For NGOs & Civil Society
Data in service of communities, not about them.
Gorebet's incident data and collective-efficacy proxies are designed for civil-society use: community program evaluation, rights documentation, public-health research, and accountability reporting — with honest disclosure of what the data cannot show.
What NGO partners can access
Civil-society organizations can request access to aggregated neighborhood-level data for the geographies relevant to their programs. Data is provided at the woreda level or above, with full methodology documentation, coverage indicators, and aggregation threshold disclosures.
We specifically support organizations working on gender-based violence, child welfare, infrastructure access, and displacement. These categories have additional privacy protections: they are not included in general data shares and require a separate data-access agreement with enhanced governance conditions.
- Woreda-level incident aggregates with coverage indicators
- Collective-efficacy proxy time series for program evaluation
- Sensitive-category data under enhanced governance agreements
- Report-desert flagging: low-coverage areas are labeled, not excluded
- Annual fairness audit results (available to data partners)
- Methodology documentation in English and Amharic
What we disclose about data limitations
Gorebet data reflects what community members chose to report through the app. It is not a census of incidents that occurred. Reporting rates vary by neighborhood, language, connectivity, and community trust in the platform. We label these gaps explicitly.
We do not claim our safety scores reflect ground-truth safety conditions. We claim they reflect community reporting behavior, which is itself a meaningful social signal — when interpreted with the methodology in hand.
Civil-society partners receive a data limitations brief as part of every data-access agreement, and we request that publications using Gorebet data include a methodology disclosure statement we provide.
Fairness and bias disclosure
What we commit to
Annual fairness audits comparing score distributions by reporter language group, approximate gender where available, and woreda socioeconomic proxy. Results published within six months of reaching analysis scale.
What we have not validated
No causal study of Gorebet's impact on outcomes. No validation of our collective-efficacy proxy against the Sampson survey instrument. AI classification accuracy not yet independently audited.
Known gaps
Coverage is lowest in areas with limited smartphone penetration, Amharic-only speakers who prefer Afaan Oromo or Tigrinya, and communities with low institutional trust. These are labeled in all data exports.