Creating a Happier, More Productive Workforce: How Most Loved Workplace® Uses Data and Machine Learning to Drive Change

Because employees play a vital role in achieving organizational success, companies must continuously strive to keep their employees engaged and motivated. Research at Most Loved Workplace® indicates that fostering emotional connectedness is the key to achieving this goal. This paper describes a new tool for listening to employees' feedback that supports a culture of feedback and predicts employees' love of the workplace and organizational commitment.

The relationship between employers and employees has changed in recent decades due to shifts in workforce requirements, generational values, and where and how employees work. This already tenuous connection was further strained by the COVID-19 pandemic, which generated unprecedented levels of burnout that led many employees to reconsider the role of work in their lives. As a result, productivity reached a historic low, with 2022 seeing the most significant annual decline in US labor productivity since 1948.

Emotional connectedness describes genuine love towards one's workplace beyond mere affective attachment (Meyer & Allen, 1991) or satisfaction. It is a mutually reinforcing dynamic that benefits both employees and the organization: employees experience a sense of belonging and psychological safety, which is the shared belief among team members that it is safe to engage in risk-taking (Edmondson, 1999). In return, employees provide loyalty, intense initiative, and goal alignment. Furthermore, emotionally connected employees are predicted to be more engaged and willing to put extra effort into doing their work with consistent feedback (Goldsmith & Reiter, 2007).

Most Loved Workplace® conducted research with over 3,600 participants to develop the validated Love of Workplace Index™ (LOWI) Pulse Validation (Carter, 2019). This survey has become the cornerstone of the Most Loved Workplace® data analytics and survey platform. The LOWI Pulse Validation contains 28 quantitative questions across five content areas that employees respond to using a 5-point scale, known as the SPARK framework.

The SPARK framework consists of five dimensions: Systemic Collaboration, Positive Vision for the Future, Alignment of Values, Respect, and Killer Outcomes. The resulting scores for each functional area and the summary SPARK index help organizations determine employees' level of emotional connectedness to their jobs.

To elicit the most profound insights possible, participants are also asked to respond to qualitative statements: "The top 3 reasons I love working for this company are..." and "The top 3 things that would make me love working for this company more are..." Similar qualitative questions have been developed across twelve Employee Experience domains including Parents and Caregivers, Diversity, Young Professionals, Women, Volunteering, Career Advancement, LGBTQ+, Employee Experience, After Action Review, Leader Effectiveness, Manager Effectiveness, and Customer Love of Company.

Our analysis of employees' qualitative feedback was conducted to create a tool for organizations that allows employees to do their best work through a culture of continuous feedback (Goldsmith & Reiter, 2007). Common themes were generated for feedback aligning with one of the five SPARK areas. These themes were then manually coded for nearly 4,000 snippets of employee responses.

The research demonstrates that the emotion of love, specifically love for one's workplace, is a central predictor of productivity and success within an organization (Carter, 2015). However, a wide range of emotions are expressed in employee feedback, and capturing these emotions is integral to assessing employees' level of organizational commitment. Feedback is categorized across four dimensions: Sentiment (positive, neutral, negative), Emotion (love, gratitude, optimism, excitement, neutral, annoyance, disapproval, anger, fear, disgust), Topics (accountability, collaboration, communication, leadership, management, resources, respect, support, team, transparency, trust, values, vision), and Themes (35 specific organizational strengths and weaknesses).

Despite the deep insights provided by qualitative data, it is cumbersome and time-consuming to extract meaningful insights from hundreds or thousands of text responses. Most Loved Workplace® turned to machine learning to instantaneously categorize employee feedback and provide organizations with an employee listening strategy that predicts employee and client outcomes.

Most Loved Workplace® built a custom BERT (Bidirectional Encoder Representations from Transformers) model with dependencies within the Sentiment, Emotion, Topic, and Thematic content. The model leverages these dependencies to make logical connections between each categorical group. For instance, employee feedback containing positive Sentiment is likely to contain Emotions like love or gratitude rather than disapproval or anger.

A rigorous train-test and validation procedure was used to ensure model accuracy. The model was trained on 1,400 human-coded responses, with test and validation sets each containing 400 human-coded responses. F1 scores for Sentiment reached 87% (validation) and 89% (test). For detecting at least one of two emotions assigned, scores reached 84% (validation) and 86% (test). These F1 scores exceed the 75% threshold generally considered minimum for deployment.

The combination of the Machine Learning approach and the gaming tool used by Most Loved Workplace® can be replicated by other organizations to gain valuable insights into employee sentiments and emotions. The approach is based upon initial research demonstrating that love of the workplace is a reliable and accurate tool to predict outcomes such as organizational commitment, employee performance, and willingness to engage in beneficial workplace behaviors (Carter, 2019). The tool can be used to build and develop a culture where employees love to do their best work with consistent feedback.