How we grade what we publish
Intelligence is only useful if you can see where it came from and how far to trust it. Every record on this platform carries a source, a reliability grade, an analyst confidence level and a time horizon. Where the evidence is thin, we say so rather than smoothing it over.
Collection disciplines
Open-source collection
Humanitarian and conflict reporting, official statements, court and procurement records, aviation and maritime notices, national press in English, French, Arabic and Somali.
Structured data
Incident, displacement, hazard and food-security datasets, ingested on a 15-minute cycle and normalised to a single Africa geography model.
Practitioner network
Regional specialists across the Horn, Sahel, Great Lakes and coastal West Africa who verify ground detail our automation cannot.
The six collection layers
Our monitoring desk is organised into six layers. Each ingested record is tagged with the layer it came from, and every article we publish is written by our analysts in our own words with the originating publisher or dataset named in the text. The full register of what we collect from — Tier 1 through Tier 3, with reliability grades and collection cadence — is published on the source registry.
Conflict & armed groups
Organised-violence event data and armed-actor reporting, used to track incident tempo and group activity by district.
ACLED · UCDP GED · GDELT
Early warning & emergencies
Disaster, displacement and food-security warnings that precede operational disruption for aid programmes.
GDACS · ReliefWeb · FEWS NET · UN OCHA
GEOINT & situational awareness
Satellite and remote-sensing signals used to corroborate reported burning, flooding and infrastructure damage.
Copernicus · NASA FIRMS · USGS · NASA EONET
Cybersecurity
Threat indicators and advisories relevant to NGO networks, donor systems and field connectivity.
AlienVault OTX · AbuseIPDB · URLhaus · CISA
Strategic & political risk
Governance, sanctions and political-economy signals that shape access, partnerships and compliance exposure.
GDELT · ACLED · OpenSanctions
Verification
Geolocation and imagery checks applied before a contested claim is allowed to influence a rating.
Reverse image search · satellite imagery · open-source research
Source reliability (A–D)
We use a NATO-style admiralty scale so that a grade means the same thing on every record, whichever analyst wrote it.
| Grade | Meaning | Applied to |
|---|---|---|
| A | Completely reliable | Official record or a source with a long, verified history of accuracy on this subject. |
| B | Usually reliable | Established outlet, agency or field contact with minor historical inconsistencies. |
| C | Fairly reliable | Credible but limited track record, or reporting outside its usual area of coverage. |
| D | Not usually reliable | Partisan or single-channel reporting with no independent corroboration. Used only alongside stronger material, never alone. |
Analyst confidence
Multiple independent sources of A–B reliability, consistent with observed patterns.
Credible reporting with partial corroboration or minor evidentiary gaps.
Single-source or contested reporting. Published with the gap stated explicitly.
Threat levels (UK five-tier scale)
Our country and site threat ratings use the United Kingdom five-tier scale, so a rating reads the same way as the UK national system. The current UK national threat level from international terrorism is SEVERE.
| Threat level | Operational meaning | Posture we advise |
|---|---|---|
| CRITICAL | An attack is highly likely in the near future. | Movement restricted to essential tasks under specific authorisation; relocation and hibernation plans held at immediate readiness. |
| SEVERECurrent UK national status | An attack is highly likely. | Enhanced controls: authorised movement only, venue restrictions, journey management on all road moves, daily reporting. |
| SUBSTANTIAL | An attack is likely. | Standing controls enforced: briefed travellers, tracked movement, no-go areas, defined check-in schedule. |
| MODERATE | An attack is possible, but not likely. | Routine controls with periodic review; heightened attention around political events and known anniversaries. |
| LOW | An attack is highly unlikely. | Baseline duty-of-care measures; monitoring continues so a change is detected rather than discovered. |
Publication gates
Africa-only allowlist. Records outside our geographic and thematic scope are rejected before analysis.
Severity and country attribution are checked against at least one independent record before a rating can move.
Automated drafting is a starting point, never a publication. A named analyst signs the assessment, horizon and confidence.
Every source URL is validated for liveness. Dead links are archived out of the library, not left to rot.
What we do not claim
- We do not present automated summaries as analyst judgement. Machine-drafted text is labelled and reviewed before it carries an assessment.
- We do not model precise casualty figures or attribute responsibility where open sources cannot support it.
- Forecasts are probabilistic judgements with a stated horizon, not predictions of specific events.