Developing Analytical Systems for XBRL Data
Learning Objectives
- Understand how to establish clear objectives for analytical data systems that align with users’ requirements
- Analyse XBRL-specific system components including taxonomy structure and filing rules documentation
- Leverage XBRL structural elements including presentation relationships and dimensional analysis capabilities
- Apply XBRL taxonomy intelligence to existing frameworks like ESRS disclosures
- Understand how to use supervisory reporting taxonomies for analytical regulatory purposes
Introduction
Building analytical systems that can effectively process and analyse reported XBRL data, such as sustainability reports, requires a systematic approach that begins with clear objectives and progresses through careful assessment of data characteristics and user requirements. The complexity of business data, which can range from highly structured numeric information to unstructured narrative content, demands thoughtful system design that matches analytical capabilities to data types and user needs.
Establishing Clear Analytical Objectives
The foundation of any effective analytical system lies in clearly defining what the system needs to accomplish. These objectives should flow directly from the established policies, regulations, and mandate of the data collector. A financial supervisor building an analytical system will have different objectives from a securities regulator, environmental agency, or policy development unit, even when all are working with similar sustainability data sources.
Defining analytical objectives requires understanding both the immediate analytical needs and the longer-term strategic goals of the institution. Immediate needs might include monitoring compliance with specific disclosure requirements, identifying trends in particular sustainability metrics, or supporting regular supervisory assessments. Strategic objectives could encompass developing new policy initiatives, supporting international cooperation efforts, or building capabilities for emerging analytical challenges.
Understanding User Requirements and Analytical Scenarios
Once objectives are established, the next critical step involves systematically gathering requirements from the users of the analytical system. This process should identify the specific analytical scenarios that users need to support, the key performance indicators they need to monitor, the types of data aggregation required, and their preferences for data visualisation and reporting.
Different user groups within the same organisation often require different analytical capabilities. Senior policymakers might need high-level dashboards that summarise trends across entire sectors, while technical staff conducting detailed assessments might require the ability to drill down into individual company disclosures and compare specific data points across multiple entities. Understanding these varied requirements early in the system design process prevents costly modifications later.
Analysing XBRL-Specific System Components
When building analytical systems based on XBRL data, additional analysis becomes necessary to understand the specific characteristics of the reporting framework being used. This step should examine the XBRL taxonomy structure, which defines the concepts, relationships, and validation rules that govern the reported data. Understanding how the taxonomy organises information provides crucial insights for designing analytical processes that can effectively leverage the structured relationships built into the data.
Filing rules documentation provides another essential input for system design. These rules define technical requirements, validation criteria, and submission processes that affect how data gets collected and stored. Understanding these rules helps ensure that analytical systems can properly interpret the collected data and identify any data quality issues that might affect analytical results.
The mechanism for identifying reporting entities also requires analysis, particularly when building systems that need to track entities across multiple reporting periods or link sustainability data with other regulatory information. This analysis should address how entity identification works within the target system and how entity relationships get maintained over time.
Assessing Data Nature and Format Characteristics
The design of analytical systems depends on understanding the nature and format characteristics of the source data. Regulatory disclosures present particular challenges because they may span multiple data formats, each requiring different processing, storage, and analytical approaches.
Structured data with rich metadata represents the most sophisticated category of information. XBRL reporting provides detailed tagging of individual numeric values, with comprehensive metadata defined in taxonomies that explain the meaning, context, and relationships of each data point. This rich structure enables sophisticated analytical capabilities, including automated validation, dimensional analysis, and complex aggregations based on taxonomy relationships.
XBRL also supports block tagging of narrative information, which captures unstructured content such as descriptive text and tables within a structured framework. While the overall structure remains machine-readable, the content itself requires different analytical approaches than pure numeric data.
Other structured formats with rich metadata, such as data based on ISO 20022 standards, provide detailed definition dictionaries that explain data meanings and relationships. These standards enable sophisticated processing while maintaining clarity about data semantics.
Structured data without metadata presents different analytical challenges. Formats such as JSON, XML, or CSV files may organise information in predictable structures, but without additional technical dictionaries or documentation, understanding data meanings and relationships becomes more difficult. These formats require analysts to infer data semantics from context or to develop separate documentation that explains the meaning of the data.
Unstructured formats such as PDF documents represent the most challenging category for automated analysis. While these formats preserve human-readable presentation, extracting specific data points for analytical purposes requires specialised processing techniques and often involves some degree of manual interpretation.
Matching Storage and Analysis Approaches to Data Types
Each data format type requires different approaches to data extraction, storage, and analysis. Understanding these requirements early in system design helps ensure that technical infrastructure can effectively support the intended analytical capabilities.
Numeric data with rich metadata benefits from storage systems that can preserve and leverage the metadata relationships. Traditional relational database systems equipped with business intelligence software provide robust capabilities for data manipulation, aggregation, and presentation when working with well-structured numeric information. These systems excel at supporting standard analytical operations such as trend analysis, comparative assessments, and statistical calculations.
Non-numeric data, particularly descriptive narrative content, requires a more sophisticated analytical approach. Traditional business intelligence tools designed for numeric analysis cannot effectively process narrative content that might contain crucial qualitative information about sustainability performance, risk management approaches, or strategic intentions.
Modern analytical approaches increasingly incorporate large language models and other artificial intelligence technologies to analyse narrative content systematically. These tools can identify themes and sentiments, extract specific information, compare approaches across multiple entities, and even assess the consistency between narrative descriptions and numeric performance data.
The choice between traditional business intelligence approaches and more modern AI-based solutions often depends on the balance between numeric and narrative content in the data, the complexity of the analytical questions being asked, and the technical capabilities available within the organisation.
Integration Considerations
Effective analytical systems often need to integrate multiple data sources and analytical approaches. A comprehensive analytical system might combine structured numeric data processed through traditional business intelligence tools with narrative content analysed using AI-based approaches, supplemented by external data sources that provide context or benchmarking information.
This integration requires careful attention to data quality, entity matching across different sources, and maintaining consistency in analytical outputs despite using different processing approaches for different data types. The system design should also accommodate future expansion as new data sources become available or analytical requirements evolve.
Building effective analytical systems for XBRL data represents a significant technical and analytical challenge, but understanding the systematic approach outlined above provides a foundation for developing capabilities that can effectively support regulatory and policy objectives while accommodating the diverse characteristics of the gathered information.
Read about storing and securing XBRL reports in this article.
Harnessing XBRL Structure for Advanced Analytics
Building analytical systems that effectively leverage XBRL data depends on understanding how to extract and use the rich structural information embedded within XBRL taxonomies and reports. Unlike simple data formats that contain only raw values, XBRL provides comprehensive metadata, relationships, and semantic definitions that can significantly enhance analytical capabilities when properly incorporated into system design.
The key principle underlying effective XBRL-based analytical systems involves preserving and leveraging the structured relationships defined within the taxonomies rather than treating XBRL data as merely another source of numeric values. This approach enables analytical systems to maintain the semantic meaning and dimensional complexity that makes XBRL particularly valuable for regulatory and policy analysis.
Utilising Presentation and Definition Relationships
The presentation relationships (or presentation tree) within XBRL taxonomies defines hierarchical relationships between concepts that reflect how information should be organised and displayed. These relationships provide a foundation for building analytical dashboards and tabular presentations that maintain the logical structure inherent in the reported data.
When designing analytical interfaces, the presentation relationships can guide the development of drill-down capabilities that allow users to navigate from summary-level information to detailed components while maintaining the conceptual relationships defined in the taxonomy. This approach ensures that analytical presentations reflect the intended meaning and structure of the reported information rather than imposing arbitrary organisational schemes.
The definition linkbase serves different but complementary purposes by defining the logical relationships between concepts. These relationships prove particularly valuable for building analytical visualisations that show how different elements interact or change over time. For instance, when analysing changes in financial position or sustainability metrics, the definition relationships help ensure that analytical presentations accurately reflect the underlying conceptual connections between different data elements.
Leveraging Dimensional Analysis Capabilities
XBRL’s dimensional structure provides sophisticated capabilities for filtering, aggregating, and analysing data across multiple dimensions simultaneously. The dimensions and their corresponding domain members can be systematically incorporated into analytical systems to provide flexible filtering and aggregation capabilities.
Dimensions enable users to view the same underlying data from multiple perspectives without losing the contextual information that gives meaning to specific values. A single sustainability metric might be reported across multiple geographic regions, business segments, or time periods. The dimensional structure preserves these distinctions while enabling analytical systems to aggregate or disaggregate information as needed for different analytical purposes.
This dimensional capability proves particularly valuable for calculating aggregates across multiple companies’ data. When companies report using consistent dimensional structures, analytical systems can perform meaningful aggregations that maintain semantic consistency. This enables sector-level analysis, benchmarking, and trend identification that would be difficult or impossible with less structured data formats.
Understanding Data Characteristics Through Type Definitions
XBRL taxonomies include detailed data type definitions that specify the nature and characteristics of each piece of information. These definitions help analytical systems understand how to process, validate, and present different types of data appropriately.
Monetary values require different processing than percentages, dates, or text descriptions. The data type definitions ensure that analytical systems can apply appropriate formatting, validation rules, and analytical operations to each data element. This understanding prevents analytical errors that might arise from treating different data types inappropriately.
Preserving Original Presentation Through XBRL Reporting Templates
XBRL reporting templates (Table Linkbase) enable the creation of complex layouts that support multi-dimensional breakdowns and sophisticated presentation patterns. This capability proves valuable for analytical systems that need to render data in formats that match original business reports while maintaining the underlying dimensional structure.
This functionality allows analytical systems to present as-reported views of business reports alongside more traditional analytical presentations. Users can examine how companies originally presented their information while also accessing the structured data for comparative analysis. This dual capability supports both detailed review of individual reports and systematic analysis across multiple entities.
Supporting Multi-Language and Multi-Context Analysis
Labels within XBRL taxonomies support multiple languages and label types, enabling analytical systems to present information in different languages while maintaining semantic consistency. This capability proves essential for international regulatory systems that need to analyse data reported in multiple languages or that serve users who work in different languages.
The multiple label types also enable analytical systems to present information at different levels of detail or for different audiences. Technical labels might provide precise definitions for specialist users, while standard labels offer more accessible descriptions for general audiences.
Incorporating Regulatory and Legal Context
References associate concepts with citations from authoritative literature, enabling analytical systems to maintain connections between reported data and the underlying regulatory or accounting standards that define reporting requirements. This capability supports the construction of knowledge graphs that link financial and sustainability concepts to their regulatory foundations.
These connections enable analytical systems to provide context about why specific information gets reported and how it relates to broader regulatory frameworks. Users can understand not just what companies reported, but also the regulatory logic underlying specific disclosure requirements. They can also access a complete wording of the underlying regulation or policy which is associated with given reported facts.
Enabling Advanced Validation and Derived Analytics
XBRL formula rules provide standardised mechanisms for defining validation rules and derived calculations within taxonomies. These capabilities can be leveraged for analytical purposes that extend beyond basic validation to include pattern analysis and trend identification.
Formula-based approaches enable analytical systems to generate derived metrics, identify inconsistencies across related data points, and apply sophisticated business rules that reflect regulatory expectations. This capability transforms taxonomies from simple data definitions into active analytical frameworks that can support complex regulatory analysis.
Integrating with Semantic and OLAP Systems
The structured nature of XBRL data enables integration with semantic modelling facilities and Online Analytical Processing (OLAP) systems that can provide sophisticated analytical capabilities. The XBRL Open Information Model (OIM) supports mapping between XBRL data and other semantic frameworks, enabling analytical systems to leverage advanced analytical tools while preserving the meaning embedded in XBRL structures.
Advanced mapping frameworks can handle the heterogeneity that exists across different XBRL filings while maintaining analytical consistency. This capability enables analytical systems to work effectively with data from multiple taxonomies or reporting periods while preserving comparability.
Building Knowledge Graphs for Regulatory Analysis
The combination of references, concept definitions, and structured relationships within XBRL taxonomies provides a foundation for constructing knowledge graphs that connect reported data with broader regulatory and policy frameworks. These knowledge graphs can support sophisticated analytical queries that span multiple domains of regulatory interest.
Knowledge graph approaches enable analytical systems to answer complex questions about relationships between reported data, regulatory requirements, policy objectives, and broader economic or environmental trends. This capability transforms XBRL data from isolated reporting information into a connected knowledge resource that can support comprehensive regulatory analysis.
The systematic utilisation of these XBRL structural elements enables analytical systems that are more sophisticated, semantically consistent, and analytically powerful than systems that treat XBRL as merely another data format. Understanding and implementing these capabilities becomes essential for building analytical systems that can fully leverage the investment in XBRL-based reporting programmes.
Case Study: Analysing ESRS Disclosures with XBRL Taxonomy Intelligence
Understanding how to leverage XBRL taxonomy information for analytical system design becomes clearer when examining specific implementation scenarios. The European Sustainability Reporting Standards XBRL taxonomy provides a comprehensive example of how detailed taxonomy structure can inform analytical system development and enhance supervisory capabilities.
ESRS XBRL Taxonomy Structure and Analytical Implications
The European Sustainability Reporting Standards (ESRS) reporting standards, developed under the Interoperable Taxonomy Architecture, demonstrate how modern sustainability reporting frameworks can provide structured data that supports sophisticated analytical capabilities. The taxonomy defines multiple data types that require different analytical approaches, each offering distinct insights for supervisory and regulatory analysis.
This diversity of data types within a single reporting framework illustrates why analytical systems must be designed to handle multiple information formats while maintaining the semantic relationships defined in the taxonomy. Supervisors can leverage the taxonomy’s structure to understand not just what information companies report, but also how different pieces of information relate to each other and what analytical approaches prove most effective for each data type.
Numerical Data Supporting KPI Calculations
The taxonomy includes concepts such as “Non-renewable energy production” with its data type defined as an “energy item” (dtr-types:energyItemType) under ESRS E1-5. This classification indicates that the data represents energy measurements that feed into broader sustainability calculations rather than standing alone as final metrics.
For analytical systems, this taxonomy information signals that such data points should be designed for aggregation and calculation purposes. Supervisors can leverage this understanding to build analytical capabilities that automatically identify relationships between underlying data and derived metrics, enabling quality assessments that verify whether reported KPIs align with their component data elements. The data type definition also ensures that analytical systems apply appropriate units, validation rules, and aggregation methods when processing energy-related information.
Key Performance Indicators with Complex Measurements
Concepts like “Energy intensity from activities in high climate impact sectors (measured as total energy consumption per net revenue)” with its data type defined as “energy per monetary item” (dtr-types:energyPerMonetaryItemType) under ESRS E1-5 represent calculated metrics that combine different measurement types. This data type definition indicates that the value represents energy consumption normalised by monetary amounts.
The taxonomy structure enables analytical systems to understand the composite nature of such metrics and to design appropriate comparative analyses. Supervisors can use this information to build analytical capabilities that properly interpret intensity metrics, compare them across entities of different sizes, and identify outliers that might indicate reporting errors or exceptional circumstances. The taxonomy-defined relationship between energy and monetary measurements also enables analytical systems to perform meaningful benchmarking while accounting for scale differences between entities.
Narrative Disclosure Analysis
Text block concepts such as “Description of key contents of policy [text block]” with its data type defined as a text block item (dtr-types:textBlockItemType) under ESRS 2 MDR-P contain narrative information that requires different analytical approaches than numeric data. The taxonomy classification indicates that this information provides qualitative context rather than quantifiable metrics.
For analytical systems, this taxonomy information guides the selection of appropriate processing techniques. Supervisors can leverage modern text analysis tools, including large language models, to systematically analyse narrative disclosures for consistency, completeness, and alignment with numeric data. The taxonomy structure also enables analytical systems to identify relationships between narrative explanations and related quantitative metrics, supporting quality assessments that verify whether companies’ qualitative descriptions align with their reported numbers.
Mixed Content Analysis
Some concepts combine narrative and numeric information, such as “Disclosure of whether internal carbon pricing schemes are applied, and if so, how these support decision making and incentivise implementation of climate-related policies and targets [text block]” under ESRS E1-8. While technically classified as text blocks, these disclosures often contain embedded numeric information alongside qualitative explanations, because of their level in the hierarchy of data model elements.
The taxonomy structure helps analytical systems understand that such disclosures require hybrid processing approaches. Supervisors can design analytical capabilities that extract both quantitative data and qualitative themes from these mixed content disclosures, enabling comprehensive analysis that considers both the numeric aspects and the strategic context provided by companies.
Structured Choice Data
Enumeration concepts like “High climate impact sector(s) used to determine energy intensity” defined as enumeration set (’enum2:enumerationSetItemType’) under ESRS E1-5 provide structured selection from predefined options. This data type enables systematic comparison across entities while ensuring consistency in terminology and categorisation.
The taxonomy definition enables analytical systems to perform categorical analysis, identify sector concentration patterns, and assess consistency between sector classifications and related numeric metrics. Supervisors can leverage this structured choice data to build analytical capabilities that automatically group entities by sector characteristics and perform sector-specific benchmark analyses.
Boolean Data for Policy Assessment
Simple true/false concepts such as “Internal carbon pricing schemes are applied” with the data type defined as Boolean (xbrli:booleanItemType) under ESRS E1-8 provide clear binary indicators that enable systematic policy analysis across large numbers of entities.
The taxonomy structure enables analytical systems to quickly identify policy adoption patterns, calculate adoption rates across different entity types or sectors, and correlate policy adoption with performance metrics. Supervisors can use this information to assess the effectiveness of different policy approaches and to identify entities that might benefit from additional guidance or oversight regarding specific sustainability practices.
Leveraging Taxonomy Relationships for Comprehensive Analysis
The ESRS XBRL taxonomy defines not only individual concepts but also relationships between them that enable sophisticated cross-referencing and validation. Analytical systems can leverage these relationships to identify inconsistencies, verify completeness, and build comprehensive views of entity performance that span multiple disclosure areas.
For supervisors, this relationship information enables analytical approaches that consider performance holistically rather than examining individual metrics in isolation. The taxonomy structure supports analytical systems that can automatically identify connections between different aspects of sustainability performance and flag situations where different disclosure areas provide conflicting information.
Supporting Different Analytical Use Cases
The structured nature of the ESRS XBRL taxonomy enables analytical systems to support multiple supervisory use cases simultaneously. Supervisors conducting detailed entity assessments can leverage the granular data and relationship information, while those performing sector-wide analysis can use the standardised structure to aggregate and compare information across multiple entities.
The taxonomy also supports both quantitative analysis of numeric data and qualitative analysis of narrative content within a unified framework. This integration enables comprehensive analytical approaches that consider all aspects of entity reporting rather than focusing solely on quantifiable metrics.
This practical example demonstrates how understanding and leveraging XBRL taxonomy structure transforms analytical system capabilities from simple data presentation tools into sophisticated regulatory analysis platforms that can support complex supervisory and policy development activities.
Case Study: Leveraging the Solvency II XBRL Taxonomy for Sustainability Risk Assessment in Prudential Supervision
The application of XBRL taxonomy intelligence extends beyond public disclosure frameworks to encompass supervisory reporting systems designed for prudential oversight and financial stability assessment. The EIOPA Solvency II XBRL taxonomy, developed under the Data Point Model architecture, demonstrates how supervisory authorities can leverage structured taxonomy information to enhance their assessment of sustainability risks within insurance supervision.
DPM Architecture and Supervisory Analytical Requirements
The Data Point Model architecture underlying the Solvency II taxonomy reflects the specific requirements of prudential supervision, where regulators need precise, comparable data for assessing financial soundness and systemic risk. Unlike public disclosure frameworks that serve multiple stakeholder groups, supervisory taxonomies focus specifically on providing supervisors with the information needed to fulfil their prudential oversight responsibilities.
This architectural difference has important implications for analytical system design. Supervisory analytical systems must prioritise precision, consistency, and integration with existing prudential assessment frameworks rather than accessibility for general stakeholders. The taxonomy structure enables analytical systems to maintain strict data quality standards while supporting the complex calculations and risk assessments that prudential supervision requires. One such example is the inclusion of Pillar 3 ESG disclosures in the EBA DPM reporting framework DPM v3.3 phase 3. The integration of ESG pillar 3 disclosures into EBA DPM and taxonomy required consistency in data modelling, e.g. reuse of already existing taxonomy elements when possible and creation of new elements in line with existing taxonomy principles. An example of such a taxonomy is the EBA’s XBRL taxonomy.
Climate Risk Exposure Assessment Through Structured Data
The EIOPA XBRL taxonomy includes concepts for reporting the monetary values of assets exposed to climate change-related transition risk and climate change-related physical risk, as captured in table S.06.04. These data points enable supervisors to systematically assess how climate-related risks might affect insurance undertakings’ financial positions and the broader insurance sector’s stability.
For analytical systems, the taxonomy structure provides essential context about how these exposure measurements should be interpreted and aggregated. Supervisors can leverage this information to build analytical capabilities that automatically calculate sector-wide exposure concentrations, identify outlier institutions with unusual risk profiles, and assess whether reported exposures align with broader market conditions and regulatory expectations.
The monetary data type definition ensures that analytical systems apply consistent currency handling and aggregation methods, enabling meaningful comparisons across institutions and over time. The table structure also enables analytical systems to understand relationships between different types of climate exposures, supporting comprehensive risk assessments that consider both transition and physical climate risks within integrated analytical frameworks.
Capital Requirement Analysis for Risk Mitigation
Concepts addressing Solvency capital requirements for risk mitigation effects across various catastrophe risk types, as reported in table S.27.01, provide supervisors with detailed information about how institutions calculate their capital requirements and the extent to which they rely on risk mitigation techniques.
The taxonomy structure enables analytical systems to understand the relationship between gross capital requirements and risk mitigation effects, supporting supervisory assessments of whether institutions appropriately calculate their capital positions. Supervisors can leverage this structured information to build analytical capabilities that automatically verify calculation consistency, identify institutions with unusual risk mitigation patterns, and assess the aggregate effectiveness of risk mitigation techniques across the insurance sector.
This information proves particularly valuable for sustainability-related supervision because many catastrophe risks increasingly reflect environmental factors such as extreme weather events or natural disasters that may be intensified by climate change. The taxonomy structure enables analytical systems to connect traditional catastrophe risk assessment with broader sustainability risk monitoring.
Product-Level Sustainability Integration Assessment
The Solvency II XBRL taxonomy includes concepts for analysing how insurance products integrate risk prevention measures, including the total amount of gross written premiums for products where allowance is made for risk prevention measures, as captured in table S.14.02. This information enables supervisors to assess the extent to which insurance undertakings incorporate sustainability considerations into their product design and pricing.
For analytical systems, the taxonomy structure provides the framework necessary to calculate meaningful metrics about sustainability integration at the product level. Supervisors can leverage this information to build analytical capabilities that assess what proportion of an institution’s business incorporates risk-prevention measures, compare product-level sustainability integration across different institutions, and identify trends in how the insurance sector incorporates sustainability considerations into core business activities.
The use of a monetary data type for the “gross written premium” concept ensures that analytical systems can perform appropriate aggregations and comparisons that reflect the scale and significance of sustainability-integrated products within institutions’ overall business portfolios.
Policy Design Analysis Through Boolean Indicators
Boolean concepts indicating whether product design makes allowance for risk prevention measures, also captured in table S.14.02, provide supervisors with clear binary indicators that enable systematic assessment of policy approaches across large numbers of institutions and products.
The taxonomy structure enables analytical systems to quickly calculate adoption rates for risk prevention measures, identify institutions that lag behind sector practices, and correlate policy design approaches with other sustainability-related metrics. Supervisors can leverage this structured information to assess the effectiveness of regulatory guidance regarding sustainability integration and to identify areas where additional supervisory attention might be needed.
The boolean data type also supports analytical systems that need to perform categorical analysis, enabling supervisors to group institutions based on their sustainability integration approaches and to assess whether different policy design strategies correlate with different risk profiles or financial performance patterns.
Supporting Comprehensive Prudential Sustainability Assessment
The integration of these different data types within the Solvency II XBRL taxonomy enables analytical systems to support comprehensive sustainability assessment that spans asset exposure, capital adequacy, product design, and policy implementation. Supervisors can leverage the taxonomy structure to build analytical capabilities that examine sustainability considerations holistically rather than addressing them as isolated reporting requirements.
This comprehensive approach proves essential for effective prudential supervision because sustainability risks often manifest through complex interactions between different aspects of insurance business. Climate-related asset exposures might interact with product design decisions and capital adequacy assessments in ways that require integrated analytical approaches to understand fully.
Enhancing Supervisory Efficiency Through Structured Analysis
The structured nature of the Solvency II XBRL taxonomy enables analytical systems to automate many aspects of sustainability-related supervisory assessment while maintaining the precision and reliability that prudential supervision requires. Supervisors can leverage the taxonomy structure to build analytical capabilities that automatically flag unusual patterns, verify calculation consistency, and identify institutions that require additional supervisory attention.
This automation enables supervisory resources to focus on analysis and judgement rather than data compilation and basic quality assessment. The taxonomy structure also supports analytical systems that can provide supervisors with comprehensive sustainability risk assessments that integrate seamlessly with broader prudential supervision processes.
The Solvency II example demonstrates how XBRL taxonomy intelligence can enhance supervisory analytical capabilities while maintaining the precision and reliability requirements that prudential oversight demands. Understanding these applications becomes essential for supervisory authorities seeking to integrate sustainability considerations into their core supervisory processes effectively.
Analytical Challenges When Data Lacks Structured Implementation
The absence of structured data formats for certain disclosure requirements creates substantial analytical challenges for supervisors and regulators. The Sustainable Finance Disclosure Regulation is one example of these difficulties, as it mandates comprehensive sustainability disclosures but requires publication in human-readable formats such as PDF documents rather than machine-readable structured data. This approach, whilst ensuring accessibility for human readers, may create significant obstacles for systematic data analysis and regulatory oversight.
Data Extraction Challenges from Unstructured Documents
When financial institutions publish disclosures as PDF documents created from word processing templates, supervisors face immediate technical challenges in extracting usable data. Unlike structured XBRL reports where data points are clearly tagged with metadata, PDF documents present information in formats designed for human consumption rather than automated processing.
Optical Character Recognition technology can convert PDF text into machine-readable format, but this process introduces multiple potential sources of error and ambiguity. The accuracy of OCR depends heavily on document quality, font choices, and layout complexity. Although the mechanisms for extraction of data from PDF documents are being constantly improved, they may struggle when dealing with tables within these documents. In particular, maintaining the relationships between column headers and data values constitutes a challenge, potentially misaligning critical information.
Modern AI capabilities, including large language models, offer more sophisticated approaches to extracting information from unstructured documents. These systems can identify and extract specific data points even when they appear in varied formats or locations within documents. However, AI-based extraction introduces different challenges, including potential interpretation errors, inconsistent extraction across similar documents, and difficulty in verifying the accuracy of extracted information without manual review.
Analytical System Complexity Without Structured Metadata
The absence of structured data and metadata fundamentally changes how analytical systems must be designed and operated. When working with SFDR PDF disclosures, data users cannot rely on predefined taxonomies that specify data meanings, relationships, and validation rules. Instead, analytical systems must incorporate complex interpretation logic to understand what extracted data represents and how different pieces of information relate to each other.
Information which in XBRL-based data is defined at the taxonomy level in the form of metadata (e.g. element data types, references to legal acts, documentation labels or relationships between elements), needs to be hardcoded at a system level when data is collected using PDF or other unstructured data formats.
Even seemingly simple true/false statements such as ‘Does this financial product have a sustainable investment objective?’ become analytically complex in unstructured formats. Whilst the response might appear straightforward, analytical systems must verify that they have correctly identified the relevant question and associated response, particularly when documents contain multiple similar questions or when institutions provide qualified responses that include explanatory text.
Challenges with Numerical Data Identification
Consider the challenge of analysing numerical data such as ‘Investee countries subject to social violations’. In a structured system, this data point would be clearly tagged with metadata indicating its meaning, units, and calculation methodology. In PDF format, analytical systems must first locate this information within potentially variable document layouts, extract the associated numerical value, and then infer the meaning from surrounding context.
The extraction process becomes particularly complex when documents use different formatting approaches, abbreviations, or presentation styles. One institution might present this information in a clearly labelled table, whilst another might embed it within narrative text. Analytical systems must accommodate this variability whilst maintaining consistency in how they interpret and categorise the extracted information. When information is expressed as monetary values, data extraction mechanisms must ensure that values and their corresponding currency definitions are processed together.
KPI Analysis Complications
Key performance indicators such as “Scope 1 GHG emissions” or “Board gender diversity” present additional analytical challenges when disclosed in unstructured formats. These metrics require not only accurate numerical extraction but also understanding of the units, scope, and calculation methodology used. Without structured metadata, analytical systems cannot automatically verify whether different institutions use consistent measurement approaches or reporting boundaries.
The lack of structured dimension information means that analytical systems cannot easily aggregate or compare KPIs across multiple institutions. With PDF disclosures, this contextual information must be extracted from narrative descriptions that may be incomplete or inconsistent across different institutions.
Narrative Content Processing Difficulties
Narrative disclosures such as “Description of policies to identify and prioritise principal adverse impacts on sustainability factors” require sophisticated text analysis capabilities to extract meaningful insights. Unlike structured text blocks that include clear semantic tags, these narrative sections appear as free-form text that may vary significantly in structure, detail, and terminology across different institutions.
Analytical systems must employ natural language processing techniques to identify key themes, policy elements, and implementation approaches within these narratives. However, the absence of structured semantic markup means that automated analysis cannot easily distinguish between substantive policy descriptions and generic compliance language, or between specific commitments and aspirational statements.
Mixed Content Analysis Problems
Content that combines narrative and numerical information, such as responses to “What is the minimum share of investments in transitional and enabling activities?”, creates particularly complex analytical challenges. These disclosures may include percentage figures, absolute amounts, calculation methodologies, and qualitative explanations within single blocks of text.
Analytical systems must extract multiple types of information from these mixed content sections whilst maintaining the relationships between different elements. The absence of structured tagging means that systems cannot easily distinguish between historical performance data, future commitments, and hypothetical scenarios that might all appear within the same disclosure section.
Quality Assurance and Validation Difficulties
The absence of structured validation rules means that analytical systems cannot automatically verify the consistency or completeness of extracted data. In structured reporting systems, built-in validation rules can flag mathematical inconsistencies, missing required disclosures, or inappropriate data relationships.
With unstructured data, developing equivalent quality assurance capabilities requires creating custom validation logic that can accommodate the variability inherent in free-form document formats. The quality assurance process may require human engagement to verify the results of inconsistency checks and decide how to proceed with the encountered errors.
Comparative Analysis Limitations
Perhaps most significantly, the lack of structured data severely limits data users’ ability to conduct systematic comparative analysis across multiple institutions and/or products. Without consistent data organisation and semantic markup, analytical systems cannot easily identify comparable metrics, aggregate sector-level trends, or benchmark individual institution performance against peer groups.
These analytical limitations fundamentally constrain data users’ capabilities, making it difficult to identify emerging risks, assess sector-wide compliance patterns, or develop evidence-based policy responses. The contrast with structured reporting approaches demonstrates how data format choices significantly affect analytical capabilities.
Understanding these challenges becomes essential for regulatory authorities considering implementation approaches for new sustainability disclosure requirements and for evaluating the costs and benefits of structured versus unstructured reporting mandates.
Summary
Building effective analytical systems requires a systematic approach that matches technical capabilities to data characteristics and user requirements. The foundation begins with establishing clear analytical objectives aligned with institutional mandates and understanding diverse user needs across organisational levels.
Data format characteristics fundamentally determine analytical possibilities. Structured data with rich metadata, particularly XBRL-formatted information, enables sophisticated analytical capabilities including automated validation, dimensional analysis, and semantic relationship preservation. Traditional business intelligence tools work well with structured numeric data, while narrative content requires modern AI-based approaches for systematic analysis.
XBRL taxonomies provide exceptional analytical opportunities when properly leveraged. The structural elements such as presentation relationships, definition linkbases, dimensional capabilities, and data type definitions, enable analytical systems that preserve semantic meaning rather than treating XBRL as simple numeric data. Practical applications in frameworks like sustainability disclosures and supervisory reporting systems such as Solvency II demonstrate how taxonomy intelligence enhances regulatory analysis capabilities.
The contrast with unstructured formats reveals significant analytical limitations. PDF-based disclosures create substantial challenges for data extraction, quality assurance, and comparative analysis. These limitations fundamentally constrain supervisory capabilities and highlight the importance of structured data formats for effective regulatory oversight.
Successful analytical systems often require integration strategies that combine multiple data sources and processing approaches while maintaining consistency and quality. The choice between traditional business intelligence and AI-based solutions depends on the balance of numeric versus narrative content and the complexity of analytical requirements.
Understanding these principles enables regulators and market participants to design analytical capabilities that effectively support oversight objectives while accommodating the diverse characteristics of information across different reporting frameworks and implementation approaches.