False positive
A false positive in mention monitoring is a post that matched a brand term but does not actually refer to the brand. Generic names like Apple, Mango, Shell, and Coach generate thousands of false positives daily because the words also mean fruit, fruit, gas station, and sports professional. Sentia uses a single-pass classifier on every ingested mention in 2026 and holds the false-positive rate under 5% on the worst-case generic-name brands.
What it is
A false positive is a relevance error. The mention monitor matched the keyword, the keyword matched the brand's entity in the configured taxonomy, but the actual post is about something else. The opposite error is a false negative: a post that does refer to the brand but the monitor missed.
The classic false-positive case is a brand whose name is also a common noun. Apple Computer versus apple the fruit. Shell Oil versus shell the seashell. Coach the brand versus coach the verb.
Why it matters
A noisy mention feed is worse than a missing one. If 80% of the items in the alert list are false positives, the marketing team stops opening the alerts. The signal-to-noise ratio is the only thing that determines whether mention monitoring is a useful workflow or a daily chore.
False-positive rate also gates downstream metrics. EMV computed over a feed with 60% noise is a fiction. Sentiment analysis on a noisy feed reports the sentiment of the noise.
How Sentia uses it
Sentia runs a single Gemini Flash-Lite pass on every ingested mention to classify relevance, sentiment, virality, and crisis risk in one call. The relevance classifier holds false-positive rate under 5% on names like Apple, Mango, Shell, and Coach. The other 95% of suppressed mentions stay searchable for audit, but never trigger alerts.
Related concepts
Related Reading
Start with one monitor. Free.
Add a brand, paste a couple of competitor handles, and see your first calibrated readout in under five minutes.