WHAT DOES ‘META
ANALYSIS’ MEAN?

Potential sources are detected through a variety of channels, including corporate career websites, that hold employer awards, HR-related blog entries or articles, professional networks such as LinkedIn, keyword-based searches via Google, cross-country comparisons, suggestions from the public or the Advisory Board. 

The fundamental aim in compiling these sources is to derive the most complete and accurate picture possible from disparate fragments. The more multifaceted the portfolio of considered criteria, the greater the likelihood of differentiating so-called ‘black sheep’ from genuine top performers. 

To safeguard methodological consistency and quality, the evaluation of such sources is conducted on an ongoing basis. The assessment process examines integrity and quality across several dimensions. These include the type of source (e.g. audit reports, surveys, public review portals), the research approach (self-enrolment mechanisms versus independent studies), and the relevance of data. Further considerations encompass validation procedures, barriers against manipulation, the availability of complaint and correction mechanisms, and whether the source operates under the oversight of academic institutions or governmental bodies. For further information, please see 2. Classifying and Standardising Sources.

This multi-dimensional approach ensures that newly integrated sources meet our quality standards, thereby reinforcing the robustness, reliability, and validity of the overall study design.

Sources are classified according to their relevance and quality in order to ensure consistency throughout the assessment process. This classification covers multiple dimensions and provides the foundation for a scientifically valid analysis.

• Content Alignment: Each source is reviewed with regard to its content and thematically assigned to one or more of the nine study dimensions. As these dimensions are interrelated, a single source may be attributed to several categories if its content is relevant across different areas.
  Geographical Relevance: Beyond content, the geographical scope of each source is assessed. Sources are categorised as regional (local significance), national (country-level relevance), or transnational/global. This includes sources applicable to wider regions such as Europe, Asia, or Africa, as well as those with global validity.
• Type of Source: The classification considers whether the material originates from a structured audit, a survey, a jury prize, an HR benchmark, a self-disclosure, a membership, or a public review portal.
• Research Approach: It is evaluated whether the data derives from independent research, externally validated processes, or self-enrolment mechanisms.
• Validation and Oversight: Each source is further assessed with regard to the robustness of its validation mechanisms, safeguards against manipulation, the availability of complaint or correction procedures, and, where possible, academic or governmental endorsement.
• Duration and Topicality: Consideration is also given to how long the source has existed. Crucial for the survey is the sustainable development of an organisation over the years. At the same time, it is ensured that each source possesses sufficient topicality and relevance to provide valid and up-to-date insights.

Through this multi-layered classification process, sources are not only thematically aligned and geographically contextualised, but also standardised with regard to quality, independence, topicality, and credibility. This systematic approach strengthens both the reliability and validity of the study results and underpins the overall scientific integrity of the assessment. 

The process of data gathering is designed to capture the maximum available information from each source, ensuring both breadth and depth of coverage. To achieve this, we employ a hybrid approach, combining automated collection with manual verification.

Automated data gathering is conducted through specialised tools and research agents that access industry-leading databases and integrate external sources such as Cognism for market and company information. This enables large-scale, efficient data enrichment and validation across hundred thousands of companies and 500+ diverse sources.

Data storage and management rely on a robust and scalable infrastructure. Postgres functions as the central high-performance database, while Supabase provides authentication and vector storage capabilities. n8n supports process automation, ensuring smooth workflows from collection to integration.

Report generation and analysis are powered by multiple Large Language Models (LLMs) such as OpenAI and Gemini, which facilitate intelligent, AI-driven analyses, dynamic evaluation logic, and contextual interpretation of the gathered data.

Manual data gathering complements the automated system: data analysts extract records directly from the respective sources, ensuring that qualitative and non-standardised information is accurately integrated.

By combining automated scalability, structured storage, and AI-based reporting with manual precision, this approach guarantees comprehensive, reliable, and scientifically validated insights.

To ensure accuracy and relevance, the dataset undergoes a rigorous cleaning process. For company name matching and the removal of duplicates, a multi-step procedure is applied that combines Natural Language Processing (NLP) techniques with advanced language models. This includes the normalisation of company names, the comparison of variations and similarities, and semantic checks using Large Language Models (LLMs) to guarantee high data quality. Manual verification complements this process in order to secure accuracy for critical data points. Once consolidated, inclusion and exclusion criteria are applied to filter out companies that are too small, inactive, or have suffered significant reputational issues.

In cases where results remain ambiguous, duplicate cleaning is supplemented by manual inspection. This time-consuming procedure is carried out according to the following principles, among others:

• Were there any renaming of the same company?
• Have there been mergers, acquisitions, or structural changes?
• Can conclusions be drawn from postal addresses?
• Identification of affiliations between entries that do not appear alphabetically together, for example:
  SC (as abbreviation), The Sample Company (with prefix), Founder’s Sample Company (including the founder’s name), etc.

In addition, there is a wide variety of industry classifications and descriptive terms. To account for this diversity and ensure comparability, industry classification is conducted automatically via the Glassdoor network, as it is globally available and provides a consistent framework for international benchmarking.

By combining automated language technologies with targeted manual checks in ambiguous cases, the study ensures a high level of robustness, reliability, and validity in the data preparation phase.

The scoring process is designed as a multi-stage evaluation procedure to ensure fairness and accuracy. The points awarded per source are directly based on its quality and relative importance. For example, a self-disclosure statement receives fewer points than comprehensive audit, which, in addition to the time and financial investment required, also involves employees and requires concrete follow-up measures based on the results.

The process begins with the evaluation of the relevance of individual data points in relation to the nine central topics of the study. Employee evaluations are only taken into account if a defined participation and representativeness threshold is reached; if this threshold is not met, the data is excluded. Once the threshold is exceeded, no further weighting adjustments are made, so that only reliable data sets are included in the results.

In addition to the overall ranking, rankings are also created for each of the nine categories, as well as differentiated region, city and industry rankings. In addition to traditional sources, the nine category rankings also incorporate topic-specific values from employee review portals – for example, the CEO score in the area of leadership. In this way, external perceptions are assigned to the respective dimensions in a meaningful way and contribute to the validity of the results.

The quality assurance process is designed to ensure data accuracy through rigorous cleaning and enrichment procedures.

Multiple plausibility checks, including anomaly detection, are employed to identify inconsistencies, such as unusually high scores supported by only a minimal number of sources. As a standard procedure, both very low and extremely high results are systematically reviewed in order to detect outliers at an early stage and to guarantee a realistic representation. These automated checks are complemented by detailed manual reviews, ensuring that all results remain both reliable and representative.

In addition, spot checks are conducted across all categories of collected data. These include comparisons with results from previous years, verification of company career page URLs, and confirmation of whether the organisation continues to operate in its current form, alongside an evaluation of the quality of the respective career environment. Further assessments also take into account potential corporate changes, such as renamings, mergers, acquisitions, or insolvency proceedings.

By combining automated detection mechanisms with targeted manual validations, the study ensures a high degree of robustness, validity, and transparency in its outcomes.

a. Structure and Composition of the Study Reports

The study reports present the individual results of each certified organisation and follow a clear analytical logic. They begin with the overall ranking, followed by the results across nine assessment dimensions, which are contextualised within both industry and size-specific benchmarks. In-depth analyses highlight strengths and areas for development, complemented by external perception data such as career websites, job advertisements, career portals, and social media. The reports conclude with prioritised recommendations for action derived directly from the results.

• Management Summary: Overview of overall performance, top-performing areas, quick wins, and identified development fields.
• Results: Detailed presentation at both overall and thematic level, differentiated further by city, region, and industry.
• Insights: Benchmarks, strengths, and optimisation potential providing contextual interpretation.
• Recommendations: Precise, prioritised measures directly based on the analysis.

The study reports are generated through a technologically robust infrastructure (Postgres, Supabase, n8n, LLMs such as OpenAI and Gemini) and are safeguarded by rigorous human quality control. This ensures compliance with the central quality criteria of scientific research: reliability, validity, and objectivity. The reports thus combine methodological rigour with practice-oriented perspectives and provide a transparent, data-driven foundation for the evaluation of employer excellence.

b. From Recruiting Performance Check to Recruiting Performance Portal

The Recruiting Performance Portal extends the former Recruiting Performance Check into a comprehensive analytical tool for organisational career environments. In addition to the systematic evaluation of career websites and job advertisements, it integrates AI-driven chat interactions that provide adaptive feedback, as well as SEO and geo-analyses and the identification of optimisation potentials. In doing so, the portal combines empirical evidence with action-oriented application logic and establishes a reliable basis for the strategic advancement of recruitment processes.

With its relaunch, the underlying evaluation logic has been fundamentally modernised. The previous 2025 system relied on a rule-based assessment structure with additive scoring logic, comparable to a multiple-choice test. By contrast, the 2026 system applies a cognitively enhanced approach powered by advanced large language models. This enables semantically differentiated and context-sensitive assessments that reflect the complex realities of contemporary recruitment practices.

Furthermore, the system allows for the integration of organisation-specific research questions, thereby generating results that are both individualised and scientifically valid. In combination with enhanced reporting, this establishes a technologically advanced, transparent, and scientifically grounded system that provides a robust basis for the evaluation of employer quality.