Wouldn’t it be great if your Chatbot wasn’t waffling?
It was August 2018 when a simple convo between Tolani Adekoya and I turned into a chance do work on a fun side project.
I had already noticed that my mobile network provider, a couple of my banks and a parcel delivery company I used were getting me to use their Webchat/Live Chat services; incentivising me through quicker responses. I really wanted to answer the question of what a good experience feels like.
After a few discussions about what we were trying to achieve, we agreed on the research objective and got things moving with most of our collabs happening remotely in the evenings and weekends.
We will use UX design principles to design an experience which helps travellers book excursions for their holiday
So why do businesses invest in Chatbots then?
Chatbots are one way that businesses can deliver services to people at a lower cost, increase the chance of making a sale and do transactions without requiring a real-time human presence.
In my experience (and many others), the current experience using Chatbots is not endearing, especially when it feels like you’re being spammed.
RESEARCH
STEP 1 — SURVEY
Getting respondents was easy, we knew the people who travelled thanks to social media, so we sent out this Typeform survey which allowed us to get a sense of whether people preferred to make their own experiences or whether they preferred booking holiday packages. It also gave us an idea of who to follow up with.
STEP 2 — USER INTERVIEWS
Having got a feel of what the mixed responses were, we wanted to dig deeper into the answers so we conducted semi-structured interviews and with their permission, we recorded and transcribed the discussions for further analysis.
Note: I’ve done 80 + user interviews in the past year, and I always learn new things when I do them. But this is was the first time I had really transcribed them — you do pick up some gems here, so I would advise doing them if you want to do them well. I also found out about Descript which gives you 30 mins free. After that, you gotta pay though.
STEP 3 — HIGH-LEVEL THEMES
At this stage, we now had qualitative and quantitative insights so Tolani and I met up one evening in Shoreditch and used thematic analysis to categorise our insight into themes. It was a time-boxed exercise to we gave ourselves 2 hours (and managed to squeeze in banter).
STEP 4 — PERSONAS
After our meet up we felt like we had enough there to start building out personas.
We had enough insights to explore a set of 3 personas. Personas evolve over time and really help in understanding who you are targeting and supporting your product, marketing and sales efforts.
DESIGN
We had now done surveys, interviews, categories everything into themes and built our personas. Now we were able to construct the first version of what the story would look like for each type of traveller.
This is where we can begin to explore the services we could offer to users and consider how we would deliver it.
STEP 5— STORY MAPPING
Helps you envision the product/service as a series of tasks for the protagonist to complete
Designing for Tina is all about guiding her along her booking experience, delighting her by showing her multiple destinations she can visit (based on proximity and budget) and learning more about her preferences as time goes on. My hypothesis is that at the beginning of her search, price is an important consideration, so as you can see it is the first Goal listed. Tina can quickly get an idea of if what she is looking for is affordable and use the helpful filters on airports, dates and accommodation; whilst we continuously show her multi-destination offers that she can snap up.
Whilst the example below is not exhaustive, it does illustrate how we can craft a chatbot experience that is relevant, focuses on getting Tina to her desired outcome and consider the information she would want to know.
Good dialogue involves listening, so whilst this is a good example of a users booking journey, in practice, I would measure how visitors progress through the funnel and improve the Chatbot with any gaps once we analyse how conversations go.
STEP 6— WIREFRAME/PROTOTYPES
Admittedly, this is the part I enjoyed the most (I love shipping 🚀)
So now we can start to build out a dialogue that asks important questions early, and considers the fringe cases.
Tina’s is a frequent flyer, so it’s important to kick off the conversation intentionally and get her to the point where she can see if you can help her or not, quickly — that’s her desired outcome. The appropriate experience is about how Tina needs to achieve it. Sidenote: there’s a great article about this here
For instance, a good question to ask is, what response do we give when Claire clicks “No”, she is not ready to see her travel options. She might be trolling the bot, but like a good politician, you should always have a good answer.
LEARNINGS
- Time-box these experiments. It’s easy for these to drag and perfection to happen when you don’t set a deadline.
- Transcribing user interviews really helps you to uncover keywords in user interviews.
- Chatbots feel exploratory. Whilst, I would be inclined to keep it open-ended, it feels like the job of the Chatbot to guide the conversation and have human-like responses for fringe cases
If this has made you think more about how you design products and even Chatbots, hold the 👏 at least 10 secs! Have a good week!
