Development record: October 29, 2026. Written retrospectively.
A drop in recorded frustration is hard to judge without knowing how much conversation was assessed. LiteLedger’s Customer frustration report shows the counts and that coverage together.
The report covers every retained human message in its visible scope, with counts for frustration, confusion, success and delight from the existing analysis pipeline. Several phrases expressing frustration in one message count once for frustration. A message can carry other signals too.
The three charts show signal rates, the selected signal’s message count, and assessment coverage. Rates use assessed messages as their denominator; unassessed messages and invalid evidence remain visible. “We haven’t looked yet” is a poor synonym for “everyone is delighted.”
Charts group the currently selected assessments by message date. Later analysis can revise earlier periods. Labels are model interpretations. Their quoted evidence is checked against the source text. Counting those labels exactly does not directly measure how someone feels.
Open Reports → Customer frustration to review the trends and supporting evidence. The report reuses existing analysis, so exploring it doesn’t start another round of model calls.