Toolbox
Fiscal Innovation in Your Hands
How is a DIARI created?
The creation of a DIARI (Information, Analysis, and Immediate Reaction Unit) within an oversight body like the Office of the Comptroller General involves a structured process ranging from conceptualization to effective implementation and operation. Below, I outline the typical steps in this process:
DocumentDefinition of an Analytical Model within the DIARI Context
An Analytical Model is a strategic asset and an advanced analytical tool designed to transform massive volumes of public data into actionable insights, enabling timely decision-making.
Document
Information, Analysis, and Immediate Reaction Directorate (DIARI), a specialized unit of the Office of the Comptroller General of Colombia, established under Legislative Act 04 of 2019 and regulated by Decrees 2037 of 2019 and 403 of 2020. Its purpose is to modernize and enhance fiscal oversight processes by incorporating advanced technological tools and innovative approaches.
FUNCTIONS AND OBJECTIVES
OF DIARI
CREATION CONTEXT
This initiative addresses the identified need to modernize fiscal oversight and keep pace with continuous technological advancements. Furthermore, it aims to provide a timely response to emergencies and critical situations affecting State assets, as noted in its institutional evolution. Through this Directorate, the objective is to optimize the surveillance of resource management by those designated as fiscal managers (gestores fiscales), in accordance with the definition established in Article 3 of Law 610 of 2000, currently in force.
The DIARI represents a significant shift in how fiscal oversight is conducted in Colombia, enabling a more proactive and efficient approach to the administration of public resources, as well as the prevention of risks arising from mismanagement or irregularities.
In summary, the DIARI is a specialized unit of the Office of the Comptroller General of Colombia, dedicated to transforming fiscal oversight through the efficient use of technology and rapid response to challenges in public resource management.
LEARN MORE HERE
Data Maturity Form
How to create a DIARI?
Development Methodology for Analytical Models
DATA MATURITY ASSESSMENT TOOL
This document provides a structured framework to diagnose the level of maturity regarding data usage, management, and exploitation within an organization. The objective is to identify the entity's current position within a maturity model and guide improvement actions toward a data-driven organizational culture.
1
Context
Data maturity is defined as an organization's ability to capture, manage, analyze, and leverage information for decision-making. Various models, such as the Data Management Maturity (CMMI), the Gartner Analytics Maturity Model, and the TDWI model, allow organizations to be classified into levels ranging from initial and reactive data management to advanced strategic use based on Artificial Intelligence and prescriptive analytics.
Assessing maturity not only provides a diagnosis but also outlines a roadmap to progressively evolve from basic practices toward a comprehensive data governance and analytics model.
2
Maturity Levels
- Level 1 – Initial: Fragmented use of data in silos; basic tools.
- Level 2 – Repeatable: Basic processes defined; centralized databases.
- Level 3 – Defined: Data governance and quality; use of BI; tactical decision-making.
- Level 4 – Advanced: Statistical models and Machine Learning; data integration.
- Level 5 – Optimized: Data-driven culture; prescriptive decisions; innovation.
3
Instructions for Use
Please respond to each item in the questionnaire using the following Likert scale:
- 1 = Not implemented
- 2 = Partially implemented
- 3 = Implemented in some areas
- 4 = Implemented in most areas
- 5 = Fully implemented across the entire organization
4
Evaluation Form
Basic Data Management
1. Data is stored in individual files without integration.
2. Information silos exist between different areas of the organization.
3. There are defined leads responsible for data capture and storage.
4. Basic data quality processes (validation, standardization) are applied.
Governance and Organization
5. There is a formal data governance framework with clear roles.
6. Policies for data quality, security, and protection are applied.
7. Data is managed centrally (Data Warehouse, Data Lake).
8. Leadership uses reliable reports and indicators for tactical decisions.
Analysis and Data Exploitation
9. Business Intelligence tools are used for reporting and dashboards.
10. Statistical or Machine Learning models are developed.
11. Internal data is integrated with external sources (customers, market, open data).
12. Data analysis influences strategic decisions.
Data-Driven Culture and Strategy
13. The organizational culture is oriented toward data-driven decision-making.
14. Prescriptive models (AI, optimization, simulation) are applied.
15. Data is considered a strategic asset (e.g., monetization, innovation).
16. The impact of data-driven decisions is measured and fed back into the cycle.
Definition of an Information System for a DIARI
Immediate Response Strategies in DIARI
Continuous Improvement Strategies in DIARI
LEGAL AND REGULATORY FRAMEWORK FOR THE CREATION OF A DIARI
The Information, Analysis, and Immediate Reaction Directorate (DIARI) is a strategic unit within modern fiscal control. This guide integrates conceptual and applied elements of the regulatory framework necessary to implement a DIARI, taking the Colombian case (a pioneer in Latin America), as a reference. Below is the normative, structural, and operational step-by-step process based on the 2019 constitutional reform and the practices adopted by the Office of the Comptroller General of the Republic of Colombia.
DOCUMENTATION
Recommendations for Replicability
- Secure a clear constitutional and legal framework that explicitly supports concomitant and preventive control.
- Define technical and organizational regulations to allow for a flexible and modern institutional structure.
- Establish robust internal units equipped with appropriate technology and highly skilled talent.
- Develop clearly defined processes and procedures for each unit within the Directorate.
- Document and publish operational protocols that are simple, clear, and fully traceable.
- Foster a culture of data governance and high responsibility in information management.
DELIMITATION OF AN INFORMATION SYSTEM FOR A DIARI
The operational core is not merely technology; it is a structure that enables the transformation of information into timely decisions for optimal fiscal control and oversight. This section explains, in a clear and accessible manner, the key components of a Directorate: its purpose, tools, infrastructure, human talent, and the data governance that regulates it.
IMMEDIATE REACTION STRATEGIES WITHIN THE DIARI
These strategies constitute the active component of the continuous monitoring mechanism used to oversee the management of public resources. They align with the new powers granted by the constitutional reform, enabling preventive and concomitant oversight in real-time (Legislative Act 04 of 2019 and Decree 403 of 2020).
Immediate Reaction Actions
Activation of Real-Time Monitoring: The process begins with the activation of the preventive oversight mechanism. Its primary inputs are the alerts generated by analytical models, based on data previously collected by the Information Unit.
Risk Identification: The Information Analysis group identifies potential risks and/or non-compliance in projects, programs, or contracts essential to the population's basic needs.
Unlocking Bottlenecks: The mission of the Immediate Reaction group is to resolve the circumstances preventing the timely completion of projects. Their goal is to ensure the project's viability and delivery to beneficiaries. Crucially, this task does not imply co-management (the oversight body does not take over the executive role).
Field Intervention Actions (In Territory)
The Immediate Reaction Unit (URI) deploys an interdisciplinary team that communicates directly with project managers and, when necessary, visits the execution site to verify the physical status of projects and infrastructure works. This on-site verification triggers actions to facilitate the timely resolution of identified risks.
Key interventions include:
- Reactivating Stalled Projects: Managing the resumption of works halted without technical justification through direct dialogue with executing entities.
- Verifying Delivery: Ensuring the effective delivery of prioritized goods and services, such as school feeding programs (PAE), equipment, or infrastructure.
- Inter-institutional Coordination: Coordinating with national or territorial entities to overcome administrative, contractual, or technical obstacles delaying execution.
These actions are carried out within short timeframes, focusing on early intervention and technical support to facilitate the efficient use of public resources.
Note on Legal Escalation: If field exercises identify signals of severe impact or non-compliance with potential legal consequences, the DIARI documents the findings and refers them to the competent authorities for formal investigation.
Results: Qualitative and Quantitative Benefits
The results of these continuous monitoring actions are categorized into: Quantitative Benefits: Savings in public funds, reduction in cost overruns, and completion of infrastructure; Qualitative Benefits: Improved public trust in institutions and the early identification of systemic risks in resource management.
Pedagogical Value and Transparency
These strategies rely on seamless coordination between the DIARI's internal units (Information, Analysis, and Reaction) and external stakeholders such as ministries, local governments, and investment funds.
This articulation ensures that alerts translate into rapid, viable, and sustainable actions, preventing fiscal control from being limited to reports without practical impact.
Inter-institutional Coordination
The outcomes produced by the DIARI promote a culture of prevention and timely action, bolstered by social control (citizen oversight). Internally, these results serve as vital inputs for: Ex-post and Selective Control: Feeding the traditional audit cycle with high-quality data; Fiscal Liability Proceedings: Providing evidence and technical grounds for formal legal processes against misuse of funds.
CONTINUOUS IMPROVEMENT STRATEGIES FOR A DIARI
Media and Social Network Communication Model
The DIARI operates within a dynamic environment based on the automated collection and processing of news through web scraping (e.g., Google News and other open sources). This unstructured data is integrated into the DIARI’s structured databases and processed using Python scripts and SPSS Modeler. The results are visualized in Power BI Report Server dashboards, which are publicly available for online consultation. This ensures transparency and provides open access to real-time dynamics of the country's public and political agenda.
Efficient Use of Available Resources
Implementing a DIARI requires more than just legislative procedures; it demands adequate funding. However, entities aiming to create such a unit can initiate preliminary analytical exercises using available public databases and existing institutional licenses. This "lean" approach allows for demonstrating value before full-scale investment.
Establishing Strategic Alliances
Collaboration with universities, research centers, multilateral organizations, and tech companies strengthens the DIARI’s capabilities. These alliances are key to:
- Accessing specialized and up-to-date knowledge.
- Incorporating advanced data analysis methodologies.
- Supporting human talent training and professional development.
- Scaling technology without relying exclusively on internal resources.
Democratization of Knowledge
An effective strategy to maximize impact is to democratize findings across different institutional areas. Through interactive and updated control dashboards, officials can interpret information and perform variable cross-referencing, generating alerts that serve as critical inputs for fiscal oversight and control.
Systematization of Best and Worst Practices
Decree 403 of 2020 (which regulated Legislative Act 04 of 2019) mandates the creation of a Bank of Best and Worst Practices. The goal is to evaluate the management and results of audited entities. This exercise strengthens decision-making, prevents the recurrence of misconduct, and provides essential criteria for refining the risk indicators in analytical models.
Alignment with Institutional Strategic Plans
To guarantee the long-term sustainability of a DIARI, it must be fully aligned with the primary plans, programs, and projects of the Supreme Audit Institution. Its objectives, outputs, and lines of action must be coordinated and supported by clear processes and procedures.
Like any core department, a DIARI must have a Mission, Vision, and specific Objectives that resonate with the high-level policies established by senior management.
Methodology for Model Creation
Phases in the Development and Execution of a Model
Technical Team and Specialized Infrastructure
Analytical Models in the Initial Phase with Open Data Sources
METHODOLOGY FOR MODEL CREATION
The creation of a model within the DIARI (Information, Analysis, and Immediate Reaction Directorate) framework is based on a rigorous methodology that ensures conceptual relevance, analytical robustness, operational functionality, and strategic value as an input for decision-making. This methodology integrates institutional analysis, data science, software engineering, data governance, and knowledge management.
The following are its fundamental phases:
Identification and Prioritization of Public Needs
Every model originates from a specific oversight or monitoring requirement:
It may respond to citizen complaints, strategic mandates from the oversight body, priority public policies, or suspected misuse of resources. Prioritization considers social impact, problem recurrence, and the feasibility of detection based on available data sources. This phase defines the type of control to be enhanced (preventive, concomitant, posterior, or selective).
Definition of Scope, Pedagogical Purpose, and Analytical Objectives
The technical objectives are established, such as:
Predicting anomalies, classifying risks, or estimating compliance. The target audience for the results (oversight units) is identified. Analytical and pedagogical hypotheses are developed, including alerting criteria and information thresholds.
Mapping and Characterization of Information Sources
Identification of structured and unstructured data sources to feed the model:
Databases, flat files, spreadsheets, open data, imagery, and non-standardized documents. Sources extend beyond traditional government entities to include any entity managing public resources. Analysis covers schemas, frequencies, fields, metadata, and interoperability requirements.
Data Processing, Traceability, and Governance
Data is extracted, transformed, and loaded (ETL) following a structured pipeline:
This process includes integrity verification, normalization, documented traceability, quality auditing, and cataloging in institutional platforms.
Technical Model Design
Based on objectives, appropriate algorithms are selected (e.g., classifiers, regressors, decision trees, neural networks): Definition of independent and dependent variables and business rules; Structuring of development notebooks; Documentation of assumptions, constraints, and calculation logic within systems like DevOps to ensure reproducibility and technical clarity.
Training, Calibration, and Evaluation
The model is trained using historical data, employing techniques such as cross-validation and grid search for hyperparameter optimization. Performance is measured by: Accuracy, sensitivity, specificity, and Area Under the ROC Curve (AUC); Quantification of False Positives and False Negatives; Workshops with Subject Matter Experts (SMEs) to validate alert consistency and operational relevance.
Operational Implementation and Production (Deployment)
Once validated, the model is integrated into the DIARI infrastructure: Defined execution cycles (daily, weekly, monthly); Output channels: Dashboards, automated reports, and real-time alerts; Established mechanisms for version control, backups, and service continuity.
Continuous Monitoring, Feedback, and Evolutionary Improvement
The model is periodically evaluated against new data: Variable adjustment and alert threshold refinement based on environmental changes; Incident documentation and continuous improvement through feedback loops with internal and external users; Ensuring the model remains dynamic and capable of anticipating risks.
Training and Social Appropriation of Knowledge
The methodology includes the design of pedagogical materials (infographics, tutorials, courses) and communication strategies (data visualizations, data storytelling): Facilitating understanding and adoption by key stakeholders; Ensuring the generated knowledge is useful, replicable, and sustainable across different institutional contexts.
This comprehensive methodology transforms analytical models into dynamic, collaborative tools backed by evidence. By combining a solid technical foundation with a focus on preventive and selective control, these models enable more efficient and transparent institutional interventions, ultimately aimed at anticipating risks and halting potential financial loss (detrimento patrimonial).
PHASES IN THE DEVELOPMENT AND EXECUTION OF AN ANALYTICAL MODEL
The development and execution process of a model within the DIARI framework follows a detailed operational roadmap, with clearly defined phases that articulate multiple teams and competencies:
Click the icon to learn more
Phase 1
Phase 2
Phase 3
Phase 4
Phase 5
Phase 6
Phase 7
Phase 8
Phase 9
Phase 10
Requirement Alignment and Scoping
Once an institutional problem or monitoring line is identified, the team conducts a formal alignment for the proposed model:
- Meetings between analysts, technical staff, and institutional stakeholders.
- Definition of the model’s scope.
- Construction of the guiding question: What needs to be detected and why?
- Definition of the model type (Predictive and/or Descriptive).
Data Dictionary Design
Once the requirements are clear, the necessary Data Dictionary is defined, containing:
- Required variables, types, and formats.
- Expected frequency and granularity.
- Data sources (Sectorized or Non-Sectorized).
- If necessary, technical sessions are held with data-providing entities to review and validate the dictionary structure.
Data Collection and Channeling
- Sectorized Entities: Monitoring of data submissions in compliance with formal resolutions.
- Non-Sectorized Entities: Data ingestion via the APPUI application.
- All data enters the management system and undergoes an automated validation process.
Data Quality Process
This ensures that information is accurate, complete, consistent, and reliable. It includes: • Profiling, cleansing, standardization, validation, and continuous monitoring.
• Data is organized through a tiered architecture (Bronze, Silver, Gold) based on its transformation level.
Analytical Model Development
The technical and analysis teams initiate the build:
- Development of scripts in Python, SQL, or other languages.
- Design of algorithms and Business Rules.
- Incorporation of Subject Matter Expertise (SME) from both thematic and technical perspectives.
- Internal conceptual validations to fine-tune the model.
Validation and Simulation
Before entering production, the model undergoes testing:
- Behavioral simulation using historical data.
- Validation of expected results and outputs.
- Refinement of alert thresholds and logic rules.
- Performance verification regarding False Positives and False Negatives.
DevOps Documentation
Every model is technically and methodologically documented, including:
- Model logic and utilized sources.
- Code versioning.
- Alert interpretation manual.
- Integration with alert production workflows.
- Final technical report.
Operational Execution
Once validated, the model enters a regular execution cycle (weekly, monthly, etc.).
resulting products may include:
- Automated alerts for analyst review to determine actionable relevance.
- Analytical dashboards.
- Written reports and inputs for mission-critical or preventive decision-making.
Monitoring and Feedback
Post-execution control is performed after each cycle:
- Comparison of results against real-world findings.
- Alert analysis to evaluate model effectiveness.
- Correction of deviations or variable adjustments.
Update and Continuous Improvement
Models are dynamic and undergo improvements when:
- New patterns are identified.
- New data sources are integrated.
- Regulatory or operational contexts change. All updates follow a Version Control cycle and are communicated to end-users.
TECHNICAL TEAM AND SPECIALIZED INFRASTRUCTURE FOR ANALYTICAL MODEL DEVELOPMENT AND EXECUTION
The implementation of analytical models within a DIARI requires an operational framework consisting of specialized technical talent, clearly defined workflows, and an advanced technological infrastructure. This ecosystem ensures the integrity, availability, utility, and security of public data, in compliance with international standards such as ISO norms.
This human and technological architecture enables not only the generation of accurate and timely models, but also their long-term sustainability, turning them into reliable tools for public oversight and strategic state intervention. It also reflects a growing level of technological maturity, in which the Supreme Audit Institution (SAI) implements advanced data-analysis techniques, optimizes audit processes, improves decision-making, and strengthens the early detection of irregularities through the smart use of technology.
VIDEOS OF ANALYTICAL MODELS
ANALYTICAL MODELS: INITIAL PHASE UTILIZING OPEN DATA SOURCES
This document presents two analytical models developed within the DIARI framework as part of an initial implementation phase. This phase is characterized by the strategic use of Open Data and official online sources. These models enable oversight and surveillance activities based on information from entities managing State resources, facilitating audit processes that are more agile, accessible, and replicable.
Infrastructure Model
The Infrastructure Model is an analytical tool developed to monitor the physical and financial behavior of public resources executed nationwide using State funds. It enables the detection of potential irregularities and facilitates timely institutional actions. Key features include: Project Monitoring: Real-time tracking of public works and construction; Risk Detection: Identification of delays, cost overruns, or abandonment of project; Geospatial Integration: Visualization of infrastructure impact across the national territory.
Click to download information
Budgetary Chain Model
The Budgetary Chain Model is a tool designed to ensure the traceability of public spending, from the initial planning phase through to final execution. It provides audit and oversight teams with consolidated information for the surveillance and monitoring of the National General Budget (Presupuesto General de la Nación). Its primary functions include: Spending Traceability: Mapping the flow of funds across different administrative levels; Execution Analysis: Comparing planned vs. actual spending to detect deviations; Data Integration: Consolidating financial data from various official sources to provide a unified view for fiscal control.
Click to download information