Analysing qualitative feedback is essential for addressing issues, understanding customer experiences, and guiding targeted product design. However, manual analysis of large open-ended datasets is time-consuming and can lead to superficial insights, which conflicts with the needs of an agile product environment.
This session will share insights from piloting LLM-based automated customer feedback analysis at MYOB. While LLMs enabled rapid first-pass categorisation, challenges like hallucinations and inconsistent output require their results to be interpreted within a broader analytics framework. Overall, LLM methods have shortened time-to-insight for qualitative data and are increasingly contributing to customer-centric insights at MYOB.
Key Takeaways
- LLMs enabled more sophisticated analysis of text data, unlocking untapped value from existing sources
- Automating thematic categorisation across large datasets has significantly reduced time-to-insight for product and design research teams
- LLM-based categorisation offered a fast first-pass analysis, but the results need to be integrated into a wider analytics framework
- Large-scale analytics with current LLM technology require pipelines to transform qualitative insights into quantitative metrics
- AI technologies enabled democratised customer analytics and cross-functional collaboration
————————————————————————————————————————————————————
Bio
Aaron Le Compte- Principal Data Scientist | MYOB | Australia
Aaron Le Compte leads the Data Science capability at MYOB, bringing over 15 years of experience in technology to the role. His career spans various fields, including research, development, data strategy, and the application of AI/ML technologies to solve complex customer problems. Aaron has worked across diverse industries, including biomedical technologies, social networks, and startup ventures, delivering end-to-end solutions that blend cutting-edge tech with business strategy.
ANZ-Stage 2 2024
Aaron Le Compte- Principal Data Scientist | MYOB | Australia