Designing for Voice: 7 quick tips for designing great Conversational UI
Alexa recently joined my family and each day that goes by she plays a bigger part in our daily life.
She wakes us up in the morning, tells us the weather, reminds us of our appointments, offers a quick news update, checks that transport lines are in order and then plays music while we get ready. She switches on the lights for me when I come home, guides me through recipes while I’m cooking and even settles disputes between me and my husband on everyday facts and trivia (prompting a “See, I told you so”).
Every day we try a new “skill”, but I have to admit the experience of using some of these skills was so frustrating. This is because Voice User Experience (VUX) design or designing Conversational User Interfaces (Conversational UI) is actually very different from designing experiences for screens.
We are now designing for dialogue, and this requires a different set of skills.
Here are 7 quick tips for great VUX:
1. Rethink your information hierarchy
Screen based applications have a specific hierarchy of information. Users often begin their journey on a home page or menu button and then drill down to the information they’re looking for. But when it comes to Alexa, users can ask for this information right from the start e.g. instead of filling out forms online or searching for specific trains on an app, users can simply ask Alexa ‘What time is the next train to Wimbledon?’
Instead of just focusing on finding out information, you may want to consider what the user wants to do with that information. For example, if the user’s asking for train times they may want to buy train tickets, check the weather at their destination or book a taxi to the station. Instead of creating a standard website Information Architecture (IA) build a decision tree to script the different paths, or “flows”, the users might choose to go down.
2. Design for how people talk, not how they type
We’ve developed a specific way of instructing tech. For example, in a search you would type “best fish restaurant in Athens”, but if you were to ask someone you would say “Do you know any good fish restaurants in Athens?” This changes again slightly when you talk to Alexa e.g. “Alexa, find me a good fish restaurant in Athens”. This speech pattern uses natural but direct phrasing, like talking to a personal assistant. Therefore, good VUX design should reflect language that’s both conversational and transactional.
3. Recognise a variety of questions and commands
No one wants to memorise hundreds of commands, so your Conversational UI needs to cater for the many different ways people speak. One person will say “Alexa, order me a taxi” another will say “Alexa, book me a ride”. Write up all the different ways someone might phrase a question or a command and make sure you include these in your recognised phrases or, as the Alexa developer site calls them, “utterances”.
4. Keep Alexa’s responses short
Longer responses tend to be more difficult to follow and remember, so Alexa’s responses need to be kept short and concise. It’s also important to avoid prompting users to answer Alexa’s question with another question as that tends to confuse her and your user will end up in a frustrating loop.
5. Try to stick to a maximum of 3 choices
When Alexa responds to a question it’s important that she only offers a maximum of 3 choices, otherwise it becomes really difficult to remember them all. Numbering these options can help. Recalling the number of the option you want is a lot easier than recalling the option itself, which could be a very long sentence.
You can also reduce pressure on indecisive users by offering an option that gives them more time to decide. For example, when using a recipe skill Alexa could ask if the user would like to “hear the details” giving them more time to decide whether they want to start cooking or search for a different recipe.
6. Add variety to your answers
None of us like to hear the same repetitive greeting or response every time we ask a question. By diversifying her answers, Alexa feels less robotic and more natural e.g. “Thanks”, “Got it”, “Okay”, “Great”, “Sure”. Variety is also a great way to add personality.
7. Personalisation is important
Have Alexa remember preferences so users don’t have to input the same information every time. This will add to the ease and efficiency of the user experience. For example, Alexa can remember users’ routes to work so when they check the trains each day they don’t need to resubmit their location every time. Alexa can even remember delivery addresses and will simply ask to confirm it when users place repeat orders.
In the future, I’m curious to see how Alexa solves issues around privacy and fraud. Imagine your children talking to Alexa and re-ordering tons of toys. Password-locking is one solution, but how do you make sure people don’t overhear it or any other confidential information? Maybe in the future Alexa will be able to recognise individual voices and adapt her responses depending on who she’s talking to.
It’ll also be interesting to see how Alexa deals with commercial bias. When we search online we’re used to seeing hundreds of results, but when we search with Alexa we expect her to intelligently pick the best answer for us. Imagine asking her for the best shampoo brand. How would she provide just one answer without being biased? I imagine search engines, like Alexa’s favoured engine - Bing, will have a big part to play in figuring out how they pick the right results and content for voice-based search.
It’s clear we still have a lot to learn, but Conversational UI is an exciting new way of interacting with technology and will definitely play a part in the future, in homes, hotels, classrooms, health centres, cars and beyond.