Case Study: Driving insights from a large Netflix dataset
Explore the live dashboard
Filter the Netflix catalog by region, genre, and popularity — then export your shortlist.
The Business Challenge (The Friction)
Code_Generated_Image (1).png
Streaming catalog evaluation and content acquisition require cross-referencing vast title libraries with external market telemetry (IMDb and TMDb). Prior to automation, catalog analytics were stalled by severe data hygiene bottlenecks across 7,786 title records and 40 fields:
- Data Corruption & Schema Violations: Spreadsheet exports routinely corrupted titles with time/date notations (e.g., coercing 7:19 into 07:19:00 and 22 July into 2026-07-22), injected invisible zero-width Unicode characters (00B), and violated First Normal Form (1NF) with unindexed, semicolon-delimited multi-value strings.
- Telemetry Gaps: Critical performance metrics suffered from high null rates—including a 34.3% void in movie runtimes and ~30% missing values in external ratings and popularity scores.
- Manual Bottleneck: Manually auditing, deduplicating, repairing, and calculating statistical yields across thousands of mixed-format titles consumed dozens of analyst hours per reporting cycle, delaying acquisition decisions and introducing human error into content valuation models.
The Solution Architecture (The Build)
The end-to-end data pipeline was structured across four automated, progressive modules orchestrated via AI and coded for deployment:
[Raw Catalog Ingestion (7,786 Rows)]
│
▼
[Module 1: Telemetry Health Check & Structural Audit]
│── Validates PK uniqueness (content_id)
│── Flags type mismatches, null distributions, & auto-format corruptions
▼
[Module 2: Normalization, Parsing & Hierarchical Imputation]
│── Strips zero-width Unicode & restores corrupted string titles
│── Splits mixed duration into runtime_minutes and season_count (Int64)
│── Executes grouped median imputation (Genre × Decade ➔ Fallback)
▼
[Module 3: Statistical Modeling & Outlier Interrogation]
│── Computes standardized Z-scores & Tukey’s Fences (Q3 + 3×IQR)
│── Calculates regional genre yield (aggregate score vs. volume)
▼
[Module 4: Executive BI & Interactive Dashboard Deployment]
│── Generates dual-axis visual yield benchmarks (Matplotlib/Plotly)
└── Cursor IDE build deployed to production portfolio (wzain.com).
Click here to view a Case Study showing the AI Agent build process.
Governance & Quality Controls (The Guardrails)
To prevent hallucinations, ensure auditability, and protect downstream reporting integrity, strict engineering controls were enforced:
- Full Data Lineage & Traceability: Added an automated Data_Imputed boolean audit flag initialized to False and systematically flipped to True across all 2,808 rows (36.06%) receiving statistical imputation. This ensures executive transparency between verified external metrics and modeled baselines.
- Hierarchical Imputation Guardrails: Missing continuous metrics (imdb_rating, tmdb_popularity, runtime_minutes) were imputed using a tiered fallback framework:
- Secondary Tier: Broad primary_genre catalog median.
- Global Fallback: Total catalog median.
- Strict Schema Typing: Cast identifier and count columns (tmdb_id, imdb_votes, runtime_minutes) from floating-point values (float64) to native nullable 64-bit integers (Int64), eliminating .0 artifacts. Dates were standardized to strict ISO-8601 (YYYY-MM-DD).
Measurable Business Impact (The ROI)
The pipeline transitioned catalog hygiene and market analysis from days of fragile manual data manipulation into an automated execution workflow:
ROI Calculations
The automated data pipeline transformed a raw, unstandardized catalog of 7,786 titles and 40 attributes into a 100% complete, analysis-ready dataset. By replacing manual spreadsheet remediation with automated schema validation, Unicode scrubbing, duration feature splitting, and hierarchical median imputation across 2,808 records, the workflow established full data integrity alongside an audit-ready Data_Imputed tracking flag. This clean foundation powered statistical modeling and regional slicing that uncovered vital commercial patterns—including a 10.1x engagement efficiency advantage for Asian anime over high-volume dramas and the power-law concentration of viral breakout hits like Outer Banks ().
Operationally, the pipeline slashed reporting turnaround from 44 manual analyst hours to under an hour per cycle, delivering a 98.2% efficiency gain and 43.2 net hours saved. Valued at an industry benchmark rate of $85 per hour, this generates $3,672 in direct labor cost avoidance per execution, scaling to an estimated $44,064 in annual savings on a monthly refresh cadence. Beyond immediate labor savings, the automated framework prevents data corruption errors and delivers rapid, audit-compliant market intelligence that de-risks content acquisition and licensing allocations.
Proven Insights
1. Top 5 US Movies vs. Top 5 US Series by Popularity & Share of Top 10
Shows the distribution of viewer demand across top-tier domestic content, highlighting the split between viral acquisition drivers and catalog retention anchors.
Key Finding: Outer Banks alone captures 28.9% of the entire top 10 cohort’s popularity volume (221.6), outpacing top-tier studio tentpoles like Avengers: Infinity War (12.4% share, 94.7 score).
2. Asia Catalog Analysis: Production Volume Share vs. Engagement Yield
A dual-axis comparison benchmarking catalog resource allocation (title volume) against per-title viewer engagement (mean popularity).
Key Finding: While Dramas occupy 31.1% of Asian production volume (505 titles), they yield only 1.3 mean popularity per title. Conversely, Anime Series accounts for just 7.7% of volume (126 titles) but delivers 13.2 mean popularity per asset—a 10.1x efficiency advantage.
3. Dominant Genre Popularity & Regional Market Share by Macro-Region
Highlights the single highest-performing primary genre in each geographic market, mapping aggregate popularity and regional concentration.
Key Finding: Content consumption patterns are heavily concentrated by geography: Crime TV Shows command 34.3% of all engagement in Latin America, Anime Series captures 31.5% in Asia, and Action & Adventure represents 16.0% in North America and 28.0% in Oceania.
Executive Summary
- Barbell Portfolio Dynamics Drive Engagement: Catalog value is split between extreme short-term viral spikes (Outer Banks with a popularity score of 221.6, ) that drive subscriber acquisition, and deep, multi-season syndicated cult classics (Cheers, Twin Peaks, Pee-wee’s Playhouse) that stabilize churn and maintain baseline viewing hours.
- Genre Efficiency Outweighs Catalog Volume: High content volume does not correlate linearly with engagement; in Asia, Anime Series generates 13.2 mean popularity per title across 126 titles, outperforming broad Dramas (505 titles, 1.3 mean popularity) by 10.1x in engagement efficiency per asset.
- Regional IP Specialization Dictates Content Strategy: Content consumption follows defined geographic preferences—Action & Superhero franchises dominate North America and Oceania, Serialized Crime/Telenovelas lead Latin America and MENA, and Animation/International Series drive Asia and Europe—demanding decentralized regional licensing rather than a uniform global catalog.
Screenshot 2026-08-30 112635.png
