Digital & Emerging Technologies

Research Methodology

Comprehensive analysis approach

Research Process & Approach

How we analyzed 12,785 reviews to derive actionable insights

Research Methodology

Comprehensive approach to extracting actionable insights from telecom customer feedback and regulatory complaints

Overview

Our comprehensive analysis employed a multi-phase approach to extract actionable insights from telecom customer feedback and regulatory complaints. We combined large-scale data extraction, AI-powered analysis, and statistical validation to understand the critical failure points driving customer frustration and CCTS escalations.

Data Collection

Phase 1: Review Extraction
  • Initial Extraction: 30,000+ app reviews collected across iOS App Store and Google Play Store
  • Providers: Rogers (MyRogers app) and Bell (MyBell app)
  • Platforms: Separate extraction for iOS and Android to identify platform-specific issues
  • Tools: Automated scraping tools with rate limiting to ensure data integrity
Phase 2: Data Selection and Cleaning
  • Deduplication: Removed duplicate reviews based on review ID and timestamp
  • Validation: Verified review authenticity and removed bot/spam content
  • Standardization: Normalized date formats, ratings, and platform identifiers
  • Final Dataset: 12,785 reviews selected for individual analysis
  • Selection Strategy: Analyzed all iOS reviews and a representative portion of Android reviews to ensure comprehensive insights while maintaining analysis quality
Phase 3: CCTS Data Integration
  • Source: Commission for Complaints for Telecom-television Services
  • Period: August 2024 - January 2025 (6 months)
  • Complaints: 15,913 formal complaints with categorization
  • Purpose: Validate app-related issues against regulatory escalations

Analysis Methodology

AI-Powered Individual Review Analysis

Key Innovation: Each of the 12,785 reviews was analyzed individually using Anthropic's Claude API rather than batch processing, ensuring nuanced understanding of customer frustrations.

Sentiment Classification

Positive/Negative/Neutral categorization with confidence scoring

Issue Categorization

Technical/Billing/UX/Features/Support classification

Severity Assessment

Critical/High/Medium/Low impact rating

Service Impact Prediction

Likelihood of requiring customer service escalation

Statistical Analysis

95%
Confidence intervals calculated
2
Platform comparison (iOS vs Android)
2
Provider comparison (Rogers vs Bell)
15+
Years of temporal analysis

Validation Process

  • Cross-Validation: Compared review insights with CCTS complaint categories
  • Sample Verification: Manual review of 500 randomly selected reviews
  • Statistical Significance: Ensured all reported differences exceeded margin of error
  • Edge Case Identification: Focused on the 0.06% experiencing breaking points

Key Methodological Decisions

1. Focus on Written Reviews

Rationale: Written reviews represent users at "breaking points"—moments of extreme frustration that predict support escalations and CCTS complaints. While only 0.06% of users write reviews¹, they reveal the exact failure modes that drive regulatory complaints.

2. Sentiment vs. Rating Analysis

Finding: Star ratings (2.64/5 average) tell a different story than app store displays (4.4/5). We prioritized sentiment analysis over ratings to understand the emotional drivers of complaints.

3. Journey-Based Categorization

Approach: Rather than simple topic classification, we mapped reviews to customer journey stages:

  • Authentication (Login/Password)
  • Transaction (Bill Payment/Plan Changes)
  • Information (Usage/Balance Checks)
  • Support (Help/Contact)

4. Platform-Specific Analysis

Discovery: iOS users showed 84.2% negative sentiment vs 58.1% for Android, revealing platform-specific architectural issues rather than general app problems.

Quality Assurance

Data Integrity

100%
Verification rate of selected reviews
100%
iOS reviews analyzed
95%
Confidence level with ±0.8% margin
5%
Manual spot-checking rate

Analysis Validation

  • AI Accuracy: Spot-checked 5% of categorizations manually
  • Consistency: Cross-validated findings across multiple analysis passes
  • Triangulation: Confirmed insights using review text, ratings, and CCTS data

Limitations and Considerations

1. Self-Selection Bias

  • Reviews represent frustrated users, not general population
  • Positive experiences likely underrepresented
  • Mitigation: Clearly distinguished "breaking point" insights from general user experience

2. Temporal Variations

  • Older reviews may reflect resolved issues
  • Recent reviews weighted more heavily in recommendations
  • App updates may have addressed some reported problems

3. Platform Constraints

  • App store reviews have character limits
  • Some technical details may be truncated
  • Supplemented with CCTS data for complete picture

Deliverables

1. Quantitative Analysis

  • Individual Analysis: 12,785 reviews individually analyzed and categorized
  • Statistical Breakdowns: By provider, platform, and category
  • Confidence Intervals: For all major metrics

2. Qualitative Insights

  • Journey Failure Points: Identification of critical breaking points
  • Edge Case Patterns: Recognition of recurring failure modes
  • Strategic Recommendations: Based on prevention opportunities

3. Strategic Framework

  • ROI Models: For edge case prevention initiatives
  • Implementation Roadmap: With realistic timelines
  • Competitive Differentiation: Strategies for market advantage
Data Collection AI Analysis Statistical Validation Strategic Insights Actionable Recommendations

This methodology ensures our findings are statistically robust, practically actionable, and strategically valuable for transforming Rogers' customer experience from reactive complaint management to proactive edge case mastery.

¹Industry estimate - requires verification with actual user base data

Back to Reports