Skip to main content

Methodology

How we calculate scores — and why they are not accusations.

Risk signal — not an accusation of crime.

AppaltiWatch uses public, versioned rules on ANAC data. Every score is explainable, tender by tender.

The 0–100 score comes from automatic checks for patterns known in literature and practice (competition, deadlines, thresholds, repeats, etc.). It is not an accusation of crime.

How the score works

For each tender we check whether documented signals are present. Each signal has a declared weight in the ruleset; weights add up to a maximum of 100.

Transparent rules

Every signal has a citizen-friendly description, visible on the tender page and in the glossary.

No black box

We do not use opaque AI models — only declared, versioned checks.

Versioned ruleset

If rules change, the version is shown on each tender so you can compare over time.

From data to your next step

1

Data signals

Automatic checks on tenders, awards, and variations published by ANAC.

2

Score 0–100

Weighted sum of active signals. Typical threshold for highlighting: ≥ 35.

3

Verify on sources

ANAC links, glossary, and tools to dig deeper — never a final verdict.

Risk signal — not an accusation of crime. Always verify on official sources.

How to read the score

Colour and badge show signal intensity, not guilt:

0–34: not enough signals to highlight the tender.
35–49: signals worth considering — useful to review.
50–59: multiple signals together — medium verification priority.
60+: high concentration — deserves attention, always confirm on ANAC.

Types of signals

Some examples of checks we apply (full list in the glossary):

Tender with a single bidderRead in glossary →
Very short bid submission periodRead in glossary →
Amount near regulatory threshold (possible contract splitting)Read in glossary →
Repeated awards to the same supplier (same CA, 12 months)Read in glossary →
Supplier in a different region than the project (weak signal)Read in glossary →
Publish or award date near an election date (temporal pattern)Read in glossary →
Award amount exceeds tender estimateRead in glossary →
Contract variation recorded on ANACRead in glossary →
Multiple variations on the same contractRead in glossary →
Work suspension recordedRead in glossary →
Mismatch between participants and declared biddersRead in glossary →
Limits and false positives

The method helps you navigate at scale, but has important limits:

  • A single signal may have legitimate explanations (emergency, narrow market, urgency).
  • ANAC data may arrive late or with missing fields.
  • Maps and rollups depend on the quality of geographic codes and localities.
  • Signal ≠ crime: only competent authorities can establish irregularities.

Practical examples

Next step

Put what you have read into practice.

For usage statistics and analytics cookies, see the Disclaimer