# Methodology

## Unit of analysis

The main unit is a registry decision row, not a unique defendant and not a criminal proceeding. A single proceeding may generate multiple registry entries; a single entry may contain several defendants.

## Main corpus

The main corpus is copied to `data/cases.csv` and `data/cases.jsonl`. It was generated from `raw/decisions-json/` using the local analytics scripts. The corpus contains decisions related to мобилизация, ТЦК, уклонение, препятствование деятельности ВСУ, ВЛК, взятки, подделки, доставка/недоставка в ТЦК and adjacent contexts.

## “Busification” proxy search

“Busification” is not a legal article. The proxy layer uses search formulas combining ТЦК/РТЦК/territorial recruitment center terms with delivery, detention, force and proxy articles including 146, 364, 365, 367, 371, 373, 374 and 426-1 КК. Automated matches are not publication-grade without manual review.

Publication-grade manual counts are in `data/busification/manual_busification_findings.csv`.

## Official Ombudsman layer

The Ombudsman layer is not treated as a rumor source. It is an official position of the Ukrainian human-rights institution. The package records official metrics separately from court verdict metrics:

- 18 complaints in 2022 and 6127 in 2025.
- 34 ERDR entries after Ombudsman appeals concerning actions of TCK/SP officials with signs of criminal offenses.
- 1 public court case described by the Ombudsman as a conviction of a TCK worker for obstructing Ombudsman activity.

These official metrics do not mean every complaint was proven in court. They show the state human-rights institution's documented escalation and formal legal response.

## Public video archive layer

The public video archive layer records busification.org as an open-source documentation project, not as a court or official state source. On the access date, its statistics page listed 7373 total videos, 3910 videos in the last year and city-level counts. This can support a claim about the scale of public visual documentation of mobilization conflicts.

Individual videos must not be described as proven crimes without case-by-case verification. For publication, verify original upload source, date, location, editing, visible actors, available official responses and whether private persons need anonymization.

## Access blocking layer

The package includes a high-risk lead layer about access restrictions around busification.org. Current collected sources point to an NCU order No. 631/3432 dated 25.08.2025 and media reporting that the restriction was made after an SBU request. This is relevant to freedom-of-information framing, but must be handled with primary documents and right-of-reply requests.

Safe publication framing at this stage: "According to dev.ua, UAblocklist and the identified NCU order number, busification.org was subject to an administrative access restriction by the National Center for Telecom Network Management. We have not yet completed a full legal search for court, sanctions or other public legal grounds."

Do not write that there was no court decision as an absolute fact until a full legal search has been completed.

## SUD.UA media/legal-press layer

SUD.UA is used as a research-lead layer. It is useful because many pages point to specific court case numbers, legal disputes, draft laws and official positions around TCK, mobilization, VLK, deferrals, registry errors and police/TCK powers. It is not treated as primary proof.

Before publication, every SUD.UA-led claim must be verified through one of the following: ЄДРСР decision text, Supreme Court text, Verkhovna Rada bill card and bill text, official ministry or Ombudsman source, police/TCK response, or direct source documents.

## Official OGP / DBR statistics layer

The OGP/DBR layer adds official national context:

- OGP Form N1 article-level pretrial statistics for 2025 and January-May 2026.
- OGP Form N2 person-level official workbooks for 2025 and January-May 2026.
- DBR annual reports and official statistical news for 2024 and Q1 2025.

The package mechanically extracts OGP rows for selected articles and stores sheet/Excel-row coordinates. These figures are official pretrial/investigation statistics, not final verdicts. DBR figures are agency-wide and must not be described as TCK-specific unless the source itself says so.

## Acquittal benchmark

Ukraine 2024 benchmark uses Supreme Court aggregate statistics: 61,434 convicted and 163 acquitted persons under final verdicts. The mobilization corpus benchmark uses only clear outcomes: convictions + acquittals.

## Caveats

- The main dataset is a collected corpus, not the full national court universe.
- 2026 is partial.
- The busification registry pass has 407 retry/missing IDs due to empty texts/timeouts.
- ERDR entries are not the same as final court convictions.
- Public video archives show open-source documentation volume, not established criminal liability.
- Access blocking claims require primary-order verification, legal-basis analysis and right of reply.
- SUD.UA materials are media/legal-press leads, not primary evidence.
- OGP/DBR statistics are official but not automatically mobilization-specific at article level.
- International acquittal comparisons are methodologically risky because countries count different procedural stages.