Call Conversation Theme & Root-Cause Extractor

What it does

Reads a call/chat/QA corpus and compresses it into 5-9 manager-ready themes, extracting each theme's root cause and a prioritized corrective action. The logic: nobody can read 10,000 calls, but a theme map can be read. The method is anchored to named canon, not gut judgment:

• Braun & Clarke Reflexive Thematic Analysis (2006, 6 phases) — the gold standard for qualitative coding: familiarization → open coding → searching for themes → reviewing → naming → reporting. Ten customers may all say "price," but why they say it differs — the theme captures the meaning behind the symptom.
• Grounded Theory (Strauss-Corbin) — open/axial coding plus constant comparison; saturation is declared when the last ~10% of calls produce no new codes.
• COPC CX Standard — every QA finding maps to a Customer / Business / Compliance critical-error class; calibration checks rater consistency, and severity (info/warn/critical) derives from this.
• 5-Whys + Pareto (80/20) — the "why" chain is driven to the human/process/system/policy layer; themes are volume-ranked and the cumulative-80% vital few are flagged.
• Kano + Impact×Effort — every action is RICE-like prioritized with a must-be/performance/delighter label.

Output: a 0-100 Conversation Health Score (weighted FCR proxy + sentiment + repeat-call + critical-theme share), a theme table, Pareto, a 5-Whys tree, and a ranked action list. Industry benchmarks (SQM 2025: ~70% FCR average, ≥80% world-class) are cited explicitly.

When to use it

When you have a transcript/QA corpus and the question is "what do customers complain about most, why, and what do we do?" Typical triggers: high repeat-call rate, low FCR, complaint clustering, VoC root cause, QA failure patterns. Not for: capacity/FTE/Erlang-C planning or pure shift optimization. This skill analyzes content, not volume.

Method / frameworks

Braun & Clarke 6-phase RTA · Grounded Theory saturation · COPC 3 critical-error classes + calibration · 5-Whys (Toyota/Ohno) · Pareto 80/20 · Kano Model (1984) · KVKK Art. 4-6 (purpose limitation, data minimization, masking). Every theme passes a 5-criterion rubric: evidence density (≥3 calls), root-cause clarity (not a symptom), volume share, actionability, COPC severity. Themes below threshold fall to gaps as anecdotes — no theme is born from a single quote.