Voice and conversation UI design best practices (+ PDF cheat sheet)
Conversation design is right here for some time already. Starting from chatbots to Google Assistant and Amazon Alexa making its way into more specific domains of our day-to-day life. Still, too often than one might imagine, the conversation fails, despite the fact that it has been designed by humans for humans. Reasons vary but perhaps, the most common is lack of investigation of human communication rules.
Today I will share the best practices of VUI taken from real apps where designers spent considerable amount of time on testing and learning. This tutorial is super easy and compelling made in the form of Dos and Don’ts with the description of each case. I also compiled a short and clear-cut PDF with examples illustrating each case (see the end of this tutorial).
TL;DR: Download the PDF cheat sheet with all the examples from this article put in short.
General
Be brief and clear
Since you design the conversation on paper first, do not forget about the fact that your users won’t be able to read (or go back to that needed point) the text. And we, humans, have quite limited short-term memory.
Diversify your language
A welcome phrase (or any other phrase repeated many times) might become a nuisance. So, write several variants of the phrase and teach your app to adapt to the context; this will provide more natural flow and add common sense to the conversation.
Greetings
Three main steps should be accomplished in your app at the stage of greeting a user:
- Be polite and welcome a user.
- Explain what the app does
- Turn the floor over to the user
Be polite
Don’t forget to use those magic words which we were taught as kids; they do magic and facilitate communication process by making user more loyal to the errors occurring during conversation.
Tell users what the app does
At first users can only guess what exactly the app does; and different users naturally have different mindsets. The best way to avoid first touch frustration is to tell users what the app does. And obviously, the faster you relieve the stress the better. So, do it straight after a welcome phrase.
Turn the floor over to the user
Make it clear for a user that now it’s their turn to start acting. And a properly asked question is the best tactics for showing users it’s high time to act.
Avoid over-promising
Basically, a user decides whether to proceed interacting with an app or not during first exchanged phrases. It’s critical to set clear value from the first seconds of interaction. Do not stumble users with overpromising since it may lead to wrong requests (something your app wasn’t ready for) and as a result to errors, miscommunication, and disappointment. Explain what to expect from the start and don’t be afraid to seem too simple. The less an app does the better it works, the clearer users see the value.
Mind the context
If you are so lucky to get a user come back to your app for more, adequate reactions of the app based on users’ invocations will be very handy. It’s really dull and annoying to hear the same phrase all over again, especially when a user changes the request.
Prompts
User’s response is needed — make it clear
First of all, you need to give users clear understanding that the system is waiting to their move. And presenting the options only doesn’t mean anything.
Help forming an anticipated answer
The most frequent reason for dialogue error is a question asked in the wrong manner. Obvious, as it seems, the system is trained to understand exact key-phrases; so, when a user picks a “wrong” phrase the whole dialogue fails. Make sure, you put a user on the right track for proceeding a dialogue.
Giving options of answers is a must
When the number of options is limited.
One question at a time, please
This is correct for human communication and it’s the only possible way for human-machine communication.
Questions last
In writing we start from the most important statements. Everything is structured when put on paper. You can go back to that point, reconsider it. In spoken language, however, we tend to place the most important info in the end of the speech.
Informational Statements
Prepare users for consuming a piece of information
When seeing a person next to us we’re generally tuned into the conversation. Talking to a computer or a smartphone implies different mindset. Users expect more formal way of communication from a machine, which actually makes it easier to understand.
Use SSML
No way around it. Machines were taught to talk to imitate human interaction. The only way to make machines more human and the conversation more real is to use SSML for speech mark-up.
Outro
And this is really just the icing on the cake. The topic of VUI is so large thanks to the pace of tech progress that I can see as far as at least three more articles on this subject. So let me know if you’re interested in the next part by clapping (you can clap several times!).
I also made a PDF with all the examples from this article put in short. Learn more — do better!
However, these are just common insights based on a lot of different apps tests. Each product is unique in terms of users it aims at. So, one of the most integral parts for a successful app launch is testing with real users. This is the only way to make a truly memorable experience via your product. And I assure you, you will be amazed by the answers and reactions real people (not your imaginary users) will show interacting with your product.
Fabble (formerly Tortu) allows for prototyping and testing your ideas at any stage — which is crucial since it enables you to save time, effort, and money. Welcome to the community of those who care about user experience.
