Extracting organizational culture from text

Why We Conduct Research?

At The Culture Factor Group, we believe in unlocking culture with data. Our research advances the understanding of how culture shapes decisions, organisations, and societies. By analysing large-scale data, we refine existing models, challenge  assumptions, and provide insights that drive better leadership, HR policies, and global strategies.

Understanding organisational culture is critical for business success, yet traditional assessments rely heavily on self-reported surveys, which are often biased. This study is relevant to business leaders, HR professionals, and researchers who seek more objective, data-driven methods for evaluating company culture.

The research introduces a new, theory-based text analysis tool that can extract cultural insights from digital text data, including employee reviews, company websites, and mission statements, offering an alternative to conventional surveys.

Purpose of the research

This study develops and validates the Dictionary of Organisational Culture and Practices (DOCP), a theory-driven text analysis tool designed to measure organisational culture from digital data sources.

• The tool is based on the Multi-Focus Model on Organisational Culture
• It assesses six key organisational culture dimensions using natural language processing (NLP) and computational linguistics.
• The goal is to provide a scalable, unbiased, and more reliable method for analysing organisational culture.

Findings

Figure 1. Dimension pole homogeneity analysis: visualisation of the dictionary allocations by experts. Numbers in parentheses refer to the dimension the pole belongs to (D1 = Effectiveness, D2 = Customer Orientation, D3 = Control, D4 = Focus, D5 = Approachability, D6 = Management Philosophy). Each arrow from the center of the chart towards the outside represents one Multi-Focus Model dimension pole. The Multi-Focus Model dimension number is in parentheses next to the pole name; hence, “goals (D1)” and “means (D1)” are opposite poles on the same dimension (D1 Effectiveness) and are, in theory, expected to be opposite of each other. To avoid clutter, not all labels are displayed. For more details, see an interactive version of this chart linked on https://osf.io/x5wqt/wiki.

The study applied six core dimensions of organisational culture using computational text analysis:

1. 1. Effectiveness (Means vs. Goal-Oriented) Process-focused vs. outcome-driven cultures.
   2. Customer Orientation (Internal vs. External Focus)  Ethics-driven vs. results-driven customer approaches.
   3. Control (Easy-Going vs. Strict) Flexibility vs. rule-based organisational structure.
   4. Focus (Local vs. Professional)  Employee loyalty to leadership vs. industry expertise.
   5. Approachability (Open vs. Closed Culture)  Transparency and inclusivity vs. exclusivity.
   6. Management Philosophy (Employee- vs. Work-Oriented)  Employee well-being vs. task completion priorities.

The DOCP tool successfully replicated the dimensional structure of organisational culture dimensions known form surveys, showing a strong overlap with survey-based cultural assessments. It demonstrated the ability of identifying cultural traits across multiple text sources.

Implications

For HR & Business Leaders: Provides a non-intrusive method to evaluate corporate culture and detect discrepancies between stated values (e.g., mission statements) and actual employee experiences (e.g., reviews).
For Researchers: Offers a validated methodology to measure organisational culture using digital data.
For Organisations: Enables continuous culture monitoring, helping leaders align culture with strategy and improve employee engagement.

Methodology

• Data Sources:
  • Employee reviews (526 companies).
  • Mission & vision statements (1,121 companies).
  • Company websites (285 companies).
• Sample Size: 26 million words analysed using computational linguistics and NLP.
• Validation: Compared against survey-based models and existing AI-driven tools.
• Analysis Techniques: Dictionary-based NLP, machine learning, and factor analysis.

Research Limitations/Implications

• Text-based analysis may differ from traditional survey results, requiring careful interpretation.
• Some text sources (e.g., annual reports) may not fully capture internal company culture.
• Future research should explore multilingual capabilities and cross-industry applications.

What is peer reviewed?

This research underwent a rigorous peer-review process, ensuring that its methodology, findings, and conclusions meet academic standards. Peer review involves expert evaluation, ensuring the study’s objectivity, reliability, and scientific validity.

• Authors: Michael Schachner, M. Murat Ardag, Peter Holtz, Johannes Großer, Carina Hartz, Hester van Herk, Michael Bender, Klaus Boehnke, Henrik Dobewall
• Journal: European Journal of Work and Organisational Psychology
• Publisher: Taylor & Francis (Informa UK Limited)
• Article Publication Date: 17 June 2024
• DOI: 10.1080/1359432X.2024.2360225