Work
Enhancing market research and uncovering deeper insights for a research company
We built a dashboard for an innovative market research
company to gain real-time insights from survey responses,
and quickly present them to its customers.
The model is deployable across all the company’s surveys worldwide, helping the company to provide richer insight with a rapid turnaround time.
Customer
An innovative and fast-growing market research company that provides insights for some of the world’s biggest brands, operating across 150 different markets.
Problem
The company collects a growing amount of free text data every day. Allowing respondents the freedom to write what they like, rather than binding them to choosing between predefined answers, has the potential to uncover a wider range of subtle insights. However, it is challenging to work with this kind of unstructured data, especially when it wants to provide insights on emerging trends to clients in real time. To help overcome this challenge, the company asked Faculty to help it enhance its innovative market research product with advanced Natural Language Processing (NLP) capability.
Solution
The objective of the project was to apply NLP techniques to 15,000 free text survey responses to identify trends relevant to one particular client, a leading global drinks supplier.
First, we built a supervised learning algorithm that accurately matches survey responses to pre-defined trends identified by the client. The model used was a multi-class classification algorithm that clustered responses into 40 categories.
Second, we applied advanced unsupervised learning algorithms to the data, to provide truly data-driven insights. This means that rather than assigning responses to pre-defined labels, they created their own by clustering responses into different distributions across topics. This allowed the data to truly tell the story, which resulted in insights to the client that were both surprising and actionable.
Because these insights reflected the unprompted thoughts and wishes of customers, not the pre-defined categories of received wisdom in the sector, they provided the client with a level of insight its competitors do not have.
Impact
The output of this work was a dashboard used by the client to gain real-time insights from survey responses, and quickly present them to its customers. The model is deployable across all the company’s surveys worldwide, helping the company to provide richer insight with a rapid turnaround time for any client.