Follow Us on Google
United Kingdom-based Nigerian researcher, Michael Usiagwu, has examined the growing use of Small Language Models in marketing analytics, highlighting their potential to process large volumes of data while warning against bias, inaccurate results and ethical risks.
Usiagwu, a researcher with backgrounds in marketing and accounting, conducted the study with Lawal Muez of Ladoke Akintola University of Technology and Johnson Chinonso of Kwara State Polytechnic.
Their research was published in the International Journal of Science and Research Archive in 2025.
The study, titled, “Small Language Models in Big Data Marketing Analytics: Addressing Bias, Accuracy and Ethical Challenges,” examined the growing application of SLMs in processing large amounts of marketing data.
The researchers noted that the rapid growth of big data had increased the need for technologies capable of analysing large and complex datasets to help businesses understand consumer behaviour and make decisions.
They described SLMs as lightweight artificial intelligence tools capable of processing textual information with relatively lower computational demands than larger language models.
“SLMs offer a practical solution for analysing large volumes of textual data, such as customer reviews, social media content, and survey responses, empowering marketers to derive actionable insights with greater speed and efficiency,” the researchers stated.
According to the study, the technology could be useful for sentiment analysis, trend detection and customer segmentation, particularly in situations where organisations require rapid analysis of large amounts of information.
The researchers, however, cautioned that the use of SLMs could also expose organisations to problems arising from biased training data.
Other News
They stated that such biases could affect marketing decisions by producing skewed customer profiles, reinforcing stereotypes or misrepresenting the preferences of some consumer groups.
“SLMs often rely on vast datasets for training, which may contain inherent biases. These biases can skew analytical outcomes, leading to misleading insights in marketing strategies,” the researchers warned.
The study also examined the accuracy of SLMs, noting that their lightweight architecture could limit their ability to understand complex or context-sensitive information.
According to the researchers, although SLMs could identify basic customer sentiment, they could struggle with complex emotions and multifaceted patterns of consumer behaviour that require deeper contextual understanding.
The researchers further raised concerns about privacy, transparency and accountability, saying organisations needed to ensure that the use of SLMs did not compromise sensitive customer information or make decisions that could not be adequately explained.
“The lack of clear explainability in model outputs can undermine trust, while improper handling of sensitive data can lead to compliance violations and reputational risks,” the study stated.
Usiagwu and his co-researchers used a mixed-method approach involving experimental testing, literature reviews and expert interviews to examine the performance of SLMs, their limitations and ways of improving their deployment in marketing analytics.
The researchers recommended measures including balanced data sampling, fairness-aware training, continuous auditing, improved transparency and stronger safeguards for privacy and accountability, saying these would help organisations maximise the benefits of SLMs while reducing the risks associated with their use.
