How I used traditional design sprint exercises to develop a CUI chatbot
Conversational UI is a hot topic right now. Before joining News Labs, I focused my efforts on the conventional web and app domain, much like most traditional UX designers. But one of my first conversations after coming into the team was about a breaking news bot that Tom, one of our developers, was building to trial for the UK general election.
Tom and I agreed that a bot would be a really exciting way to provide users interested in the minutiae of a story with up-to-date detail. For example, in a case such as the devastating Grenfell fire where thousands of people were affected, there would be a strong desire to get every update as quickly as possible. Our Live pages reload when updates are published, but they require navigating to the BBC website and remaining on the page to follow a story. The only push notifications we offer are through the BBC News app, and these are limited to high-level breaking news stories across all topics.
Tom and I agreed on a direction we could see working really well for the bot, but I felt that we were in danger of solutionising the problem. Instead, I organised an ideation workshop with everyone involved on the project. The workshop brief loosely encapsulated working with a messaging bot and breaking news, but the group ended up breaking out of these parameters — which I hesitantly endorsed.
I started the workshop off by talking through existing competitor products that currently permeate the news domain. We also discussed which platforms we could potentially look to build on. Facebook is currently the frontrunner with 1 billion monthly users having access to over 20,000 bots. However, 7 billion users have access to SMS, which could increase our reach to less developed countries. WhatsApp, on the other hand, is the biggest messaging platform worldwide; however, their bot functionality is currently limited. As a group, we agreed that using Facebook Messenger would be most beneficial due to the reach and current technical opportunities.
How might we…?
Our first exercise was the How might we…? Each participant wrote ‘How might we…’ and then finished the sentence with an objective they felt was the most important to achieve in this project. The objective here was to simplify the brief as much as possible to give the group something manageable to work towards. Once they had written all their HMWs, I asked them to vote for their favourite.
The winner was…
How might we reach an audience that don’t currently engage with the news?
Now that we had a brief, it was time to sketch. I gave each person an A3 sheet of paper and asked them to fold it in half three times, leaving them with eight sections. I gave the group ten minutes to sketch down as many solutions as possible relating to the brief, which they then had to present back to the group. I asked the group to be as creative as possible, and not to worry if their ideas weren’t feasible.
Once everyone had presented their ideas, I asked them to do the task again. By doing this once, we scraped off the ideas that sit at the surface of our creative consciousness. By doing it a second time, we dug deeper to find the treasure that sits within all(most) of us. The group presented back some really compelling and thoughtful ideas, ranging from gamifying the news to sharable quizzes.
I then divided everyone into three groups and assigned each group two ideas from the Crazy 8s exercise. Each group had to produce one ultimate idea, they were then asked to answer “The Beetle” questions.
- Who is it for?
- Why would people use this again?
- What is special about this idea?
- What happens if you reverse this?
- What makes this different?
- How could you grow this idea?
- How does the experience behave?
- Who does it remind you of?
We were left with three pretty great ideas to play with.
A breaking news video hosted on YouTube. Once the user finishes the video, an interactive poll determines which video appears next.
‘Graffiti sharing’. A user can chop and change certain elements of a breaking news article that they wish to share on social media.
Content specification for a Facebook Messenger bot. A user has the option to pick dark, light, fun or sad stories when they start a conversation with BBC News.
These ideas would admittedly be difficult to execute, but that was the point of this workshop: to get the team to collaborate, open their minds and think outside the box. It’s my job to take those ideas and rein them in to make them viable and functioning.
I built three interactive prototypes. The first idea centered around a user being sent a breaking news story in a video format within Facebook Messenger. This idea focusses on the onward journeys. When a user starts to watch the video, the onward journeys aren’t related to the video content. However, after the user has watched the video for longer than 30 seconds, the onward journeys change to related videos because we could then assume that the user is interested in the content, therefore they may want more information on that topic.
The second prototype was focussed around sharing. When a user wants to share a story, they currently have no option as to what they share. This idea gives the user the option to share certain elements of an article, e.g. if a user has read a quote they really like from an article, they can share only that quote instead of the whole article.
The third prototype was focussed around giving the user a choice of story type within Messenger. After the user reads a story, they’re given the option to either read the ‘next most important’ story, or a ‘quick and light’ story. The difference between these two options are what our editorial team believe is the most important story to read, as opposed to a lighter shorter story — most likely an entertainment article.
Once I’d built the three prototypes, I was keen to put these ideas in front of potential users. I went out to Oxford Street armed with my phone and a notebook and engaged in some guerrilla research with the public. I spoke to six people ranging from 23–36, male and female, about their device habits. I was keen to find out which messaging apps they currently use, if they’d encountered bots before and whether or not they found them useful. I showed them all three prototypes and we discussed what they liked and disliked about these ideas.
The headlines were:
- The more visual media content types from prototype 1 were welcomed more than others, but the change to onward journeys was barely noticed.
- The ‘Graffiti sharing’ in prototype 2 was deemed “cool” but most people said they wouldn’t use it.
- Prototype 3 got the best response: most people liked the idea that they were able to choose which story to read next.
To conclude this research, it was clear that although no participants explicitly rejected any of these ideas, there wasn’t necessarily a clear winner. Instead of reimagining and improving on the initial idea that we had before the workshop, it felt as if we were overcomplicating what is a very simple concept. With that in mind, we banked our ideas and started to build the simplest version of a breaking news bot that we possibly could.
In the initial stages of concept design, it’s important to learn quickly and efficiently. We spent no money on user research, we minimised the amount of the team’s time we required and we built very basic low-fi prototypes to work with. This saved time and money, allowing us to be more agile with further project development.
So where are we now?
We’re continuing to implement our first proposition. Although the ideas from the ideation session became fully functioning viable prototypes that could potentially be put forward, there definitely wasn’t a clear winner that I’d feel confident enough to push through into our designs.
Although these ideas aren’t being immediately implemented, it is important to go through the design process to validate the initial thinking. On the other hand, it’s important to save these ideas as you never know what’s going to happen six months down the line — so look out for GRAFFITI SHARING in the future (…or maybe not).