As the UK went into COVID-19 lockdown, Tommy’s launched its first voice skill: Tommy’s Midwife, a voice skill on Google Assistant and Amazon Alexa to support mums and dads-to-be. Being able to offer this extra source of support in a very anxious time, when pregnant women weren’t able to have face-to-face time with their midwife, was brilliant.
One in four pregnancies sadly end in loss in the UK and Tommy’s mission is to change this by making pregnancy safer for everyone. We fund and carry out pioneering research into the causes and prevention of baby loss. And my team, the Tommy’s PregnancyHub, provides midwife-led information and support to women and their families, no matter what stage of pregnancy they are at or what their experiences have been. Ever determined to reach more people we set ourselves the challenge to use voice technology to deliver our trusted advice and information to parents-to-be.
Why voice technology?
14% of UK households now have voice-activated speakers and it is predicted that 50% of searches will be by voice this year. But when people talk about ‘voice’, they aren’t just talking about in-home speakers, such as Amazon Alexa and Google Home. Voxly Digital’s latest report, Time To Shine, indicated the usage of voice assistants has also boomed in the UK since lockdown. This is an important point as using websites can be harder for people with literacy difficulties, lower computer literacy or accessibility needs, or for those people whose first language isn’t English. And there has also been a huge rise in voice-powered voice assistants, such as Google Assistant, which has 500 million monthly users worldwide.
With so many people using voice technology we knew developing ‘Tommy’s Midwife’ skill was an important way of us reaching them and helping them navigate the huge amount of overwhelming and sometimes conflicting pregnancy information available online. Working with Voxly Digital and Mindshare UK we set out to build Tommy’s Midwife as a one-stop voice solution to bring this information into one place and alleviate any added anxiety caused by trying to find useful answers.
Helped by our team of midwives and our audience we developed 400+ answers to our most frequently asked questions, from mental health, nutrition, exercise and maternity leave, to complications, symptoms and labour and birth.
So how does it work?
The skill is a free-search voice app, which means that people can use natural language to ask a question, for example, ‘What do contractions feel like?’, ‘Is it safe to take painkillers?’ or ‘Can coronavirus harm my baby?. In order to do this, the team at Voxly built a bespoke, voice search engine completely from scratch. We had to be confident that this was working well, especially given the sensitive nature of the questions and answers we’re providing. Once we had decided to provide a Q&A type product, there were four key steps to the build:
- Auditing the content – What works on a website doesn’t always work in voice. We had to make all of our content voice friendly and accessible by voice.
- Building topic categories – Voice-ready content was sorted into five core categories, which are used by the user to help refine their question.
- Leveraging tags and utterances – We spent a lot of time looking at how we would tag the content to make sure we surfaced the right answer. We used primary and secondary tags as well as utterances that were loaded in and matched directly to the user’s question. This involved significant hours of user testing and refining.
- Using category and title refinement – For questions that don’t match an utterance or tags, we serve the user a range of categories or answer titles so they can choose the answer they want to hear.
Tips for people thinking about using voice
Ask users what they want – We had a firm idea in our minds about what we wanted to offer but when we asked our audience, they didn’t want that at all. They wanted to use voice to ask quick questions about pregnancy and get a short, useful answer from a trusted source – hence Tommy’s Midwife.
Build for Google and Alexa – We wanted our skill to be as accessible as possible so make sure to build for Google and Amazon devices. This means that the skill isn’t limited to people who have expensive home speakers but also for those with the Google Assistant app on their phones.
Add in-depth analytics and testing right from the start – We used Google Analytics and a dashboard to tag everything to make sure we could refine the algorithm during testing. This has been essential for reporting and for improving the live skill.
Use a live CMS – We can edit and amend any copy straight away if we need to. When the COVID-19 lockdown started to hit, we added information to the skill in a matter of hours.
Add feedback mechanics into the skill itself – The numbers of people using the skill is much lower than for a website, for example, so having user feedback to show the worth is important. From our statistics, 70% said they would recommend the skill, 75% said they found it useful and 80% said they felt supported by Tommy’s after using it.
This was such a fun and exciting project – but it was also a lot of hard work for a small team! When we first started, we had no idea how hard it would be to actually deliver the right answers to people – something we had to be confident in given the sensitive questions people would be asking. As a team who has always worked on websites, it was also hard to change our way of thinking about user journeys. But I’m sure that voice is here to stay so am looking forward to seeing the innovative ways charities embrace it in the future.
You may also like Why voice technology is becoming hard for charities to ignore.
Image: Mati Mango on Pexels