Search quality and zero rows

Use this page when a search returns unexpected matches, misses expected language, or exports rows with zero metric values.

False positives

A false positive is a matched sentence that contains a query term but does not represent the intended concept.

Common fixes:

• Add an AND NOT exclusion.
• Quote an exact phrase.
• Replace a broad term with a more specific phrase.
• Split a broad construct into separate searches.

False negatives

A false negative is relevant language that the query misses.

Common fixes:

• Add synonyms.
• Add phrase variants.
• Add plural forms manually.
• Use wildcards only when they do not overmatch.

Zero rows

CSV exports include rows for all earnings calls in the selected date range, even if no sentences satisfy the query.

Zero Exposure means the query did not match sentences in that earnings-call transcript. It does not mean the company had no real-world exposure to the topic.

Keep zero rows when merging searches or calculating aggregate statistics unless the analysis design explicitly requires filtering them out.

Mixed sentiment sentences

In the main Risk Tool counts, a sentence that contains both positive and negative sentiment words can contribute to both positive and negative. The net contribution to sentiment is zero because Sentiment is positive - negative.

Dataset-level sentiment filters use stricter logic. See Metric definitions.

Coverage and variation

A topic can have many matches in the broad corpus but little variation in the target sample. A topic can also be conceptually important but rare in earnings-call language.

Check coverage and variation before interpreting a search as evidence for a research question.