Vendor Data Quality & SLA Monitoring Framework

📌 Project Overview

This project demonstrates a lightweight, scalable framework for evaluating and monitoring third-party data vendors using standardized data quality metrics, SLA monitoring, and side-by-side vendor bake-offs.

The goal is to enable data-driven vendor decisions, improve downstream reliability, and maintain a clear audit trail for compliance and internal transparency.

🎯 Objectives

1. Define a clear, measurable definition of Data Quality
2. Compare multiple vendors using quantitative bake-off metrics
3. Monitor SLA performance and identify breaches early
4. Translate business requirements into technical evaluation logic
5. Provide transparent documentation for audits and stakeholder alignment

🗂️ Project Structure

To run the SQL notebook, please visit folder >models>Exploratory_data.ipynb

To view management report, please visit folder >output>management_report.md

To view output files from data analysis, please visit folder >output

vendor-data-quality/
├── README.md
├── data/
│   ├── vendor_a.csv
│   ├── vendor_b.csv
│   ├── united_states.csv
├── models/
│   ├── Exploratory_data.ipynb
├── docs/
│   ├── data_quality_definition.md
│   ├── vendor_bakeoff_methodology.md
│   └── data_lineage.md
├── output/
│   ├── SLA_breach_report.csv
│   ├── vendor_missing_value_flag.csv
│   ├── management_report.md

📊 Data Model (Mock Vendor Records)

Each vendor provides criminal record-like datasets with the following schema:

Column	Description
record_id	Unique record identifier
vendor	Data provider name
county	Jurisdiction
dob	Date of birth (PII)
ssn	Social Security Number (PII)
disposition	Case outcome
record_date	Source record timestamp
ingest_time	Time data was ingested

Mock data intentionally includes:

• Missing PII
• Delayed ingestion
• Inconsistent dispositions between vendors
• This simulates real-world vendor data variability.

🧮 Defining Data Quality

1. Data Quality Dimensions
   Data quality is defined across four weighted dimensions:

• PII Completeness
• Presence of DOB and SSN
• Disposition Accuracy
• Valid and interpretable case outcomes

2. Freshness (Latency)
   Time difference between record date and ingestion

3. Coverage
   Jurisdictional availability

Weighted Scoring Model

Data Quality Score =

35% PII Completeness + 30% Disposition Accuracy + 20% Freshness + 15% Coverage

Weights can be adjusted based on jurisdiction risk, compliance requirements, or downstream product sensitivity.