feat: Add Streamlit dashboard with Blueprint compliance (v2.1.0)
Dashboard Features: - 8 navigation sections: Overview, Outcomes, Poor CX, FCR, Churn, Agent, Call Explorer, Export - Beyond Brand Identity styling (colors #6D84E3, Outfit font) - RCA Sankey diagram (Driver → Outcome → Churn Risk flow) - Correlation heatmaps (driver co-occurrence, driver-outcome) - Outcome Deep Dive (root causes, correlation, duration analysis) - Export functionality (Excel, HTML, JSON) Blueprint Compliance: - FCR: 4 categories (Primera Llamada/Rellamada × Sin/Con Riesgo de Fuga) - Churn: Binary view (Sin Riesgo de Fuga / En Riesgo de Fuga) - Agent: Talento Para Replicar / Oportunidades de Mejora - Fixed FCR rate calculation (only FIRST_CALL counts as success) Technical: - Streamlit + Plotly for interactive visualizations - Light theme configuration (.streamlit/config.toml) - Fixed Plotly colorbar titlefont deprecation Documentation: - Updated PROJECT_CONTEXT.md, TODO.md, CHANGELOG.md - Added 4 new technical decisions (TD-014 to TD-017) - Created TROUBLESHOOTING.md with 10 common issues Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This commit is contained in:
sujucu70
366
docs/GAP_ANALYSIS.md
Normal file
366
docs/GAP_ANALYSIS.md
Normal file
@@ -0,0 +1,366 @@
|
||||
# GAP ANALYSIS: CXInsights vs BeyondCX Blueprints
|
||||
|
||||
> **Generated**: 2026-01-19
|
||||
> **Blueprints Analyzed**: 4 documents in `docs/blueprints/`
|
||||
|
||||
---
|
||||
|
||||
## Executive Summary
|
||||
|
||||
CXInsights currently implements a **subset** of the BeyondCX blueprint requirements. The project has strong foundations for RCA (Root Cause Analysis) but is missing several key frameworks defined in the blueprints.
|
||||
|
||||
| Framework | Coverage | Status |
|
||||
|-----------|----------|--------|
|
||||
| **Ventas (Sales Analysis)** | ~40% | Partial - conversion tracking, some agent failures |
|
||||
| **Close The Loop (CX)** | ~30% | Partial - CX drivers only |
|
||||
| **FCR (First Call Resolution)** | ~5% | Missing - no FCR detection |
|
||||
|
||||
---
|
||||
|
||||
## Blueprint 1: Contexto BeyondCX (General Framework)
|
||||
|
||||
### What it defines:
|
||||
- Role: Expert auditor for call analysis (energy sector - Endesa B2C)
|
||||
- Three strategic frameworks: Ventas, Close the Loop, FCR
|
||||
- Rules: Evidence-based, no invented info, professional analysis
|
||||
|
||||
### CXInsights Alignment:
|
||||
|
||||
| Requirement | CXInsights Status | Notes |
|
||||
|-------------|-------------------|-------|
|
||||
| Evidence-based analysis | ✅ Implemented | `evidence_spans[]` required for all RCA labels |
|
||||
| No invented info | ✅ Implemented | LLM must quote transcript |
|
||||
| Professional output | ✅ Implemented | Structured JSON with traceability |
|
||||
| Three frameworks | ⚠️ Partial | Only partial coverage of each |
|
||||
|
||||
---
|
||||
|
||||
## Blueprint 2: Análisis de Llamadas de Venta (Sales Framework)
|
||||
|
||||
### What it defines (5 KPIs):
|
||||
|
||||
#### KPI 1: Tasa de Conversión de Venta
|
||||
| Blueprint Requirement | CXInsights Implementation | Gap |
|
||||
|-----------------------|---------------------------|-----|
|
||||
| Determine Venta/No Venta | ✅ `CallOutcome.SALE_COMPLETED/SALE_LOST` | Covered |
|
||||
| Max 5 key factors for success | ⚠️ `lost_sales_drivers[]` | Different structure |
|
||||
| Max 5 factors for failure | ⚠️ `lost_sales_drivers[]` with RCALabel | Missing detailed actions |
|
||||
| Origin attribution (agent/client/company) | ❌ Missing | No origin field |
|
||||
| Specific corrective actions | ❌ Missing | Only `reasoning` field |
|
||||
|
||||
**Gap Details:**
|
||||
- CXInsights captures WHAT caused lost sale (driver_code)
|
||||
- Blueprint requires: WHO is responsible + HOW to fix it + EXAMPLE from call
|
||||
|
||||
#### KPI 2: Efectividad de Campañas
|
||||
| Blueprint Requirement | CXInsights Implementation | Gap |
|
||||
|-----------------------|---------------------------|-----|
|
||||
| Campaign effectiveness analysis | ❌ Not implemented | **MAJOR GAP** |
|
||||
| Product-message fit detection | ❌ Not implemented | |
|
||||
| Customer feedback capture | ❌ Not implemented | |
|
||||
| Segmentation analysis | ❌ Not implemented | |
|
||||
|
||||
**Gap Details:**
|
||||
- CXInsights has no concept of "campaign"
|
||||
- No way to track which campaign a call belongs to
|
||||
- No analysis of product-customer fit
|
||||
|
||||
#### KPI 3: Habilidades del Equipo de Ventas
|
||||
| Blueprint Requirement | CXInsights Implementation | Gap |
|
||||
|-----------------------|---------------------------|-----|
|
||||
| "Buen Comercial" / "Necesita Mejora" classification | ❌ Not implemented | **MAJOR GAP** |
|
||||
| Agent strengths identification | ⚠️ Partial via drivers | Implicit only |
|
||||
| Agent weaknesses documentation | ⚠️ `OBJECTION_NOT_HANDLED`, `POOR_PITCH` | Limited |
|
||||
| Coaching recommendations | ❌ Not implemented | |
|
||||
|
||||
**Gap Details:**
|
||||
- No agent skill scoring or classification
|
||||
- No explicit "good practices to replicate" output
|
||||
- No coaching action recommendations
|
||||
|
||||
#### KPI 4: Argumentarios y Objeciones
|
||||
| Blueprint Requirement | CXInsights Implementation | Gap |
|
||||
|-----------------------|---------------------------|-----|
|
||||
| Script quality evaluation | ❌ Not implemented | **GAP** |
|
||||
| Objection handling quality | ⚠️ `OBJECTION_NOT_HANDLED` driver | Binary only |
|
||||
| Improvement recommendations | ❌ Not implemented | |
|
||||
|
||||
**Gap Details:**
|
||||
- CXInsights only detects IF objection was handled poorly
|
||||
- Blueprint requires HOW to improve + specific examples
|
||||
|
||||
#### KPI 5: Ciclo de Venta
|
||||
| Blueprint Requirement | CXInsights Implementation | Gap |
|
||||
|-----------------------|---------------------------|-----|
|
||||
| Actions that lengthen sales cycle | ❌ Not implemented | **GAP** |
|
||||
| Friction points in call | ⚠️ Events (HOLD, SILENCE) | Partial |
|
||||
| Optimization recommendations | ❌ Not implemented | |
|
||||
|
||||
---
|
||||
|
||||
## Blueprint 3: Close The Loop (CX Framework)
|
||||
|
||||
### What it defines (5 Pillars):
|
||||
|
||||
#### Pilar 1: Mejorar Experiencia de Cliente (CX)
|
||||
| Blueprint Requirement | CXInsights Implementation | Gap |
|
||||
|-----------------------|---------------------------|-----|
|
||||
| "Buen CX" / "CX Mejorable" classification | ❌ Not implemented | No binary CX outcome |
|
||||
| Positive CX factors | ❌ Not implemented | Only negative drivers |
|
||||
| Friction points | ✅ `poor_cx_drivers[]` | Covered |
|
||||
| Customer feedback (explicit/implicit) | ❌ Not implemented | |
|
||||
|
||||
**Gap Details:**
|
||||
- CXInsights only captures PROBLEMS (poor CX drivers)
|
||||
- No capture of positive CX factors
|
||||
- No explicit CX quality rating
|
||||
|
||||
#### Pilar 2: Reducir Fuga de Clientes (Churn)
|
||||
| Blueprint Requirement | CXInsights Implementation | Gap |
|
||||
|-----------------------|---------------------------|-----|
|
||||
| "Sin riesgo de fuga" / "En riesgo de fuga" | ❌ Not implemented | **MAJOR GAP** |
|
||||
| Churn risk factors | ❌ Not implemented | |
|
||||
| Origin attribution (company/agent/client) | ❌ Not implemented | |
|
||||
| Prevention factors | ❌ Not implemented | |
|
||||
|
||||
**Gap Details:**
|
||||
- CXInsights has no churn risk classification
|
||||
- `CallOutcome.CANCELLATION_*` exists but no risk prediction
|
||||
- No churn drivers taxonomy
|
||||
|
||||
#### Pilar 3: Eficiencia Operativa
|
||||
| Blueprint Requirement | CXInsights Implementation | Gap |
|
||||
|-----------------------|---------------------------|-----|
|
||||
| Process inefficiencies | ⚠️ `COMPLEX_PROCESS`, `SYSTEM_ERROR` | Very partial |
|
||||
| Agent efficiency analysis | ❌ Not implemented | |
|
||||
| Optimal efficiency factors | ❌ Not implemented | |
|
||||
| Process improvement proposals | ❌ Not implemented | |
|
||||
|
||||
#### Pilar 4: Valor del Cliente (Customer Value)
|
||||
| Blueprint Requirement | CXInsights Implementation | Gap |
|
||||
|-----------------------|---------------------------|-----|
|
||||
| Explicit customer needs | ❌ Not implemented | **MAJOR GAP** |
|
||||
| Latent/implicit needs | ❌ Not implemented | |
|
||||
| Upsell/cross-sell opportunities | ❌ Not implemented | |
|
||||
| Customer value maximization | ❌ Not implemented | |
|
||||
|
||||
#### Pilar 5: Talento Interno
|
||||
| Blueprint Requirement | CXInsights Implementation | Gap |
|
||||
|-----------------------|---------------------------|-----|
|
||||
| Positive agent behaviors | ❌ Not implemented | |
|
||||
| Agent skills to replicate | ❌ Not implemented | |
|
||||
| Improvement areas | ⚠️ Agent-related drivers | Limited |
|
||||
| Coaching recommendations | ❌ Not implemented | |
|
||||
|
||||
---
|
||||
|
||||
## Blueprint 4: FCR (First Call Resolution)
|
||||
|
||||
### What it defines:
|
||||
|
||||
| Blueprint Requirement | CXInsights Implementation | Gap |
|
||||
|-----------------------|---------------------------|-----|
|
||||
| First call vs repeat call detection | ❌ Not implemented | **CRITICAL GAP** |
|
||||
| Factors causing repeat calls | ❌ Not implemented | |
|
||||
| Churn risk combined with FCR | ❌ Not implemented | |
|
||||
| 4 categories: Primera Llamada (sin/con riesgo) + Rellamada (sin/con riesgo) | ❌ Not implemented | |
|
||||
|
||||
**Gap Details:**
|
||||
- CXInsights has `CALLBACK_REQUIRED` driver but no FCR detection
|
||||
- No mechanism to identify if call is first contact or repeat
|
||||
- No churn-FCR cross-analysis
|
||||
|
||||
---
|
||||
|
||||
## Output Format Comparison
|
||||
|
||||
### Blueprint Required Output Structure:
|
||||
```
|
||||
Factor Clave 1: [etiqueta breve]
|
||||
- Descripción objetiva basada en transcripción
|
||||
- Por qué ayuda/impide [objetivo]
|
||||
- Qué tiene que hacer el agente para [corregir/replicar]
|
||||
- Ejemplo/fragmento de la transcripción
|
||||
|
||||
Separados por: "··· "
|
||||
```
|
||||
|
||||
### CXInsights Current Output Structure:
|
||||
```python
|
||||
RCALabel(
|
||||
driver_code="PRICE_TOO_HIGH",
|
||||
confidence=0.85,
|
||||
evidence_spans=[
|
||||
EvidenceSpan(
|
||||
text="Es muy caro para mí",
|
||||
start_time=45.2,
|
||||
end_time=47.8
|
||||
)
|
||||
],
|
||||
reasoning="Customer objected to price"
|
||||
)
|
||||
```
|
||||
|
||||
### Gap:
|
||||
| Blueprint Field | CXInsights Field | Gap |
|
||||
|-----------------|------------------|-----|
|
||||
| Etiqueta breve | `driver_code` | ✅ Equivalent |
|
||||
| Descripción detallada | `reasoning` | ⚠️ Too brief |
|
||||
| Acciones correctivas | ❌ Missing | **GAP** |
|
||||
| Ejemplo con fragmento | `evidence_spans[].text` | ✅ Covered |
|
||||
| Origin (agent/client/company) | ❌ Missing | **GAP** |
|
||||
| Max 5 factors | No limit | ⚠️ Different |
|
||||
|
||||
---
|
||||
|
||||
## Summary: Required Changes
|
||||
|
||||
### HIGH PRIORITY (Core Functionality Gaps)
|
||||
|
||||
1. **Add FCR Detection Module**
|
||||
- Detect first call vs repeat call
|
||||
- Track call reason/topic
|
||||
- Link to churn risk
|
||||
|
||||
2. **Add Churn Risk Classification**
|
||||
- Binary: "Sin riesgo" / "En riesgo de fuga"
|
||||
- Churn drivers taxonomy
|
||||
- Prevention factors
|
||||
|
||||
3. **Add Agent Skill Assessment**
|
||||
- Binary: "Buen Comercial" / "Necesita Mejora"
|
||||
- Skill dimensions scoring
|
||||
- Coaching recommendations
|
||||
|
||||
4. **Enhance RCALabel Structure**
|
||||
```python
|
||||
class RCALabel:
|
||||
driver_code: str
|
||||
confidence: float
|
||||
evidence_spans: list[EvidenceSpan]
|
||||
reasoning: str
|
||||
# NEW FIELDS:
|
||||
origin: Literal["agent", "customer", "company", "process"]
|
||||
corrective_action: str | None # What to do to fix
|
||||
replicable_practice: str | None # What to replicate
|
||||
```
|
||||
|
||||
### MEDIUM PRIORITY (Enhanced Analysis)
|
||||
|
||||
5. **Add Customer Value Analysis**
|
||||
- Explicit needs detection
|
||||
- Implicit/latent needs
|
||||
- Upsell opportunities
|
||||
|
||||
6. **Add Campaign Tracking**
|
||||
- Campaign ID field
|
||||
- Campaign effectiveness metrics
|
||||
- Product-customer fit analysis
|
||||
|
||||
7. **Add Positive Factors Output**
|
||||
- Not just problems (drivers) but also successes
|
||||
- "Buen CX" factors
|
||||
- "Effective sales" factors
|
||||
|
||||
### LOW PRIORITY (Refinements)
|
||||
|
||||
8. **Enhance Output Format**
|
||||
- Limit to max 5 factors per category
|
||||
- Add structured corrective actions
|
||||
- Match blueprint output structure
|
||||
|
||||
9. **Add Script Quality Analysis**
|
||||
- Argumentario quality scoring
|
||||
- Objection handling quality
|
||||
- Improvement suggestions
|
||||
|
||||
---
|
||||
|
||||
## Taxonomy Extension Required
|
||||
|
||||
### New Driver Categories Needed:
|
||||
|
||||
```yaml
|
||||
# CHURN RISK DRIVERS (New)
|
||||
churn_risk:
|
||||
PRICE_DISSATISFACTION:
|
||||
description: "Customer unhappy with pricing"
|
||||
SERVICE_DISSATISFACTION:
|
||||
description: "Customer unhappy with service"
|
||||
COMPETITOR_INTEREST:
|
||||
description: "Customer considering competitors"
|
||||
CONTRACT_ENDING:
|
||||
description: "Contract ending soon"
|
||||
REPEATED_ISSUES:
|
||||
description: "Customer has called multiple times for same issue"
|
||||
|
||||
# FCR DRIVERS (New)
|
||||
fcr_failure:
|
||||
INCOMPLETE_RESOLUTION:
|
||||
description: "Issue not fully resolved"
|
||||
MISSING_INFORMATION:
|
||||
description: "Agent didn't provide all needed info"
|
||||
PENDING_ACTION:
|
||||
description: "Action pending from company side"
|
||||
UNCLEAR_NEXT_STEPS:
|
||||
description: "Customer unclear on what happens next"
|
||||
|
||||
# POSITIVE CX FACTORS (New - inverse of poor_cx)
|
||||
good_cx:
|
||||
QUICK_RESOLUTION:
|
||||
description: "Issue resolved quickly"
|
||||
EMPATHETIC_SERVICE:
|
||||
description: "Agent showed empathy"
|
||||
PROACTIVE_HELP:
|
||||
description: "Agent anticipated needs"
|
||||
CLEAR_COMMUNICATION:
|
||||
description: "Agent communicated clearly"
|
||||
|
||||
# AGENT SKILLS (New)
|
||||
agent_skills:
|
||||
EFFECTIVE_CLOSING:
|
||||
description: "Agent closed sale effectively"
|
||||
GOOD_RAPPORT:
|
||||
description: "Agent built good rapport"
|
||||
OBJECTION_MASTERY:
|
||||
description: "Agent handled objections well"
|
||||
PRODUCT_KNOWLEDGE:
|
||||
description: "Agent demonstrated product knowledge"
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## Recommended Implementation Phases
|
||||
|
||||
### Phase 1: Core Gaps (Essential)
|
||||
1. Add `churn_risk` field to CallAnalysis
|
||||
2. Add `fcr_status` field (first_call / repeat_call)
|
||||
3. Add `origin` field to RCALabel
|
||||
4. Add `corrective_action` field to RCALabel
|
||||
|
||||
### Phase 2: Enhanced Analysis
|
||||
5. Add positive factors capture (good_cx, effective_sales)
|
||||
6. Add agent skill classification
|
||||
7. Add customer value analysis
|
||||
|
||||
### Phase 3: Campaign & Optimization
|
||||
8. Add campaign tracking
|
||||
9. Add sales cycle analysis
|
||||
10. Match exact blueprint output format
|
||||
|
||||
---
|
||||
|
||||
## Files Requiring Modification
|
||||
|
||||
| File | Changes Required |
|
||||
|------|------------------|
|
||||
| `config/rca_taxonomy.yaml` | Add churn, FCR, positive factors |
|
||||
| `src/models/call_analysis.py` | Add churn_risk, fcr_status, origin fields |
|
||||
| `src/inference/prompts.py` | Update prompt for new fields |
|
||||
| `src/inference/analyzer.py` | Parse new output structure |
|
||||
| `src/aggregation/` | Aggregate new metrics |
|
||||
| `src/exports/` | Include new fields in exports |
|
||||
|
||||
---
|
||||
|
||||
**Última actualización**: 2026-01-19
|
||||
Reference in New Issue
Block a user