Common Voice is a crowdsourcing project started by Mozilla to create a free database for speech recognition software. They discovered that voice assistants & other software were more receptive to western males. To achieve this, they approached Digital Business Lab to collect voice data clips from a wide range of users across various social media platforms in the APAC region to help build up their dataset of voices.
Awareness & Conversion
Digital Business Lab produced a series of social media visuals and videos with a strong female focus combined with softer tones to attract a predominately female user base to donate their voices to Mozilla’s platform.
We have also created localised content in each target market’s native language to be more personable and relatable while running ads across different platforms, such as Facebook, YouTube, Yahoo!, and WeChat.
Finally, we analysed and optimised the weekly performance of the campaign to ensure that Common Voice was generating the target hours of voice data daily.
Content Strategy - Social Media Content Creation
🎙️Digital Business Lab generated over 43 hours of voice clip data in two months despite Mozilla Common Voice’s limited brand presence in the chosen regions. The campaign performed well through our understanding of market dynamics as we utilised diverse social media platforms across the APAC region. The achievement is supported by the plethora of experiences credited to our in-house team of experts per market.
💡As a result of Digital Business Lab’s continued optimisation efforts throughout the campaign, we learned that:
- Voice donations usually occur on mobile devices
- Notably, older generations (ages 40-60) are keener to participate
- Users prefer more technical content than creative messaging when trying to grasp our budding brand
About Common Voice (Mozilla)
Common Voice, from Mozilla, is a publicly available voice dataset, powered by voices of volunteers around the world. People who want to build voice applications can use the dataset to train their machine learning models. At present, voice datasets are also underrepresented, this means voice-enabled technology doesn’t work for all that many languages and where it does work, it may not perform optimally. Common voice strives to change that by mobilising people everywhere to share their voices.