Data Quality Guardrail

Python data-quality portfolio project for schema drift, freshness, null, duplicate, and range validation across operator-owned datasets.

Portfolio takeaway: "Data quality becomes operationally useful when failures are ranked, explained, and routed before they contaminate downstream decisions."

---

Project Overview

Attribute	Detail
Language	Python
Runtime Shape	FastAPI + CLI
Domain	Dataset health and validation workflows
Check Families	schema drift · freshness lag · null spikes · duplicate collisions · range violations
Output Modes	JSON API · terminal summary
Primary Users	analytics engineering · revops ops · data platform

---

Executive Summary

Data Quality Guardrail models the sort of service teams use when operational reporting is only as trustworthy as the pipelines feeding it. Instead of treating dataset checks as a passive notebook exercise, the project ingests a structured dataset contract, runs high-signal validations against fresh records, scores the severity of what it finds, and returns evidence-backed issues with next actions.

The repo is intentionally built as a Python service layer rather than a frontend artifact. It shows how dataset reliability can be treated as an operating system concern: validate the shape, inspect the drift, score the damage, and route action before bad data pollutes forecasting, attribution, customer intelligence, or executive briefings.

---

Validation Flow

dataset contract + records
        |
        v
typed validation request
        |
        +--> schema drift checks
        +--> freshness lag checks
        +--> null spike checks
        +--> duplicate collision checks
        +--> range violation checks
        |
        v
severity-scored quality report

---

Validation Families

Schema Drift

• unexpected columns
• missing required columns
• type expectations that no longer match the feed

Freshness Lag

• stale loads that undermine current-state reporting
• delayed ingestion windows on operational datasets

Null Spike

• missing critical identifiers or metrics
• sudden completeness regression

Duplicate Collision

• repeated primary keys or event identifiers
• inflated counts on downstream models

Range Violation

• values outside accepted floors or ceilings
• unrealistic conversions, revenue, or health signals

---

Usage

Create a Virtual Environment

python -m venv .venv
.venv\Scripts\activate
pip install -e .[dev]

Run the API

uvicorn app.main:app --reload

Open the Docs

http://127.0.0.1:8000/docs

Run the CLI Summary

data-quality-guardrail

Run the Tests

pytest

---

Sample Output

Data Quality Guardrail
======================
Dataset: revops_pipeline_snapshot
Rows analyzed: 12
Overall score: 89

[CRITICAL] freshness_lag (score 89)
Summary: Dataset freshness is materially outside the allowed reporting window.

---

Screenshots

Hero Capture

API Summary

Validation Breakdown

Proof Layer

---

Industry Applications

Revenue Operations

• stop stale pipeline snapshots from distorting forecast and coverage calls
• catch duplicate opportunity rows before they inflate board-facing numbers

Growth Analytics

• surface conversion-rate anomalies before attribution models drift
• flag null campaign fields before channel reporting is trusted too far

Customer Intelligence

• prevent broken health-score feeds from contaminating churn or lifecycle views
• validate freshness on intervention datasets before operators act on them

---

What This Demonstrates

• Python added meaningfully through a real validation service, not a token script
• Pydantic models and FastAPI used for operational data checks
• data quality modeled as a severity-ranked response problem
• CLI and API outputs shaped for real operator use
• evidence-backed reporting instead of vague “data looks off” summaries

---

Future Enhancements

• add historical comparison windows for trend-aware alerting
• support CSV upload and object-store ingestion paths
• export markdown incident summaries for data-quality review
• add rule packs for SaaS revenue, lifecycle, and experimentation datasets
• emit webhook-ready escalation payloads for orchestration systems

---

Tech Stack

Portfolio Links

• LinkedIn
• Kinetic Gain
• Skills Page
• GitHub

---

Part of mizcausevic-dev's GitHub portfolio, with a focus on backend systems, growth operations, data reliability, and operational decision tooling.