Glassdoor review analysis says employee voice can expose what ratings miss

A new analysis of 604,320 public Glassdoor reviews argues that standard workplace ratings often track the same underlying signal, while employees’ written comments reveal what recommendation scores miss. The findings raise new questions about why negative employee voice appears to affect some employer brands more than others, including Salesforce and McDonald’s.

Why it matters: - Patrick Diogenia’s analysis suggests five familiar workplace ratings may be less distinct than they appear, which matters for employers using Glassdoor-style dashboards to judge workforce health. - The research says written employee comments can add signal that star-style ratings do not capture, especially around whether employees would recommend an organization. - The findings also point to a possible employer-brand effect, where some companies appear more resilient to negative employee voice than others. - That has implications for employee Net Promoter Score, listening programs and how organizations interpret workforce feedback.

What happened: - Diogenia analyzed 604,320 public Glassdoor employee reviews for a newly published research guide, Glassdoor Employee Reviews: What They Reveal About Employee Voice and Employer Brand. - The analysis compared structured workplace ratings with employees’ written descriptions of what they liked and disliked about employers. - The review set included historical data across multiple employers, with Salesforce and McDonald’s used as examples at opposite ends of one observed pattern. - The project is presented as independent research and is not affiliated with, sponsored by or endorsed by Glassdoor.

The details: - Ratings for senior management, culture and values, career opportunities, compensation and benefits, and work-life balance overlapped so much that combining them into one overall workplace measure caused virtually no loss in predictive performance. - Negative employee language was more informative than positive language in distinguishing whether an employee would recommend the organization. - Negative recommendations tended to use sharper, more specific language. - Employees who still recommended their employer often described shortcomings in more qualified terms. - Across the employers studied, the link between negative employee commentary and recommendation varied meaningfully. - Salesforce showed a weaker relationship between negative employee commentary and recommendation in the historical dataset. - McDonald’s showed a stronger relationship between negative employee commentary and recommendation in the historical dataset. - The analysis says those company-level differences are exploratory and not a ranking of employer quality. - The project also says the differences do not measure current employee experience or prove that one employer has a stronger brand than another. - The source materials are available in the full analysis and the supporting project materials.

Between the lines: - The results suggest recommendation scores can be useful as a high-level pulse, but they may compress different dimensions of workplace experience into one number. - Diogenia’s interpretation is that employer-brand resilience may help explain why negative employee voice carries different consequences across companies. - The project does not directly measure that resilience, so the idea remains a hypothesis for future research. - Because the dataset is self-selected, historical and public, the findings are associative and predictive rather than causal. - The research also has not been peer-reviewed, which limits how far the results can be generalized.

What's next: - Diogenia says future research will examine employer-brand resilience more directly. - Organizations using employee listening programs may need to combine recommendation metrics with richer text feedback and other evidence. - The broader question going forward is not just whether employee sentiment moved, but what employees think is happening underneath the score.

The bottom line: - A single recommendation number can summarize sentiment, but employee voice may be better at showing why the number changed and whether an employer’s brand can absorb the hit.