Frontier Airlines is a Denver-based ultra-low-cost carrier founded in 1994, serving more than 100 destinations across the United States and select international markets. Its model emphasizes affordable fares and point-to-point routes, making customer experience and operational reliability essential to differentiation.
This project is a full, end-to-end analysis that gathers public customer reviews, standardizes the data, and presents the findings in an interactive report. The goal is to clarify how travelers experience Frontier and where improvements would create the most value.
Key questions include: Where are Frontier’s strengths and pain points across routes and aircraft? How do patterns compare with similar airlines? Which issues represent quick wins versus longer-term initiatives, and how should resources be prioritized?
- Scope: 3,000 Frontier Airlines reviews from 2015 to 2025, sourced from AirlineQuality.com (Skytrax), covering multiple airlines.
- Goal: Identify what drives Frontier’s customer satisfaction and turn those insights into actionable improvements, benchmarked against other ULCCs for context.
The dashboard can be accessed here.
For a comprehensive explanation of the ELT pipeline and its workflow, refer to this repo skytrax_reviews
The extraction layer automates the process of collecting, cleaning, and staging Skytrax review data for analysis. It integrates multiple technologies to ensure data reliability, scalability, and security.
Key Technologies:
- Python & Pandas – for web scraping, parsing, and initial data handling
- Apache Airflow – to orchestrate and schedule ETL tasks
- AWS S3 – for cloud-based data storage and versioning
- Docker – to containerize and deploy workflows consistently
- Snowflake – as the data warehouse for downstream transformation
Workflow Summary:
- Scraping: Iterates through the Skytrax airline index, retrieves paginated review pages, and extracts structured fields (ratings, comments, flight details, passenger metadata, and category-specific scores).
- Cleaning: Standardizes date formats, handles nulls, and enforces consistent schema and data types using Python scripts.
- Staging: Uploads cleaned data to AWS S3 with IAM-based access control, encryption, and versioning.
- Loading to Warehouse: Copies cleaned datasets from S3 to Snowflake using the
snowflake_copy_from_s3operator. - Task Orchestration: Airflow DAG manages dependencies in this order
scrape_skytrax_data → clean_data → upload_cleaned_data_to_s3 → snowflake_copy_from_s3.
This layer refines raw review data to ensure consistency and readiness for transformation.
Stack: Python 3.12.5, Pandas, NumPy, Matplotlib, Seaborn
Key Steps:
- Column standardization - snake_case, special-character cleanup
- Date formatting - ISO 8601 for submission and flight dates
- Text cleaning - verification flag extraction, nationality normalization
- Route parsing - origin, destination, and connections
- Aircraft standardization - unified Airbus/Boeing naming
- Rating conversion - numeric Int64 fields for analysis
Cleaned outputs feed directly into Snowflake for transformation.
This layer transforms cleaned data into structured models to support analytics and reporting.
Stack: dbt (Core), Snowflake, Airflow, GitHub Actions
Data model
- Dimension Tables
| Table | Purpose |
|---|---|
dim_customer |
Passenger information |
dim_aircraft |
Aircraft attributes |
dim_location |
Airport/city keys for origin, destination, transit |
dim_date |
Calendar table for submission & flight dates |
- Fact Table
fct_review_enrichedOne row per review per flight with quantitative metrics Incremental dbt jobs maintain data freshness and minimize warehouse cost
Data Quality Framework:
- Schema and relationship tests
- Custom business-logic assertions (e.g., ratings within 0–10)
- Freshness/completeness checks. CI/CD triggers on code pushes, PRs, weekly runs, and manual executions
- SQL (Snowflake): Data cleaning, preparation, and exploratory analysis
- Python (Matplotlib, Seaborn): Visualizations (e.g., heatmaps) and exploratory analysis
- Mode Analytics: SQL/Python integration and interactive dashboards
- GitHub: Version control and documentation
3.2.1. Data Extraction (Snowflake)
- Queried 121K+ Skytrax airline reviews stored in Snowflake
- Filtered 3K+ Frontier Airlines records
- Applied SQL joins, CTEs, and window functions to clean, normalize, and prepare structured datasets
3.2.2. Integration (Mode Analytics)
- Connected Snowflake directly to Mode Analytics for real-time querying
- Combined SQL and Python within Mode for exploratory data analysis (EDA)
3.2.3. Exploratory Data Analysis (EDA)
- Used SQL to identify key service metrics (seat comfort, cabin staff, food, Wi-Fi)
- Applied Python (Matplotlib, Seaborn) to generate trend visualizations, including heatmaps for satisfaction drivers
3.2.4. Visualization & Insights
- Built interactive dashboards in Mode Analytics to highlight customer pain points and route-based performance differences
- Designed visuals for comparative analysis across service categories
3.2.5 Version Control & Documentation (GitHub)
- Documented SQL scripts, Python notebooks, schema diagrams, and dashboard outputs in GitHub
- Used GitHub for version control, collaboration, and project presentation
- Frontier’s overall sentiment is relatively weak: the average rating is low, and the proportion of reviewers who would recommend the airline signals improvement regarding customer service.
- In Economy, satisfaction is most strongly tied to Food & Beverages and Cabin Staff Service, passengers consistently cite those as differentiators.
- In Non-Economy, Seat Comfort and Food & Beverages stand out: negative feedback in these areas correlates strongly with overall dissatisfaction.
- When compared with other ULCCs, Frontier shows particular underperformance in the recommendation rate.
- Prioritize service enhancements in Economy Class, especially around Entertainment and Seat Comfort, since this class represents the majority of reviews and customer base
- Reevaluate investment in First Class, as demand is low compared to other segments, suggesting opportunities to optimize costs or reallocate resources
- Low sample size limits the reliability of review accuracy
- Aircraft type shows minimal impact on customer satisfaction; focus should shift to other service dimensions
- Airlines in the same price range deliver higher customer satisfaction, giving them a competitive advantage, while many Frontier customers feel the value does not match the price paid
- Frontier should prioritize improving key service metrics to enhance perceived value and close this competitive gap
- Predictive & Early-Warning – Build models that forecast satisfaction/NPS by route and aircraft and flag routes likely to underperform.
- Text Intelligence – Analyze review text to extract themes and sentiment (with multilingual support) to explain the “why” behind scores.
- Expand Data Sources – Integrate DOT on-time and complaint data, fare and ancillary fee data, airport/aircraft metadata, weather, and holiday calendars to provide richer context.