September 10, 2026
Data Governance and Financial Risk in the Forest Sector: Empirical Evidence in a Brazil-Finland Perspective
Fernanda Desimon Testa da Silva; Juliano Souza Vasconcelos
DOI: 10.22167/2675-6528-202602216
Article derived from a Final Course Work (TCC), with content based on the student’s original work and adapted to the editorial format of the Revista E&S with the support of the ResumeAI tool, an artificial intelligence solution developed by the Pecege Institute for textual synthesis and organization.
Summary
Informational fragmentation in the Brazilian forestry sector was the starting point of this research, which investigated how data governance impacted analytical capacity, financial risk, and the competitiveness of forest management, through a comparative study between Brazil and Finland. 332 documents were submitted, including 307 CEPEA/Esalq/USP reports (2001–2025) and 25 sectoral reports (Bracelpa, ABRAF, Ibá, 2003–2025), to an automated extraction pipeline based on OCR, text mining, and regular expressions, yielding 13,059 structured records. Sensitivity analysis of Net Present Value (NPV) and lexical sentiment analysis were performed. Only 12.1% of cost records contained the Effective Operational Cost filled out, and none presented a calculable net margin. Uncertainty in input data fluctuated the NPV of a forestry project by more than R$ 20,000/ha, evidencing the distortion of data-driven decision-making when algorithmic sophistication did not replace the quality of the source data. Lexical sentiment analysis confirmed the transition of Brazilian sectoral reports from a statistical to a promotional format. In contrast, the Finnish model demonstrated the viability of integrated governance, with consolidated forest inventory as a pillar of industrial intelligence. The results indicated that strengthening data governance was a prerequisite for the sector’s transition to an evidence-based planning system.
Keywords: competitiveness; informational fragmentation; artificial intelligence; digital sovereignty; decision-making.
1. Introduction
Currently, data occupies the center of geopolitical and economic disputes, being frequently compared to commodities like oil due to its strategic value. In the context of digital transformations driven by Artificial Intelligence, the ability to process large volumes of data has become an indispensable asset for the competitive advantage of nations and organizations (Provost and Fawcett, 2013; Turban and Volonino, 2013; Brazil, 2025). Just as with raw materials, the value of data lies in its treatment: when integrated into statistical and Machine Learning models, they are transformed into critical information and knowledge for sovereign decision-making.
Despite Brazil possessing historical institutions for systematic data collection, such as the Brazilian Institute of Geography and Statistics and the Informatics Department of the Unified Health System, significant barriers to the integration and reuse of its databases persist. Although the country shows good open data rates, a structural flaw in the reuse pillar is observed: the data is available, but lacks a workable format (Lafuente, 2020). The absence of standardization results in deficient interoperability, keeping informational assets fragmented in institutional silos and increasing analytical latency and operational risk (Belli et al., 2024). Brazil’s digital governance structure has advanced through successive frameworks — from the Marco Civil da Internet (Brazil, 2014) and the General Data Protection Law (Brazil, 2018) to Decrees No. 12.069 and 12.198 (Brazil, 2024) —, however, the practical implementation of interoperability between agencies remains incipient.
This fragmentation manifested itself in a particularly critical way in the forestry sector, where access to operational data is crucial for short- and long-term strategic decisions, both in ecosystem conservation and production efficiency. Data gaps limit the application of Data Engineering and Data Science techniques, tools that are now indispensable for the competitiveness of the private sector and for the development of adaptation plans in the face of extreme climate events. The modernization of the sector depends not only on technological advances but also on governance that integrates technical competence and systemic interoperability, given that analytical effort requires the cross-referencing of heterogeneous sources — foreign trade data, records from regulatory bodies, land registries, and geospatial monitoring platforms — in a context of scarcity of specialized human capital (Silva and Costa, 2024).
From the perspective of operational efficiency, the quality of decisions is intrinsic to the precision of informational inputs; inconsistencies feed the Garbage In, Garbage Out principle, generating uncertainties and financial losses (Turban and Volonino, 2013; Ferrari and Silva, 2017). For the forestry sector, data integrity has been the foundation for the transition from intuitive management to rational management, where information is organized to achieve sustainable objectives with efficiency (Schneider, 2004). The absence of standardization and the lack of articulation between public and private databases introduce latent operational and financial risks, capable of catalyzing failures in management execution and budget overruns. In this sense, data governance acts as a prerequisite for the transition from empirical management to evidence-based decision-making, providing the necessary inputs for models that mimic biological and economic reality (Anderson, 2015).
In contrast, international models such as the Finnish one have demonstrated the viability of integrated governance, in which forest biophysical data are open by law, national inventories are continuously updated by remote sensing, and owners exercise sovereignty over their information through digital platforms (Finland, 2011; Sitra, 2021; Metsäkeskus, 2023). The contrast between the two scenarios highlighted that the efficiency of the bioeconomy depends not only on the gross availability of data but on the existence of digital ecosystems that convert them into intelligible assets. Faced with this scenario, the present research seeks to answer the following question: how do differences in data governance between Brazil and Finland impact the analytical capacity, financial risk, and efficiency of evidence-based forest management? To this end, the work adopts an empirical approach, subjecting real data from the Brazilian forest sector to an automated extraction and structuring pipeline, and confronting the results obtained with the informational maturity of the Nordic model.
2. Material and Methods
2.1. Classification of the research
The research is exploratory in nature and quali-quantitative in approach (Yin, 2015). It combined documentary analysis and comparison of governance models (qualitative dimension) with automated extraction of 13,059 records, lexical sentiment analysis, and Net Present Value (NPV) simulation (quantitative dimension). The methodological design investigated the maturity of data governance in the forestry sector, analyzing the relationships between private market practices, public policies, and State digital sovereignty models. The study focused on the fragmentation of forest management into operational silos (silviculture, harvesting, and logistics) in Brazil and Finland, treating fragmentation as a critical data governance failure and a vector of operational risk.
2.2. Database
The data for analysis were obtained from public and private entity databases, organized into three categories, as synthesized in Table 1.
Table 1. Data sources used in the research
| Category | Source | Period | Documents | Main data |
|---|---|---|---|---|
| Sectoral reports | Brazilian Association of Pulp and Paper [Bracelpa] | 2003–2008 | 6 | Planted area, production |
| Brazilian Association of Planted Forests Producers [ABRAF] | 2006–2012 | 7 | Area, production, exports, indicators | |
| Brazilian Tree Industry [Ibá] | 2014–2025 | 12 | Area, production, ESG indicators | |
| Market Information | Center for Advanced Studies in Applied Economics [CEPEA/Esalq/USP] | 2001–2025 | 307 | Forest product prices |
| Production costs | Future Farm Project — CNA/Senar | Several | 36 | COE, COT, revenues, implementation costs |
| Public data | SNIF, IBGE | Several | — | Complementary official statistics |
| “Finnish” benchmark | Luke, Metsäkeskus, Metsään.fi, Sitra | Several | — | Governance, “Fair Data Economy”, inventory |
Source: Elaborated by the author (2026)
2.3. Environment configuration
The development was carried out in a macOS environment, using pyenv for Python 3.12 version management and virtualenv for dependency isolation. The tools used in the data “pipeline” and their functions are detailed in Table 2.
Table 2. Tools used and functions in the data “pipeline”
| Tool | Function | Processed sources |
|---|---|---|
| Tesseract OCR | Optical character recognition of scanned PDFs | Bracelpa, ABRAF |
| Poppler (pdf2image) | Conversion of PDF pages to images (300 DPI) for OCR | Bracelpa, ABRAF |
| pdfplumber | Extraction of text and tables from native PDFs | CNA/Senar, Ibá, CEPEA |
| tabula-py | “Fallback” for native table extraction | CEPEA |
| Pandas | Structuring in DataFrames and CSV export | All |
| SQLite | Relational storage with normalized “schema” | All |
| Regular expressions (re module) | Identification of numerical and textual patterns | All |
Source: Elaborated by the author (2026)
2.4. Extraction and Storage Pipeline
Due to the heterogeneity of the documents, two complementary extraction “pipelines” were built: one for scanned documents (images) via OCR and another for native PDFs, with direct extraction of text and tables. The selection logic and techniques of each flow are synthesized in Table 3.
Table 3. Comparison of the two extraction “pipelines”
| Characteristic | OCR Pipeline | Pipeline pdfplumber |
|---|---|---|
| Sources | Bracelpa, ABRAF, Ibá (sector reports) | CNA/Senar, CEPEA (newsletters) |
| Nature of the PDF | Digitized, image-intensive | Digital natives |
| Main tool | Tesseract OCR (via pdf2image at 300 DPI) | pdfplumber (with “fallback” to tabula-py) |
| Preprocessing | Automatic rotation detection via OSD; correction of rotated pages | Not necessary |
| Extraction mode | PSM 6 (single block) with preservation of word spacing | extract_tables() with classification by keywords (COE, COT, revenue, margin, IMA) |
| Treatment of data in running text | Regular expressions on OCR lines (double/triple space separation) | Regular expressions on native text (e.g., “deployment cost of R$ 5,367.26/ha”) |
| Numerical conversion | BR Format (1,234.56) → computational (1234.56) | |
| Traceability | Source_Origin Column in each record | Source_Origin Column in each record |
| Source code | Author’s repository |
Source: Elaborated by the author (2026)
The data from both streams were stored in a SQLite relational database with a normalized “schema” (tables: documents, costs, cost composition, technical indicators, and market prices) and in CSV replicas for a “data lake”, allowing structured queries (SQL) and exploratory analyses (Pandas).
2.5. Heuristic Refinement with Artificial Intelligence
The extraction “pipelines” were optimized through iterative heuristic tuning, with the aid of two commercial Artificial Intelligence (AI) models. Table 4 details the role of each tool in the process.
Table 4. Use of Artificial Intelligence in the data “pipeline”
| Tool | Model | Pipeline stage | Functions performed |
|---|---|---|---|
| Gemini 3 Flash (Google) | Native multimodal | Pre-processing and OCR “pipeline” (section 2.4) | Automation of massive “downloads”; dynamic creation of extraction “scripts”; rapid prototyping for documents with complex layouts |
| Claude 4 Opus “high-thinking” (Anthropic) via Cursor | LLM with code execution | Refinement and pdfplumber “pipeline” (section 2.4) | Consolidation of historical databases; extraction of production costs (CNA/Senar); retrospective analysis of work versions |
Source: Elaborated by the author (2026)
2.6. Analytical methods
Regarding the extracted and stored data, two complementary analytical methods were applied, synthesized in Table 5.
Table 5. Analytical methods applied to the extracted data
| Method | Objective | Scope | Technique | Reference |
|---|---|---|---|---|
| Text mining | Capture operational data in free text (non-tabular) | 25 sectoral reports (Bracelpa, ABRAF, Ibá) | Search for exact match of 28 typical forestry operations + synonyms (Appendix D); regular expressions for adjacent quantitative data | |
| Sentiment analysis. | Quantify the change in tone of reports over time | 3 representative reports: Bracelpa (2004), ABRAF (2012), Ibá (2025) | Domain-specific lexical approach (Loughran and McDonald, 2011), adapted to the forest sector | Loughran and McDonald (2011) |
Source: Elaborated by the author (2026)
Text mining sought 28 typical forestry operations (Appendix D) and adjacent quantitative data in sectoral reports, using regular expressions to capture non-tabular information and validate extraction. Sentiment analysis followed Loughran and McDonald’s (2011) lexical frequency classification approach, adapted to the Brazilian forestry sector. A domain lexicon in Portuguese (Appendix B) was built, composed of 16 positive, 17 negative, and 16 ESG terms, iteratively refined to remove ambiguities. The normalized sentiment “score” was calculated as the frequency of terms per thousand words, allowing for comparison between documents. The analysis of three representative documents served as a qualitative indication of the tone change over the period.
2.7. Context of the Finnish “benchmark” and Brazil-Finland comparative matrix
For the comparison between the governance models of Brazil and Finland, a matrix was used that crossed governance indicators (transparency, interoperability, legal compliance, data sovereignty) with competitiveness indicators (digital infrastructure, informational transaction costs, access to certification). The analysis was systematic, considering data sustainability and governance as continuous processes of institutional improvement.
The Finnish “benchmark” is based on the concept of “Fair Data Economy”, which balances technological innovation with state sovereignty and citizens’ digital rights. This framework is enabled by legal and technical instruments, such as the Information Management Act (Tiedonhallintalaki) and the decentralized infrastructure of the Suomi.fi portal (Finland, 2019), which promote open data via APIs. The “MyData” model ensures individual sovereignty over information, allowing control over access to personal data (Sitra, 2021).
The management of Finnish natural resources is integrated by Metsäkeskus, Luke, and Syke. The Forest Information System Act (Metsätietolaki) democratized access to biophysical data of nature (Finland, 2011), separating it from personal data (protected by GDPR). This openness is supported by massive and continuous data collection via LiDAR and satellite imagery for high-precision digital forest inventories.
The Metsään.fi portal allows owners to view a digital twin of their forest and grant access to companies, optimizing the industry’s logistical planning (Metsäkeskus, 2023). Networked harvester machines send real-time data, dynamically updating the national inventory. Table 6 summarizes the structural differences identified between the two forest data governance models.
Table 6. Comparative matrix of forest governance: Brazil vs. Finland
| Source | “Score” Positive Sentiment (%%) | “Score” Negative Sentiment (%) | Positive ESG Terms (n) | Negative ESG Terms (n) |
|---|---|---|---|---|
| Bracelpa (2004) | 3,24 | 0,00 | 192 | 19 |
| ABRAF (2012) | 6,18 | 1,87 | 367 | 110 |
| Ibá (2025) | 19,02 | 10,17 | 331 | 28 |
Source: Elaborated by the author (2026)
3. Results and Discussion
Information fragmentation and the hidden cost of data acquisition
Brazil generated and published a large volume of data for multiple purposes. However, the absence of consistent standards in the modeling of these databases and the complex way they were presented severely hindered the analytical reuse of this information, both in the public and private sectors.
In the public sector, institutions such as the Brazilian Institute of Geography and Statistics (IBGE) and CEPEA/Esalq/USP provide a large volume of information, but the architecture of their platforms often hinders large-scale data export. In the private sector, the main difficulty was the way information was made available, with data diluted in sectoral reports or statistical yearbooks of various and regionalized formats.
Most of the effort in this research consisted of locating, extracting, and consolidating these bases, demanding data engineering to transform 332 unstructured documents into permanent information assets. The resulting infrastructure, an SQLite relational database and CSV repository, represented a technical contribution. The cost of this effort was conservatively estimated at over 150 hours of technical work.
Evolution of sectoral reports (2003 – 2025)
National sectoral reports underwent significant changes in content and tone throughout the analyzed period. The documentation went through three distinct phases: in the first (2003–2006), Bracelpa’s statistical yearbooks prioritized quantitative data in charts and tables, favorable for automated extraction. In the second phase (2007–2014), ABRAF’s yearbooks combined technical analyses with tables, charts, and images, resulting in a mixed and partially extractable format. In the third phase (2015–present), Ibá’s reports began to prioritize qualitative content, with texts and images over tables and charts, adopting a predominantly promotional and difficult-to-automate extraction format.
This paradigm shift demanded more aggressive data collection and “data wrangling” techniques, given the absence of public APIs or structured data deliveries. The migration from statistical to promotional reports represented a reduction in the supply of reusable data by the Brazilian private sector, in contrast to the trend of increasing openness observed in Finland.
Sentiment analysis of reports
Lexical sentiment analysis corroborated the transition observed in the reporting phases. The positive “score” grew almost sixfold between 2004 and 2025 (from 3.24‰ to 19.02‰), while the negative one jumped from zero to 10.17‰. Bracelpa’s (2004) zero reflected the purely statistical nature of that document, with neutral technical language. The simultaneous increase in both polarities, with a positive/negative ratio of 1.87:1 in 2025, indicated progressively more emotionally charged language and a deliberate promotional bias. Bracelpa (2004) presented a neutral and statistical profile; ABRAF (2012) reflected a cautious tone, consistent with the economic crisis of the period; and Ibá (2025) exhibited pronounced ESG optimism. From 2015 onwards, reports began to prioritize communication objectives over statistical ones, significantly reducing their usefulness as a source of structured data for quantitative analysis.
Structured data extraction and integrity diagnosis
To quantify the effort of obtaining and the quality of the available data, automated extractions were conducted on monthly bulletins from CEPEA/Esalq/USP (307 bulletins, 2001–2025), sectoral reports from Bracelpa, ABRAF, and Ibá (25 reports, 2003–2025), and cost reports from CNA/Senar (36 documents).
CEPEA Bulletins (307 documents, 2001-2025)
Two complementary “pipelines” were developed: extraction of native tables from PDFs and extraction of prices and percentage variations from narrative text via regular expressions applied to OCR files. The result was the acquisition of structured records, distributed according to Table 7.
Table 7. Structured records extracted from CEPEA/Esalq/USP reports (307 documents, 2001–2025)
| Data type | Records | Period | Method |
|---|---|---|---|
| Domestic market prices (min/avg/max by region) | 3.163 | 2002-2007 | Native PDF Tables |
| Exports (value, average price, quantity) | 6.510 | 2002-2023 | Native PDF Tables |
| Cellulose and paper (domestic prices) | 599 | 2003-2024 | Native PDF Tables |
| Percentage price variations (by region and product) | 982 | 2002-2025 | Regular expressions on narrative text (OCR) |
| Absolute prices (R$) mentioned in the text | 264 | 2002-2025 | Regular expressions on narrative text (OCR) |
| Sum | 11.518 | 2002-2025 |
Source: Elaborated by the author (2026) with the aid of AI
Despite the significant volume, the temporal coverage revealed structural discontinuities: from 2009 onwards, CEPEA reports replaced price tables with graphs, making absolute price data by region inaccessible for automated extraction. For the period 2009–2025, it was only possible to extract percentage variations mentioned in the narrative text, useful for trend analysis, but insufficient to feed NPV models that require absolute values.
Sectoral reports (25 documents, 2003 – 2025)
The extraction of reports from Bracelpa, ABRAF, and Ibá required the construction of specific “parsers” for each organization due to variations in table formats, state nomenclature, and document structure. A total of 1,541 records were obtained, as shown in Table 8.
Table 8. Structured records extracted from the Bracelpa, ABRAF, and Ibá sectoral reports (25 documents, 2003–2025)
| Data type | Records | Period | Sources |
|---|---|---|---|
| Planted area (by state, species) | 1.179 | 1980-2024 | Bracelpa, ABRAF, Ibá |
| Exports (by product, in millions USD) | 52 | 2002-2012 | ABRAF |
| Production (cellulose, paper, panels) | 9 | 2011-2024 | Ibá |
| Economic indicators (VBP, employment, taxes) | 301 | 2006-2024 | ABRAF, Ibá |
| Sum | 1.541 | 1980-2024 |
Source: Elaborated by the author (2026) with the aid of AI
Planted area data dates back to 1980 due to retrospective historical series in ABRAF yearbooks. The “Production” category presents only 9 records because Ibá’s reports provide production data as consolidated national indicators in narrative text, not in standardized tables, hindering automated extraction. In total, the extraction from CEPEA and sectoral sources (Tables 7 and 8) resulted in 13,059 structured records from 332 documents, requiring three distinct Python scripts with over 500 lines of domain-calibrated regular expressions.
CNA/Senar Reports (36 documents)
Of the 36 CNA/Senar documents processed by the same automated “pipeline”, only 33 cost records were extracted. Table 9 presents the fill rate of critical fields for financial analysis.
Table 9. Financial field fill rate extracted from CNA/Senar reports
| Variable | Records with data | Fill rate |
|---|---|---|
| Effective Operational Cost (COE, R$/m³) | 4 of 33 | 12,1% |
| Total Operational Cost (COT, R$/m³) | 14 of 33 | 42,4% |
| Implementation Cost (COT, R$/ha) | 12 of 33 | 36,4% |
| Revenue (R$/m³) | 3 of 33 | 9,1% |
| Net Margin (R$/m³) | 0 of 33 | 0,0% |
Source: Elaborated by the author (2026) from the SQLite database, with the aid of AI
No records presented a calculable net margin; only 9.1% contained revenue and 12.1% contained Effective Operational Cost (COE), essential variables for viability analysis. In addition to the low fill rate, the data presented artifacts in three dimensions: geographical (region fields corrupted by OCR), temporal (numerical values interpreted as dates), and financial (implementation costs ranging from R$ 989/ha to R$ 29,500/ha). Excluding the extremes, the operational range was between R$ 3,000/ha and R$ 14,000/ha. The technical “pipeline” worked correctly, indicating that the problem lay in the quality and standardization of the source data.
The contrast between the volume of records obtained from CEPEA and from sectoral reports (Tables 7 and 8) and the poverty of financial data from CNA (Table 9) revealed a pattern: Brazil produced descriptive data in reasonable quantity (planted area, exports, macro indicators), but the operational data necessary for financial decisions — costs, revenues, margins — remained practically inaccessible in structured format. This fragmentation was not restricted to CNA reports; a review of 12 academic studies published between 1999 and 2020 revealed that each research collected primary data independently, in different regions and base years, reporting costs in incomparable units. This scarcity was not recent, with Rodrigues (1991) already warning about the non-existence of reliable experimental data for Brazilian forest management. The results of this research demonstrated that this effort did not materialize, and the macroeconomic models that underpinned Brazilian public policies never included forest resources as a relevant variable (Bacha, 2004).
The distortion of data-driven decision
The most revealing aspect of this exercise was the methodological path. Before the present extraction round, an attempt to structure the sectoral reports using the Gemini Flash model (Google) resulted in CSVs with numerically corrupted values, such as eucalyptus areas in scientific notation and numeric fields filled with zeros. Given the impossibility of working with this data, the original quantitative approach of financial viability analysis via NPV was replaced by lexical sentiment analysis, not because it was more relevant, but because it was the only viable technique with the obtained data.
This phenomenon was termed “data-driven decision distortion”, occurring when the research question is shaped by data availability, rather than the other way around. In the Brazilian forest context, an analyst seeking to assess the viability of a project cannot find the necessary data in a structured format; if they have AI tools, they obtain corrupted data; and they often end up needing to reformulate their analysis around what is possible to extract, losing the ability to answer the original question.
Illustrative scenario: NPV and the financial risk of uncertainty
To demonstrate the concrete financial impact of uncertainty in the input data, a sensitivity analysis was built by crossing the two parameters with the greatest variability: the deployment cost and the standing timber price.
Table 10. Sensitivity analysis of NPV (R$/ha) for eucalyptus, 7-year rotation, 8% annual rate.
| Implementation cost | Price R$ 100/m³ | Price R$ 130/m³ | Price R$ 160/m³ |
|---|---|---|---|
| R$ 3,000/ha (minimum CNA, excluding artifacts) | +1.236 | +6.014 | +10.793 |
| R$ 5,500/ha (literature reference) | -1.264 | +3.514 | +8.293 |
| R$ 14,000/ha (CNA Jaguariaíva/PR) | -9.764 | -4.986 | -207 |
Source: Elaborated by the author (2026) from data extracted from Senar/CNA reports, with the aid of AI. Values in R$/ha, rounded
Table 10 revealed that, depending exclusively on the deployment cost data used, the same forestry project would present a positive NPV of R$ 10,793/ha (optimistic scenario) or negative of R$ -9,764/ha (pessimistic scenario), an oscillation of over R$ 20,000/ha. This meant that a Brazilian forestry manager attempting to assess the viability of a eucalyptus project based on publicly available data would not be able to determine, with reasonable confidence, whether the investment would generate profit or loss.
This result empirically validated the “Garbage In, Garbage Out” principle in the forest context: uncertainty in the “inputs” is not a theoretical risk, but a concrete barrier to rational decision-making. The contrast with the Finnish model showed that this uncertainty is not inherent to the sector, but a product of data governance.
Governance comparison: Brazil versus Finland
Table 11 summarizes the structural differences identified between the two forest data governance models.
Table 11. Comparative matrix of forest data governance: Brazil vs. Finland
| Dimension | Brazil | Finland |
|---|---|---|
| Legal framework for data | LGPD (2018), Decrees 12.069/12.198 (2024) | Data Administration Act (2019), Forest Information Act (2011), GDPR |
| Forest biophysical data | Fragmented between IBAMA, ICMBio, SICAR, and the states | Open by law since 2011 (Metsätietolaki) |
| Forest inventory | IFN with a 20×20 km grid, manual in-person collection, long cycles, and punctual sample coverage; private inventories without integration to the public system | Continuous, national, via airborne LiDAR |
| Operational costs | Diluted in PDFs of various sector reports | Trackable via real-time “harvesters” data |
| Availability format | PDFs, infographics, promotional reports | Open APIs, structured data, Metsään.fi portal |
| Effort for analysis | Months of data engineering (OCR, regular expressions, heuristics) | “Download” direct + analysis |
| Owner sovereignty | Scattered data among agencies with little or no integration | Metsään.fi Portal with digital twin of the forest |
| Interoperability | Deficient among federal, state, and private organizations | Integrated via Suomi.fi and national standards |
| Specialized human capital | Scarce in the environmental sector for data processing | Integrated training within the innovation ecosystem (Sitra) |
Source: Elaborated by the author (2026)
The fundamental difference was not in the quantity of data produced — Brazil generated significant volumes — but in the governance architecture that made them usable. The Finnish legal mechanism of separating biophysical data from personal data created an ecosystem where information flows between owners, companies, and government with low friction, while Brazil kept its informational assets in institutional silos and formats that were difficult to reuse. This paradigm difference manifested with greater clarity in the function assigned to the forest inventory: Finland consolidated it as a pillar of industrial intelligence, while Brazil largely confined it to environmental management and enforcement. The result was an informational transaction cost that, in Brazil, made financial analyses unfeasible that would be trivial in the Finnish context.
In summary, the research demonstrated that informational fragmentation in the Brazilian forestry sector constitutes a quantifiable financial and operational risk. The automated extraction of 13,059 records from 332 documents demanded over 150 hours of technical work, but revealed a critical asymmetry: descriptive data were obtained in a reasonable quantity, while elementary financial variables presented filling rates below 12.1% for the Effective Operational Cost, and no calculable net margin. Sensitivity analysis indicated that uncertainty in input data can fluctuate the NPV of a forestry project by more than R$ 20,000/ha, compromising the return on investment assessment. The phenomenon of “data-driven decision distortion” was observed, where the lack of data quality forced the reorientation of financial viability analysis towards sentiment analysis. This analysis, in turn, confirmed the transition of sectoral reports from a statistical format to a predominantly promotional one, reducing their analytical utility. The comparative study with Finland highlighted that the central issue is not the volume of data, but the governance that makes it usable, with the Finnish model promoting a low-friction information ecosystem, in contrast to Brazilian institutional silos. Strengthening data governance, including standardization and API access, is a prerequisite for the Brazilian forestry sector’s transition to evidence-based planning.
4. Conclusion
This research investigated how data governance differences between Brazil and Finland impact analytical capacity, financial risk, and the efficiency of evidence-based forest management. It was found that informational fragmentation in the Brazilian forest sector constituted a significant barrier, requiring considerable effort to extract and structure 13,059 records from 332 documents. However, a critical asymmetry was observed in the findings: while descriptive data were obtained in a reasonable quantity, essential financial variables, such as Effective Operational Cost, showed fill rates below 12.1%, with no calculable net margin. Sensitivity analysis demonstrated that uncertainty in this input data could fluctuate the Net Present Value of a forestry project by more than R$ 20,000/ha, compromising the determination of return on investment. This scenario materialized the phenomenon of data-driven decision distortion, where the lack of quality in informational inputs forced the reorientation of financial feasibility analysis towards lexical sentiment analysis. This, in turn, confirmed the transition of Brazilian sectoral reports from a statistical format to a predominantly promotional one, reducing their analytical utility. In contrast, the Finnish model demonstrated the viability of integrated governance, with consolidated forest inventory as a pillar of industrial intelligence, facilitating data flow and reuse.
The study’s main contribution lies in the empirical demonstration that informational fragmentation in the Brazilian forestry sector is not merely a technical limitation, but a quantifiable financial and operational risk. Strengthening data governance, which includes standardizing publication formats, creating public APIs for sectoral data, clearly defining roles in the information custody chain, and investing in specialized human capital, is a prerequisite for the sector’s transition to an evidence-based planning system. Limitations acknowledged include that the sentiment analysis was based on a reduced sample of documents and that the extraction results reflected the sources and tools available during the study period. For future studies, it is suggested to replicate the research with expanded corpora, build standardized forestry data repositories, and evaluate the economic impact of adopting open data standards in the sector.
Bibliographic References
ASSOCIAÇÃO BRASILEIRA DE PRODUTORES DE FLORESTAS PLANTADAS (ABRAF). Anuários estatístico da ABRAF 2007 – 2013: ano base 2006. Brasília: ABRAF, 2007 – 2013.
BRASIL. [Decreto nº 12.069, de 21 de junho de 2024]. Dispõe sobre a Estratégia Nacional de Governo Digital. Brasília, DF: Presidência da República, [2024
INDÚSTRIA BRASILEIRA DE ÁRVORES (IBÁ). Relatórios Anuais 2014–2025. Brasília: Ibá, 2014-2025. Disponível em: https://www.iba.org. Acesso em: 06 fev. 2026.
PROVOST, F.; FAWCETT, T. Data Science for Business. Sebastopol, CA: O’Reilly Media, 2013.
TURBAN, E.; VOLONINO, L. Tecnologia da informação para gestão. 8. ed. Porto Alegre: Bookman, 2013.
Article originating from the Course Conclusion Work of the Specialization in Data Science and Analytics of the MBA USP/Esalq
To learn more about the course, click here and access the MBX Academy platform