| name | review-sentiment-shift |
| description | Weekly diff of review sentiment baseline for an Amazon ASIN — detects when 4-5★ themes regress or 1-3★ themes accelerate. Use when running `review-sentiment-shift {ASIN}` weekly via `/loop weekly`. Companion to `review-velocity-monitor` (which tracks count/rating); this tracks theme shifts. |
review-sentiment-shift — Theme-Level Sentiment Diff
Reviews shift in 3 ways: count, rating, and theme. review-velocity-monitor tracks count + rating. This skill tracks themes — what customers are loving/complaining about.
Prerequisites
Invocation
review-sentiment-shift {ASIN}
review-sentiment-shift {ASIN list}
Output
/tmp/cro-research/{ASIN}-sentiment-shift-{date}.md
Phase 1 — Pull Current
Calls review-mining {ASIN} (cached output ok if recent) → current themes.
Phase 2 — Compare to Baseline
State at /tmp/cro-research/.sentiment-state/{ASIN}.json — last quarter's themes + frequencies.
Compute:
- 4-5★ theme regression (theme dropped >20% frequency)
- 1-3★ theme acceleration (theme grew >20% frequency)
- New themes appearing
- Themes disappearing
Phase 3 — Alerts
🔴 Critical:
- A previously dominant 5★ theme dropped (e.g. "great quality" went from 40% → 18%) — quality regression signal
- A new 1-3★ theme appeared (something started failing)
🟡 Watch:
🟢 Stable: themes consistent
Phase 4 — Output
# Sentiment Shift — {ASIN}
**Period:** {prior baseline} → {now} | **Status:** {🟢/🟡/🔴}
## Theme Trajectory
### 5★ Drivers — Stable / Declining
| Theme | Prior % | Now % | Δ |
|-------|---------|-------|---|
| "great quality" | 42% | 18% | 🔴 -24% |
| "perfect size" | 30% | 32% | 🟢 stable |
| ... | ... | ... | ... |
### 1-3★ Objections — Stable / Accelerating
| Theme | Prior % | Now % | Δ |
|-------|---------|-------|---|
| "broke after week" | 5% | 22% | 🔴 +17% NEW PATTERN |
| ... | ... | ... | ... |
## What Changed
🔴 Quality regression: "great quality" dropped 24 points; "broke after week" emerged. Investigate manufacturing / supplier change.
## Recommended Actions
- Pull recent 1-3★ verbatim — `review-velocity-monitor {ASIN}`
- If pattern confirms: trigger `objection-killer` for the new objection
- If quality is real (not perception): escalate beyond CRO scope — manufacturer issue
## Save Snapshot
Updated baseline saved for next-week diff.
Reference Files
~/.claude/skills/cro/review-analysis.md
- Companion skill:
review-velocity-monitor (count/rating)
Quality Bar
Auto-Triggers
/loop weekly review-sentiment-shift {ASIN}
- After A/B test ends (did sentiment shift?)