1. What is counted

Fatal figures count people who died. Serious non-fatal figures count distinct events. The two units remain separate. Linked survivors in one attack count as one non-fatal event unless independent evidence establishes separate incidents. A survivor who later dies moves to the fatal register and is not counted twice.

2. Which date is used

• Use the incident or death date when supported.
• If the death date is unknown, a documented discovery or recovery date may be used as a provisional proxy.
• Article, arrest, charge, hearing and judgment dates do not determine the incident month by themselves.
• Undated allegations remain outside monthly totals.

3. Evidence and classification

Evidence labels distinguish reported allegation, official confirmation, corroborated account, provisional classification and adjudicated finding. Several outlets repeating the same police statement are one source lineage, not several independent confirmations.

A homicide involving a woman is not automatically classified as gender-related. The record must show an intimate-partner, sexual-violence, family-violence or other supported gender-related nexus. Where that nexus, causation or identity remains unresolved, the case stays in an uncounted lead table.

4. Gender fields and legal status

Victim and actor gender fields reproduce attributable reporting and are not inferred from names or images. Unknown gender remains visible. Arrested or accused people are not described as convicted perpetrators. Fatal-person and non-fatal-event denominators remain separate.

5. Privacy and safeguarding

Children and living survivors are not named. Exact ages, small localities, identifying images, home addresses and confidential medical details are withheld when they could identify a child or survivor. Public availability elsewhere does not require republication here.

6. Corrections and limits

Case IDs remain stable. Material changes record the review date, old and new claim, supporting source, reason, change type and effect on monthly or cumulative figures. Archived months remain open to stronger evidence.

This is an incomplete public-source register. It cannot estimate national incidence, prevalence or perpetrator distributions. Official aggregate datasets remain separate from individually indexed cases, and no single Africa-wide total is calculated from these registers.