A successful voice app is one that your users come back to time and time again. But how do you keep them interested when the novelty has worn off? Simple – it all comes down to identifying your product’s ‘hook’ and connecting with your user’s state of mind. Here’s how.
A few weeks ago we revealed an exciting collaboration with the Shiken team, culminating in the release of an Alexa Skill. We’d already played around with our office Amazon Echo and created a number of quick, fun skills to get used to Alexa’s capabilities and limitations. However, the Shiken revision skill required a different approach.
Voice apps are fairly new to the market, so it’s all the more important to make them appealing and establish a pattern of regular use. We knew we had plenty of research to do, especially around how audience behaviour and user experience would differ between a web platform and an Alexa Skill.
Here’s what we discovered and how we used our learning to create a useful skill.
Nir’s habit-forming model
We knew that in order to create a successful Alexa skill, it needed to be a product that users would come back to after the initial thrill had worn off. We’re fans of Nir Eyal’s Hooked: How to Build Habit-Forming Products, so we looked at creating a ‘hook’ for our Alexa Skill.
Hooks are ‘experiences designed to connect the user’s problem with the company’s product with enough frequency to form a habit’. Eyal’s famous hook model comprises four key steps that successful companies follow to create sticky products that users can’t get enough of.
One of the first questions we had to ask was – why were people going to use our skill? What mood were they going to be in when they opened it? We needed to think about what the needs of our users would be and how we would satisfy them.
For example, uncertainty on what to eat for dinner might lead someone to open up a recipe book skill. If we could work out what the regular trigger for using our skill was, we could set about providing the solution. And if the user was satisfied with the outcome, they’d likely open the skill again the next time they had that problem. After all, a skill that the user would pick up on a whim won’t necessarily translate to repeat usage.
Shiken’s audience mainly comprises of young medical students. They’re short on time, always on the go and multitasking their way through daily life. There are often times when it simply isn’t possible for them to sit down with a text book or jump on the Shiken web app to revise. Therefore, our trigger was the need for quick medical revision material – fast and hands-free.
For the hook to be effective, users need to take a simple and efficient step to receive their reward. Voice technology could not make this any easier – the students only have to talk to Alexa to get what they want.
But first we needed to think about the utterances a user might use. Utterances are the specific phrases that people use when making a request to Alexa. These can be hugely varied — just think of the number of ways someone can ask you for the time.
This is where our flair for communication came in. When developing a skill, utterances have to be coded to tell Alexa what to expect. This meant us typing out dozens of very slight variations of questions and statements — basically anything we thought a user would actually say to get the result they want. For example, when asking for help someone might say:
- “Please help me”
- “What do I do?”
- “What do I say?”
- “How does it work?”
- “Can you help please?”
- … and so many more
Now the skill has launched, Alexa analytics tools are extremely useful because they can monitor what our users are actually asking. From that data, we can see whether we’ve missed any utterances and add them in before it becomes a problem.
The reward stage is where users experience the satisfaction of reaching their goal. For Alexa Skills, this can be the trickiest step to perfect as the value of the reward rests entirely on Alexa’s response. There are two ways that this can go wrong:
a) Your skill doesn’t understand what’s been asked and answers incorrectly. There is nothing more frustrating than Alexa failing to handle your request. Unfortunately, aside from adding as many utterances as we could think of, there’s not much else we can do to mitigate this issue. Luckily, this problem should diminish as Amazon improves its technology.
b) Your skill’s tone of voice/ personality is alienating. We made sure to add some personality to the Shiken revision skill. When revealing the user’s results, instead of sounding like a mindless drone, we had Alexa insert randomised quirky sayings such as:
- “Better luck next time!”
- “You’re getting the hang of this!”
- “Oh dear”
Despite this, Alexa’s voice can still sound robotic at times – especially if she’s reading out longer phrases and sentences where there should be natural breaks and changes in tone. However, through speech synthesis markup language, or SSML, we’re able to control her pronunciation, intonation, timing and emotion. For example, sometimes we wanted Alexa to whisper certain phrases or pause to take a breath for emphasis. All this allowed us to inject personality and quirks into the Shiken revision skill – making it both approachable and memorable.
This is possibly the most important stage in the model, as it requires the user to gives something back. The hook is only complete when the user invests something – passively or actively – back into the skill. This step ensures they will return to it.
For the Shiken revision skill, we enabled users to link their Shiken web account with the Alexa Skill. This gave users direct access to questions they had purchased on the web store and recorded their progress using the skill on their online analytics dashboard. This made it possible for the students to monitor their performance across all platforms in one place.
What happened next?
This exploration work gave us clear understanding of how the Shiken Alexa Skill would solve a problem for its users and the confidence to begin building the skill. It definitely made it easier to design the experience and stay focused on the end goal.
By identifying the killer ‘hook’ and designing the app around it, we can be satisfied that we’re helping Shiken’s users achieve their objective.
The next step in the process was to establish some design guidelines for the project – no mean feat when voice apps are such a new technology! We’ll be exploring what we did for this step in a follow-up post – keep an eye out for it on Facebook and Twitter.