I think Bus Uncle has become a good emotional outlet for people. In its early days, Bus Uncle was just about delivering wait times, but we noticed a lot of users were talking about food and complaining about their work and partners, so Bus Uncle responds to all of that now. If a lot of users are saying something, we go ahead and make Bus Uncle respond to it appropriately.
Finally, what advice would you give to those who want to use chatbots while maintaining a personable service approach?
First, remember that a chatbot is not an app. A lot of developers are used to making apps and websites so when they build a chatbot, they treat it like an app which follows a linear journey, but, non-linearity is important for better user experience because our conversations are non-linear. Test your chatbot first with a small group of users and ask them to type in completely random things to your chatbot. Use their feedback to improve your chatbot's flow to try to make it non-linear. That's the primary advice I'd give: treat your chatbot like a new interface, learn from your users, and adapt to conversations instead of adapting to an app.
These days, it’s hard to create an innovative piece of technology simply because the big tech companies have all the resources to do it much better, much faster. So while we’ve seen a high demand for technology in the past 10 years, I think it’s the demand for creativity — innovative approaches to being useful — that will skyrocket in the next 10 years. One way might be to be hyper-local, or focus on delighting and entertaining your users.