Construction of the Culture 500 dictionary began by selecting topics that companies frequently included in their corporate values or that had been researched extensively.

Topic Selection for the Culture 500 Dictionary

For each selected topic, researchers built a preliminary dictionary of high-fidelity words and phrases, then had two team members analyze a random sample of at least 20 reviews containing each term. The bootstrapping process generated tens of thousands of terms, which human experts reviewed and refined.

Bootstrapping Process for Culture 500 Dictionary Terms

For each term in the Culture 500 dictionary, the methodology specified its component words, close variants, sequence and distance between words in a phrase, exclusion terms, and common misspellings.

Per-Term Specifications in the Culture 500 Dictionary

The dictionary-construction process took three years and produced a dictionary of over 20,000 highly accurate terms mapping to more than 90 topics.

Scale and Duration of the Culture 500 Dictionary Construction

To classify reviews into topics that matter to managers and researchers, the Culture 500 team built a custom dictionary from the ground up by combining machine learning tools with human expertise. This approach dramatically improved the accuracy of their models.

Google

To accurately classify text into topics that matter to managers and researchers, the Culture 500 researchers built a custom dictionary using machine learning tools and human expertise. After a three-year process, the dictionary contained over 20,000 highly accurate terms mapped to more than 90 topics.

MIT SMR/Glassdoor Culture 500

The varied language in Glassdoor reviews paints a nuanced picture of corporate culture but defies simplistic textual analysis. When the Culture 500 team initially applied standard natural language processing tools to the Glassdoor data, the results were disappointing.

Limitations of Standard NLP for Culture 500 Analysis

MIT Sloan Management Review project page on measuring culture in leading companies, including the Culture 500 dataset. https://sloanreview.mit.edu/projects/measuring-culture-in-leading-companies/

MIT Sloan Management Review

The MIT Sloan Culture 500 ranks major companies across nine dimensions of corporate culture using Glassdoor reviews. It also celebrates cultural excellence through public recognition and identifies areas where influential companies can improve their culture.

MIT Sloan Culture 500 Rankings of Corporate Culture

The Culture 500 interactive tool provides a snapshot of how frequently and positively employees within a company talk about the Big Nine values.

Culture 500 Snapshot of a Single Company's Cultural Profile

The Culture 500 allows users to benchmark companies against their competitors on each of the Big Nine values.

Competitor Benchmarking View in Culture 500

Glassdoor reviewers use rich and varied language when describing culture. That language paints a nuanced picture of corporate culture, but it also defies simplistic approaches to textual analysis.

Glassdoor Open-Text Reviews for Culture Insight

To detect potentially problematic reviews, the Culture 500 researchers identified positive spikes: company-months in which both the number of reviews and the overall company rating were two or more standard deviations above the monthly mean.

Positive Review Spike Detection in Culture 500

Custom Culture 500 Dictionary Combining Machine Learning and Human Expertise

Nearly all organizations in the Culture 500 are for-profit companies and most are headquartered in the United States. Some international companies with large U.S.-based workforces are also included, but regardless of headquarters location, only reviews from employees based in the U.S. are used in the analysis.

For-Profit US-Based Composition of the Culture 500 Sample

The Culture 500 organizes its companies into 33 clearly defined industries, with an average of 18 companies per industry. In the tool's industry bar graphs, the number to the right of each bar denotes the count of companies included in that industry.

Learn Before

Related

Learn After