TfL TravelBot: Designing the conversation

As you may already know if you’re following this blog, we recently released the TfL TravelBot on Facebook Messenger. If you haven’t read them yet, Steven and Charul’s posts will give you a bit of background.
Check out TravelBot here or search for TfL TravelBot in the messenger application. In this post I will explore the reasons for introducing a conversational bot and our learnings around the design of conversation.

Diverse backgrounds, cultures and lifestyles mean that we all use different words to talk about things. This can become frustrating when you’re trying to find something on a website.

In our team, we try to label things in a way that most users will understand, but are well aware of the fact that we will never be able to cater for everyone. This means that some users have to change the way they think to match what they are looking for.

Giving control back to the user

Conversational interfaces have the potential to address this problem and give control back to users by letting people speak naturally.

By giving the opportunity to ask a question as you would in a normal conversation, we reduce the amount of steps someone needs to take to get to a specific piece of information. In most cases a question leads to a direct answer. This means users don’t have to browse a menu and gradually drill down to the information they need or make sense of search results.

Getting information around bus arrivals for example. Changes from having to navigate to a specific application, planning a route and selecting a stop to asking ‘When’s the next 365?’.

Building what you ask for

The TfL TravelBot is helping us to understand the ways that customers talk about our products and services. For the first time we are able to see when customers aren’t getting what they expect and change a product to suit specific needs instantly and to respond to changing needs continuously.

For example, when we shared the TfL TravelBot with a restricted group we discovered that people were looking for a Tube map, something that we hadn’t thought to be a priority for our MVP.

The benefit of being able to respond to needs instantly meant that we were able to build and test this functionality before we had started any formal user testing or even launched an MVP product.

Understanding the Restrictions

Conversational interfaces have many benefits but also come with restrictions.

Conversational platforms struggle to display complex data so we need to understand when chat isn’t the best place and when to direct a user somewhere more appropriate.

For example, asking a question like ‘What’s the status of the District line?’ only requires a response of a few words, but if you’re asking for ‘The status of the underground network’ a live web view will display the information in a much more understandable way.

We will be exploring this with future iterations of the TravelBot where directing a user to a specific place within will provide a better experience but uses the benefits of a conversational UI to reduce the amount of time they spend finding it.

Future exploration

Conversational interfaces are great when the user knows what they’re trying to do, learn or fix. However, whilst conversations create a more focused experience, they also remove the ability for users to explore or discover things.

We need to think about whether this is the type of interaction that chat should be used for and whether or not we should be designing new types of conversations to facilitate discovery.

TravelBot is learning all the time so please keep the feedback coming.