14-Dec

UX

Tips & Tricks for Better UX in Voice Interfaces

4 min read

·

By Ingar Røsok Kvalheim

·

December 14, 2019

Me: “Hey Goog..”\ *Bang* The sound of my girlfriend’s palm hitting the light switch, and the room goes dark. With a tear from the corner of my eye I go to sleep, well aware that I won’t be turning the lights off with my voice tonight either. \ \ A true and heart-breaking story from my life, and one that I guess many technology lovers out there can relate to. Because even though smart speaker sales numbers has been going through the roof over the last few years, my impression from talking to friends and reading articles is that voice user interface (VUI) still struggles to find its place among the mainstream user. One of the main reasons for this is bad UX. Nielsen Norman Group formulated it well comparing todays voice assistants with mobile phones from year 2000.\ \ I know the struggle. I’ve made two voice skills for Amazon Alexa, advised on voice projects at work and read a lot of articles. So, I’d like to summarise a couple of important things to be aware of and some tips and tricks when making a VUI:

--

  • Ask yourself “Why is VUI the right interface for this solution?”. This is for example not the best interface for selling paintings. In my opinion, one of the biggest problems for VUI so far is that it needs to prove itself useful and not only “nice to have”.\ \ Tips & tricks: People use VUI when their hands are busy - for example to set a timer when cooking or to find a song while driving. And they use it when it’s faster to ask a question than trying to find the answer by themselves - for example getting the outdoor temperature or the name of a capital.

--

  • Speaking and listening share the same short-term memory and working memory as recall and problem solving. So thinking while speaking requires quite an effort. The consequences of this is that people struggle remembering answers, options and commandos voice assistants give them.\ \ Tips & tricks: You have to prioritise hard on the amount of information to give the user. Too much information, and you lose your user. This is mobile first level 100. Also if you have several options for the user, don’t give them more than three at a time. Let the user know they can ask for more options instead if needed.

--

  • “What can I do in this VUI? And how do I do it?” These are questions your users will have.\ \ Tips & tricks: Offer a tutorial for first time users. Offer guidance, for example when users say “help”. If you have other interfaces explain the VUI here as well. This is also good marketing, people won’t find your VUI by themselves.

--

  • There’s a lot of questions to consider when it comes to branding. How does your brand sound? Male or female? Funny or serious? Old or young? Excited or calm? One voice or several voices? Any other sounds including voices? \ \ Tips & tricks: If possible, use a human voice instead of a built-in robot voice. This will create a more engaging experience. The use of sounds is also a powerful tool for evoking feelings and create engagement.

--

  • Words matter more than ever. The chosen words are the ones that dictates how the user understands your product, and that shapes their experience.\ \ Tips & tricks: End your sentence with a question, and avoid placing it in the middle. People tend to answer the questions immediately. This is, by the way, not the time and place for those rhetorical questions trying to pretend to be a great author.

--

  • Don’t tell the user what to say. Like the old phone bots telling you to press 1 for this and 2 for that. I can’t imagine anybody associating that with a great experience. \ \ Tips & tricks: Try to make a normal conversation. Ask questions instead: “Do you want this or that?”

--

  • Unfortunately, a lot of things still go wrong when using VUIs. This is well documented both in studies and in user testing.\ \ Tips & tricks: Blame the assistant, not the user, when telling the user that something went wrong. Test your VUI to see if it’s some place where the voice assistant easily gets the response from the user wrong. Especially in Norway responses easily gets misunderstood because of all the dialects. Offer a phone call to a real person or information of where to get help if something goes wrong several times.