What’s next for voice emotion analysis?
In this latest series we are talking to tech leaders who offer solutions within the research, data, and insights space. To shine a spotlight on disruptive technologies and better understand the value they bring.
Phebi uses a combination of emotion and sentiment analysis to unlock new insights previously hidden in spoken words to understand what’s in people’s heads and hearts at scale.
Day One Strategy was a pioneer of voice technology within market research when we highlighted its potential in our 2019 EphMRA paper: Voice the next frontier. In this paper, we shared results of a pilot with Phebi, showing that voice open-ends provided longer, richer answers, as well as offer new insights on the emotions expressed.
But now to the interview…
Mike, can you talk us through the fruition of Phebi and how it came to be?
My background is in large-scale operational market research, but I have always been interested in the more innovative side of things. Initially, I set up a company called Cognicient where we used multiple data sources for brand tracking and we could see how AI voice assistants were becoming mainstream.
We thought there might be some interesting opportunities to use voice rather than typing to access System 1, immediate responses and to make surveys more interactive. This was the origin of the idea that my business partner @Konstantin Morjan and I had when we founded the company in 2019. We built a prototype in 2019 and shortly after won the IIEX (Insight Innovation Exchange) award for Innovation in Research Technology.
Why do think there is so much appetite for voice-enabled market research?
Firstly, it’s a better experience for participants, and a richer source of insight for the researcher. The non-conscious voice analysis can tell you something new. Text analytics can give you sentiment and emotion from words, but people don’t always mean what they say.
Often physicians say they’re interested in prescribing a brand, but the analysis shows that they are calm and relaxed. This can suggest they’re unlikely to act due to lack of engagement in how they speak.
Another reason is real-time use. We were reviewing a new treatment with a physician who said he was interested in prescribing, but the real-time analysis showed low levels of engagement. That allowed the moderator to challenge the physician who revealed that he didn’t have any patients, so they were never actually going to be able to use it.
Phebi helps interact with people and gives us an immediacy of insight and it’s been proven time and time again to be a great proxy for engagement.
And what about outside of health? Are you seeing the same traction there?
The plug-in tech is embedded in many platforms such as Decipher, Confirmit and Askia. Increasingly, we’re becoming these platforms partners for voice. And it is a broad range of sectors including health, finance, construction and even politics.
In a recent project for a construction brand, the client’s brand was getting a low engagement score. People had more engagement with competitors, and this was helping them grow their companies. This highlighted how they needed to be more brand driven.
What have been some of the challenges along the way?
People say they want to innovate, but not all want to learn something new, so we must be flexible in how we use technology.
Naturally, given the nature of voice or video, there are some concerns about privacy and where the data is stored.
Another challenge has been managing expectations. People expect Phebi to come up with an amazing insight. Many times it does, but it’s better when it’s used across projects to get a steady stream of learning rather than looking for one Eureka moment.
What advice would you give research agencies looking to convince clients about the value of voice analysis?
People say ‘I don’t need voice analysis to tell me someone is happy, I can read people myself’. But if you’re Finnish, can you really read people in Egypt? And can Egyptians read people in Germany? Can you read them consistently? There are instances where the answer is no.
There’s a paper that our partners at Shapiro & Raj conducted about violence in America. The people in the paper are saying these awful things, and yet they’re coming across as happy. We were thinking, well, there’s something wrong here, and you realise that they’re relieved to talk about it and get it off their chest.
You pioneered the use of voice, but how can you continue to differentiate Phebi from competitors?
We’re looking at call centres to understand what the caller feels and how the representative influences this. This allows call centre employees to answer appropriately and with empathy, which has opened up new areas for us. This is especially true when reps use scripts – it’s very difficult to learn why some reps are better than others at developing empathy with the caller.
We will continue to focus on platforms and also provide companies like Day One with a full-service offer to make it easier to work with us.
And lastly, other companies tend to have a more stand-alone, end-to-end offer whilst we are increasingly becoming embedded in the platforms – we see Phebi as a partner for voice rather than just a supplier.
Ben Lorkin – Senior Research Director at Day One Strategy
Mike Page – Founder & CEO of Phebi.ai
Phebi AI uses a combination of emotional and sentiment analysis to unlock new insights previously hidden in spoken words to understand what’s in people’s heads and hearts at scale.
Day One Strategy is a specialised healthcare market research agency that combines the latest technology with human intelligence to better understand customers’ wants, needs and behaviours. We start every day as if it’s our first, full of enthusiasm and a desire to try new things. If you feel the same and would love to leverage the combined power of tech and human intelligence for smarter, faster insight then get in touch: email@example.com