This article is being featured in Special Feature: Mind Meets Machine

Data analytics has transformed video marketing, replacing traditional methods like focus groups and surveys with a more comprehensive and data-driven approach. In a survey titled ‘Marketing in a Post-Covid Era’, the CMO survey (September 2022) found that nearly half of its respondents relied on analytics to shape marketing decisions (as opposed to 38 per cent before the pandemic) while 63.3 per cent studied online consumer behaviour. With the help of Artificial Intelligence (AI)  and Machine Learning (ML) tools, marketers can now extract even deeper insights from unstructured data like videos, audio and images, accelerating this transformative shift.

Hannah Chang, Associate Professor of Marketing, explains that before the technological breakthrough that enabled such tools, there was always a “bottleneck in data analysis” since the unstructured nature of video and audio data made it difficult for analysts to extract meaning. 

Human coders had to convert what she dubbed “this messy data” into organised formats manually, and this was a long and laborious process that involved creating categories and codes to represent different variables in a video, such as visuals, narration, music and text. With the help of ML algorithms, it is now possible to automate this process at scale, allowing businesses to uncover consumer behaviour insights based on countless datasets within seconds and at a minimal cost. 

The power of voice in video marketing

In her research, Assoc Prof Chang has uncovered a new insight into using narration in video marketing. According to her, the human voice has always played a key role in the communication of messages but most especially, persuasion. For a long time, practitioners and researchers saw voice as important because it influenced the listener’s perception of the narrator. For example, she shared, a listener might be persuaded by the confidence or warmth exuded by the narrator, based on the tone of their voice.

Even so, understanding how to optimise ‘voice’ in a video remains a challenge. She explains, “Every person’s voice is unique, because of the differences in their accents, pronunciations, and natural tones due to the structural features of their larynx and vocal tract.”

Assoc Prof Chang’s research has led to a discovery that the number of narrators can also play an important role in the persuasiveness of the content after she and her collaborators used ML tools to analyse large-scale, real-world datasets from a range of crowdfunding and advertising videos. They observed that the number of narrating voices had a consistent and significant impact on downstream behaviours, such as an increased likelihood of purchase. 

To understand this phenomenon further, they used an AI voice model to design synthetic voices that were comparable on different fronts. Participants of her experiment were shown the same video, with some watching a version that had different voices narrating the same message in the soundtrack, while others had a single narrator. 

The experiment revealed that marketing videos are more persuasive when they use multiple narrators, especially when delivering a clear message. The research team called this effect the “voice numerosity effect”. 

Assoc Prof Chang says, “People are not necessarily paying attention to the message that is being said to them when a marketing video is playing on their screen. They might be looking at the video’s aesthetics or texting friends. Yet when there is a change in the narrator’s voice to carry on the message, it unconsciously draws people's attention back to what is being said in the video and to think about this message. Our evidence suggests that this is how the voice numerosity effect works”. These findings underscore the importance of the human voice in videos as a potential strategic design element. 

Together with her collaborators, she is now looking at additional consequences of the voice numerosity effect to gain a more holistic understanding of this effect and its conditions for success.

A new toolkit for businesses of all sizes 

Assoc Prof Chang believes that the rise of user-friendly cloud analytics platforms presents a great opportunity for business practitioners to tap into AI and ML tools and gain valuable insights into customer behaviour. These platforms do the heavy lifting by using pre-trained models, making it easy for non-technical users to convert massive amounts of unstructured data into organised formats such as reports, charts and graphs. 

For instance, a company can upload past video marketing campaign data to uncover viewer trends like average viewing times and the impact of onscreen images or narration on audience engagement. Armed with these insights, business practitioners can make informed decisions to create more effective and captivating videos in the future. This straightforward process empowers businesses to unlock the full potential of their data and enhance their video marketing strategies.

Click the icons to learn more

Accessing public data and respecting privacy laws  

Some might assume that only established organisations with extensive data collection would be able to benefit from such tools, but Assoc Prof Chang challenges this notion. She explains that through the use of Application Programming Interfaces (APIs), data can be collected from public platforms like YouTube and Instagram. This means that even SMEs can access the diverse datasets needed to uncover consumer trends by analysing videos from external sources and identifying the best practices for viewer engagement.

She says, “With so many data sources and analysis toolkits publicly available these days, practitioners might find themselves overwhelmed by the number of options at hand. At this point, we are still keeping up with the growth curve, but in a year or two, we will be able to find more systematic ways to explore and utilise these online resources”. 

She also reminds businesses about the importance of respecting data privacy rules and staying informed about evolving international data protection guidelines. To ensure responsible data usage, Assoc Prof Chang suggests adopting standard practices such as obtaining explicit consent from users for data usage and aggregating information to prevent individual identification. By following these principles, companies can maximise the value offered by ML-assisted data analytics while learning about consumer behaviour responsibly.