1 minute readReference note
What makes reputation data reliable enough to use?
In brief
Reputation data is usable when its records, definitions, and transformations support the conclusion being drawn. A large dataset remains unreliable if entity matching, coverage, or classification is wrong.
A monitoring record needs the query, entity, source or platform, date, language, response mode, and captured result. Duplicate records, missing observations, retries, and changed collection conditions should not silently alter the denominator.
Analytical categories also need testing. A negative phrase about a competitor can be assigned to the brand; a historical allegation can be labeled as a current finding. Important classifications should be checked against examples, with uncertainty and corrections retained.
The output must distinguish observation from inference. A score summarizes a defined sample; it is not a complete measure of public belief or a causal estimate of revenue. Reliable data makes those boundaries inspectable.