Skip to content
Browse docs
Docs / Choose what to build

How Annsa classifies feedback

Four readings of every signal, in 16 languages, before anything is ranked.

01Four readings

Every piece of feedback is read for intent, the job it relates to, urgency and sentiment. One ranked list can then hold a bug report, a feature request and a compliment.

02The four intents

Bug: something is broken. Feature: an explicit request for something new. Improvement: friction in something that exists. Praise: something worked. Ambiguous messages lean Improvement, not Bug.

03Grouped by problem, not wording

"CSV export takes two minutes" and "I wait ages for the CSV download" are one problem in two voices, so they fold into one priority. "Export to PowerPoint" is a different request and stays apart. A message that raises two things becomes two signals; the same customer saying it twice is one voice.

0416 languages

English, Spanish, French, German, Portuguese, Japanese, Korean, Chinese, Italian, Dutch, Arabic, Hindi, Russian, Turkish, Polish and Swedish. They rank on one list.

05Correct it

Move a signal, rename a priority, correct an intent. The correction sticks and feeds Decisions and principles.

Your call: when the grouping is wrong, say so once. A finding your agent files isn't customer feedback; see Choosing what to build.

Next guideImporting feedback