The conversational interface is not a new platform. It’s slowly merging into our lives. The most common interface would be the Interactive Voice Response(IVR). When you dial the number, it directs you to specific categories. Then it dives in deeper to narrow it down where the user needs help.
For example, if I call a phone service, they ask you if you speak English or Spanish. Then if I select English, IVR asks if you need to know your payment, plan, store location/hour, or talk to an agent.
The conversational interface for voice assistant and chatbots work very similarly to IVR.
For our UX assignment, we had a chance to conduct a travel prototype with Alexa. Building a wireframe for a site is a lot easier compared to the conversational interface.
First of all, there are not many user case studies compared to the website. Also, when I asked people with Alexa, they only use basic functionalities. Such as asking for the weather, or playing music.
Same as designing an app or the site, the first thing first is come up with a How Might We statement. How might we plan a school trip with 20 students and adults to Washington D.C.?
Storyboard
Then we started building a storyboard. The actual storyboard that illustrates where the conversation is happening, what do they talk, and what would they address. By imagining the conversation, it helped me out what questions I would ask if I was with them.
While building a storyboard, I was aware of the consumer marketing funnel: awareness, familiarity, consideration, purchase, and loyalty. Since our goal is to figure out how to enhance the customer experience using a conversational interface. We did not need to use all five steps to direct users to be loyal customers. Also, at the purchase level, people would not say out loud their credit card numbers. So when it comes to this funnel, I was considering the first three steps.
Awareness
To let customers know what we do, we have to show or tell them what we do. It can be through word of mouth, SEO, social media ads, display ads or email newsletters, etc. To connect awareness and Alexa, I integrated a question such as ‘Can you recommend historical cities in the U.S.?’ The question may not seem to relate to awareness. But if the Alexa answers the article/blog post from our client, then it builds awareness by clicking into the site and reading about the article.
Familiarity
Once they engage with our site, the pre-customers need to familiar with using the site. Then they can ask, ‘How can I plan a school trip?’ Of course, we have the information on site. So we can directly give them what we offer and show them different plans to plan a trip.
Consideration
After they are familiar with the site and the plan, the pre-customer would like to consider the price. ‘How much would it cost to go on a school trip?’ to get a quote on the current site, the pre-customers have to put their contact information. Such as their name, email, address, school name, school address, phone number, and professions. If the pre-customer is on this stage, they will request for a quote. This is the final step that Alexa would direct them to this page.
After the storyboard
To build this prototype in a real-life, I had to think of different utterances that people might say but have the same results from Alexa. Here is a screenshot of my spreadsheet for one output. If the user asks blah blah blah, then Alexa will answer blah blah blah.
This utterance can range between hundreds and thousands. Also, Alexa can answer one at a time. I can’t ask ‘Can you find a historical city in the U.S. for a school trip?’ For Alexa, a historical city is one variable and a school trip is another variable. As you see in the utterance sheet, there are too many ways to say one thing. If the variable change, Alexa would not be able to give you the answer and would recommend you to speak in a different way. This is where the frustration comes in.
Conclusion
I honestly hated Alexa. Because she can’t understand or does not give a relevant answer that I need. Also, I don’t like the fact that I have to talk the way that Alexa would understand.
When I Google search, I don’t ask word by word. I use a fragmented sentence and use keywords that would get the best results. But while I was building this prototype, I realized how much effort is necessary to make a conversational interface. Just writing out different utterances was a lot of work. Thinking about different problem space and that everyone has a different question is such a frustration.
That doesn’t mean I would use Alexa from now on. But at least I understand how it works and how much effort is put on for this conversational interface.