Model facts with provenance
Store value, source, date, confidence and validation. Separate the fact from its interpretation and the project's status.
Despegue / Data intelligence
From scattered information to useful business intelligence.
We designed and built a system researching investments and supplier needs, structuring evidence into browsable records and analysis.
Ignacio Aredez's own initiative
Live system · case documented as of October 2026
01 / The problem and its context
Despegue
Information about major investments is spread across announcements, decisions, companies and sector sources. A supplier business needs to understand where its activity fits, who buys and what to do next.
Who did what
Ignacio defined the opportunity, product and data model, and built the research, validation, publishing and analysis process. AI handles volume; direction and accountability for the result remain human.
A view of what was built
A data base turned into a product
The interface lets people explore supplier needs and open detailed records. Its architecture connects research, data, analysis and publication checks.
Explore this part of the product ↗
02 / Engineering decisions
From sources to analysis
Need: understand supplier opportunities
01Research and structure
- Identified sources
Announcements, decisions and documentation
- Data model
Value, source, date, confidence and validation
02Cross-check and correct
- Research and refutation
Independent paths by claim type
- Human review
What is published and its evidence
03Publish and reuse
- Data with provenance
Visible references and confidence
- Company analysis
Reusable base for assessing fit
↺ Refuted data is recorded and checked for recurrence.
System decisions and boundaries
Facts, interpretations and investment status are modelled separately. Validation and regression checks help preserve corrections.
Cross-check before publishing
Use independent research and refutation, rules by claim type and human review. The methodology states the verification paths.
Preserve corrections
Record refuted data and run checks to detect its reappearance.
Turn research into a product
Reuse a data and analysis base for the public site and reports on supplier businesses' fit.
Operations, automation and controls
How it works and is maintained
Data engineering
Value, source, date, confidence and validation status form the information unit. Facts, estimates and interpretation are distinguished so readers understand each conclusion's basis.
Research and refutation
Research and cross-checking are separated, with rules by claim type and human review. Official sources are required to verify the content of state acts.
Consistent publishing
Data and analysis feed pages and reusable outputs. Schema, reference, translation and derived-surface checks help detect divergence when the base changes.
Corrections that persist
Refuted data retains its reason and its reappearance is checked. Procedures and lessons guide further research and review of published work.
03 / How it is evaluated
How it is evaluated
Validation of schemas, references, confidence labels and refuted-data regressions. A confidence level expresses the evidence available; review does not turn an estimate into a fact or guarantee the absence of errors.
What was built
A public intelligence system with data linked to sources and a reusable base for company-specific analysis. You can inspect published information and follow its references.
Live system · case documented as of October 2026
Explore Despegue ↗What we can bring to your project