A multinational extractives company with a workforce of tens of thousands operating in several countries and continents.
It is a constant challenge for the company to keep in touch with employee sentiment in a business comprising a large multilingual workforce performing a wide range of roles and often in very isolated locations. Faculty was asked to help design a digital employee engagement campaign that uses data science to build a more sophisticated understanding of the drivers of employee sentiment.
Traditional survey techniques tend to ask people to give discrete answers to questions. These surveys usually provide quantitative data, particularly where respondents are asked to score answers on a scale. We developed a program, based on natural language processing (NLP), that turns unstructured answers into structured data.
To exploit this new capability, we asked a series of deliberately open questions, such as ‘What makes you proud of your work?’ and ‘In what areas could we improve?’ Employees were free to raise whatever was on their mind, rather than simply responding to management prompts. We used NLP and clustering models to extract patterns in the employees’ responses.
By using data science, we raised the quality of the employees’ answers and made their responses more informative. Our approach allowed the management team to gain a much clearer understanding of what mattered to employees. For example, management expected that pay would be the biggest driver of engagement. But NLP modelling showed that when we clustered the responses employees gave to open questions, they were more interested in talking about the quality of relationships with their team.
Uncovering these insights opened up a new approach to employee communications, focused much more closely on what actually matters to employees. This went on to increase overall engagement, as measured by an employee net promoter score, by 8%.