UX in Chatbot Design

Views: 1656

For the past few months, the IT community has been busy discussing whether chatbots actually work and whether people truly need them. Skeptics often cite their own unsuccessful chatbot projects as examples while complaining about low retainment, poor conversion or short-lived conversations. Often the reason for these failures is that during development, they fail to take into account the UX features in chatbot design that are crucial for this particular channel of communication. So we’ve decided to share our approach to designing virtual assistants in terms of user experience. Our method is based on the projects we’ve done of our customers, as well as data we’ve gathered on the efficiency of the bots created on our platform.

Don’t Create False Expectations

Today, there’s no virtual assistant in the world that is capable of adequately maintaining a conversation with a human on any topic and solving any problem. However, upon first contact, a lot of chatbots say something along the lines of,

“Hello, I’m a chatbot, let’s talk.”

The user may form false expectations, which will lead to inevitable disappointment.

In their greeting, chatbots should immediately indicate what exactly they can help the user with. Like this, for example,

“I am a finance chatbot. Ask me about exchange rates or stock pricing.”

Don’t Exceed the Readability Threshold

As soon as the virtual assistant begins to bombard the user with huge walls of text, the contact with the user is lost. People love chatbots precisely because it’s way of communication which eliminates the need to read through long texts and explanations.

On our platform, one message is limited to 120 characters, but we recommend keeping the number of characters under 90.

Keep In Mind the Difference Between Devices and Instant Messengers

Chatbot-creation platforms allow for the use of several channels simultaneously: one virtual assistant can be easily integrated with Skype, Viber, Facebook Messenger, Slack or a website. However, this advantage has a downside – each channel has its own nuances. What works well on one website may be incompatible with other sites. In addition, a message can fit perfectly on the screen of one device but require scrolling on another. Designing chatbots requires the same testing process as working with webpages or apps.

During testing, you need to keep in mind that instant messenger users often have the keyboard active, so, in effect, only half of the screen is left for text.

Use Buttons

Although the chatbots’ ability to understand natural language is the developer’s main source of pride, it isn’t of great value to users. Understandable buttons allow the user to quickly get the desired answer, which is much better than a ‘natural dialogue’ which can lead to misunderstandings. For a effective conversation to take place, use buttons in the following cases:

  1. To offer multiple-choice answers for a limited number of options;
  2. To identify the user’s primary need;
  3. To specify the request in case of misunderstanding.

If you’d like to preserve the magic of the natural dialogue and at the same time make sure that the chatbot understands the user correctly, use the buttons for clarification.

“Did I understand correctly: would you like to reserve a room?”

Such a yes-no question will prevent unexpected incidents and significantly affect on ux in chatbot design.

Steer Users to a Clear Request Format

Every person writes and talks differently, and this must be taken into account in chatbot development. Suppose the chatbot’s system can react to phrases like “I want pizza,” “I want salad,” “I want cake,” “I want coffee,” and so on. If the user asks for a “lunch for three,” the chatbot won’t understand the request. In this case, the chatbot can tell the user what type of demand the bot is expecting,

“For me to make an order, type “I want” and indicate the name of the food.”

P.S. While surfing on the web we’ve run into a UX-designer’s blog named Max Glenister where he collected a list of resources for the user experience of chatbots. We reccommed you to take a fast peek – here