7 reasons people aren’t using your voice UI
Voice interfaces are finally ready for prime time, thanks to mature technology and real user needs. But you wouldn't always know it looking at some products that are out there.
Based on our experience, here are seven common mistakes you probably need to fix to help your voice UI be simpler, more satisfying to use, and ultimately successful.
1) You’re not being clear about what it can do
When your input is just a microphone, people have no idea what your product can (and cannot) do. So they try crazy things beyond the capacities of your system and end up disappointed.
Think about framing your voice interface, through its visual design, its name, and the way it introduces itself. By having your voice interface ask a specific question (like ‘What clothes are you looking for?’) you’ll make sure that users focus on topics it can answer, rather than oddball questions like ‘Who is Keyser Söze?’ or even simply ‘When is the shop open?’.
2) You’ve ignored privacy and social contexts
Research is clear: people are happiest using voice interfaces at home, in their cars, and sometimes in the ‘privacy’ of a crowded street. At work, or in static public spaces like the train, it’s a different matter. Asking people to shout the name of the medical condition they need help with is unlikely to be a good idea in most contexts. Oh, and voice alerts don’t go down well in most situations either.
So if people aren’t using your voice interface, you may not have thought about privacy and social contexts. Are you expecting them to use voice in inappropriate settings?
Don’t confuse this with meaning voice can not be used for sensitive information. In the privacy of their home, people might prefer to dictate secret information such as their Netflix password rather than typing it on an awkward on-screen keyboard.
3) You don't care who is speaking
Home assistants allow anyone speaking to your device to quickly transfer money, read your calendar, or order things for you on Amazon. That may not be what you want. Even a TV programme mentioning such a case ended up triggering unintended purchases.
It's great that we're finally moving away from the hassle of logins and passwords, but voice UIs still need to behave differently depending on who's speaking to allow for proper usage control and privacy.
Make your system aware of who's speaking by identifying their unique voice signature. This technology is already here: First Direct bank uses voice signatures instead of passcodes. But remember to provide a backup option. You don't want to be locked out because you've got a sore throat.
4) You don't remember anything
Most conversations involve a series of related statements or commands (‘What was the last message from my husband?... Send him a reply.’). Voice systems that lose track of the conversation (answering ‘Searching the web for Send him a reply...’) are infuriating, as we’re used to people being great at keeping track of the thread of a conversation.
Make your system smart enough to keep key information in mind, and build on it to understand future queries. So when you ask your voice interface to "Show me the shirts on offer", it shouldn’t fall over when you say ‘Just the blue ones’.
5) You rely only on voice.
There are tasks that a purely voice UI will never be good at. Providing a long list of options by speaking them out, one after the other after the other will always be boring and unusable.
Consider mixing voice and graphical UI, using one's strength to compensate the other's weaknesses, to easily display many options in a manageable manner. Take Amazon’s Echo device - it has a companion mobile app so that when you want to see the shopping list you’ve built up through voice, instead of it being read out painfully item by item, it can be viewed and edited on screen.
6) You don't think about the audio context
If you are in a noisy environment like driving a car, a factory floor, or listening to loud music, your system will struggle. It's obvious but often forgotten: voice UI require the product to be able to distinguish the voice clearly.
And if you’re designing voice interfaces, consider explaining why the system is having trouble understanding (‘Sorry, I couldn’t hear that clearly - it’s noisy right now.’). It helps users understand why there’s a problem and how they might fix it.
If you control the hardware, more complex audio capture systems would work better, as Amazon did with the Echos’ seven microphones array. If the only audio input available is a smartphone mic, consider providing an alternative interface, either physical buttons or on-screen.
7) You've removed all alternatives
You want to increase the setting on your ‘smart’ radiator, but the wifi is slow and Siri’s not responding. So what now? I guess you just have to freeze.
Smart homes will allow for greater energy efficiency, enable automations so you have to do less, and can magically predict your needs. But your home should be usable at all times - even when there’s a technical glitch. And you need to be sure that your guests can turn on the light in the spare bedroom without needing to install an app.
So no matter how exciting your new experience, leave an obvious, reliable, ‘non-smart’ way to control your device, whether it's about increasing the heating or…opening a pod bay door.
While voice technologies are ready for prime time, getting people to use them is not a technology problem. It begins with understanding how to get people and machines talking.