CoverageReportsCalendarReproducibilityContactVerlumia Intelligence ES EN FR
verlumia Support
Doctrine · principles & method · doesn't change monthly

Governance · reproducibility

Trust is proven

We don't ask you to take our word. Every figure we publish is accompanied by everything needed to verify it: the sealed output, the exact methodology it was computed with, the official input and the hashes that tie it all together. Download it and verify it yourself.

The proof

What you can check

Two things, with nothing more than what your system already includes:

  1. Fixity. That the published datum is exactly what we sealed, byte for byte — comparing the file's SHA-256 hash against the one recorded in its own manifest.
  2. Reproducing the headline. That the number comes from the official datum via a fixed, public formula. For fraud, the 75.8 headline can be reproduced with a calculator from the SESNSP figure.

The headline is not a model. July 2026's 75.8 is the index of the official SESNSP datum (10,995 case files) transformed by the frozen scale: (10995 − 5406.58) / (12777.65 − 5406.58) × 100 = 75.8. The nowcast (an estimate that uses Google Trends) lives in the annex, never in the headline.

Download by edition

Reproduction packages

Fraud · July 2026 edition · Mexico grade A · measured
Sealed output + hash manifest + exact ICC/1.3 methodology + official SESNSP input + README with step-by-step verification. Headline 75.8 "High".
Download package (.zip) →
Retroactive fraud series · 2015–2026 grade A · measured
The 139-month measured series that feeds the trend + its manifest + exact ICC/1.4 methodology. Same seal (4651ab20…) cited by the public map.
Download package (.zip) →
Cyberbullying · 2021–2024 grade B · survey
Sealed series + pipeline record + ICC/1.4 methodology + INEGI's official anchors (MOCIBA). Headline 21.0% (2024), a survey estimate — no teal.
Download package (.zip) →

Each package ships a README (English README_EN.md and Spanish LEEME.md) with verification instructions and a SHA256SUMS.txt to check integrity in a single command.

How to verify

Four steps

  1. Download the edition's package and unzip it.
  2. Check integrity: sha256sum -c SHA256SUMS.txt (Linux/macOS) or Get-FileHash (Windows, PowerShell).
  3. Compare the output and methodology hashes against those declared in the edition's manifest.json.
  4. Follow the README to re-obtain the headline from the official input.
Scope

What it includes today — and what's next

These are verification packages: they let you check the fixity of the datum and reproduce the headline with the methodology and the official input. Re-running the full end-to-end computation with the pipeline code is a later publication decision by the Institute, and will be announced when ready. Nothing here depends on that step: everything is verifiable today.

Sources and licenses. SESNSP (crime incidence, fraud) under "Libre Uso MX"; INEGI/MOCIBA (cyberbullying) under INEGI's «Términos de Libre Uso de la Información» (Free Use of Information terms) — both permit redistribution with attribution. Google Trends (the nowcast signal) is Google's data: it is not redistributed, only its integrity hash is recorded.

No generative AI takes part in the published data: it is a construction and audit tool, never the source of the data. Teal signals measured truth only (grade A). We always distinguish a measured fact from an estimate. Governed by the five guiding principles + an ethos of non-fabrication and auditability. Questions about methodology or auditing? Write to metodologia@verlumia.org.