Protective intelligence · source quality
Before a threat report becomes a briefing
A feed can contain thousands of reports without containing thousands of distinct events. The useful question is what each report can support, and what remains unknown.
Start with the claim, not the volume
A report is an item of information. An event is something that happened. Several reports may describe one event; one report may describe several. A feed count cannot be relabelled as an event count without a documented way to handle that difference.
This page explains how to evaluate that chain. It presents source-review questions, not a new finding about crime in Brazil or a tested result from a protective operation.
Find the original source
Ask who first reported the claim and what evidence they had. A first-hand account, an official record, a press summary and a repost are different kinds of source. Record the distinction before describing them as corroboration.
Two websites repeating the same report do not supply two independent confirmations. The number of pages is not the number of separate observations. A useful source note shows where each claim began and how later accounts relate to it.
Keep time and place attached
A new post can refer to an old event. Retain the event date separately from the date of publication and the date it entered a feed. When a date is uncertain, keep that uncertainty visible.
Location needs the same care. A city name added to a broad news item does not establish the street or neighbourhood of an incident. State totals, metropolitan totals and municipal totals answer different questions.
Make a count reproducible
- Define the counting unit: reports, distinct events, people or administrative records.
- Specify the start and end of the period and the geographic boundary.
- Describe how repeat coverage is identified and what happens when records cannot be matched.
- State which sources were included, what was missing and when the extract was saved.
- Keep the version used for publication so later additions do not silently change a historical claim.
Write the uncertainty into the briefing
A report can be worth investigating before it is strong enough to support a firm conclusion. Separate what a source states from what you infer and what still needs checking.
A small checked subset is not automatically representative of everything collected. Sources may cover some places, incidents or languages more thoroughly than others. A ratio between a checked subset and an incomplete feed cannot show the prevalence of threats in a city.
A short source-review card
| Field | What to record |
|---|---|
| Claim | The exact statement being assessed |
| Origin | Original source and accessible evidence |
| Time and place | Event date, publication date and geographic scope |
| Independence | Whether supporting accounts share an origin |
| Limits | Missing information and plausible alternative explanations |
| Version | Saved extract or document used for the conclusion |
Common questions
Does a large monitoring feed prove broad coverage?
No. Volume alone does not establish geographic coverage, independence or accuracy. Those properties require separate evidence about the sources and the processing.
Can I cite the old 5,291-to-49 comparison?
Do not use it as a finding about independently verified events or city risk. The original inference has been withdrawn; this page now explains the source-quality questions that need to be resolved.