Data Quality for Beginners: 7 Checks Every Analyst Should Automate – Nic | Data

Author

Nichodemus Amollo

Published

November 10, 2025

Why Data Quality Is Your Real Job

If your data is trash, your:

Models are misleading
Dashboards tell the wrong story
Policy recommendations can hurt people

Data quality is not “extra” work—it’s core to being a serious analyst.

7 Checks to Automate on Every Dataset

Missingness patterns by variable and group
Uniqueness of IDs
Range checks for numeric variables
Category consistency for factors/coded responses
Cross-field logic (e.g., age vs date of birth, pregnancy vs sex)
Duplicates and near-duplicates
Date/time sanity (ordering, impossible dates)

Automate these in:

R (with tidyverse/janitor)
Python (with pandas)

How to Turn This Into a Portfolio Project

Take any public health or development dataset
Build:
- A script that runs all 7 checks
- A short report or dashboard summarizing issues
Include:
- A “recommended data cleaning plan”
- Examples of how findings explain weird results

Great analysts are paranoid about data quality—and employers love that.

---
title: "Data Quality for Beginners: 7 Checks Every Analyst Should Automate"
subtitle: "Stop trusting dirty spreadsheets—learn how to catch problems before they hit your models"
author: "Nichodemus Amollo"
date: "2025-11-10"
categories: [Data Quality, Data Cleaning, Monitoring & Evaluation]
image: "/img/blog-analytics.svg"
---

## Why Data Quality Is Your Real Job

If your data is trash, your:

- Models are misleading
- Dashboards tell the wrong story
- Policy recommendations can hurt people

Data quality is not “extra” work—it’s core to being a serious analyst.

---

## 7 Checks to Automate on Every Dataset

1. **Missingness patterns** by variable and group  
2. **Uniqueness** of IDs  
3. **Range checks** for numeric variables  
4. **Category consistency** for factors/coded responses  
5. **Cross-field logic** (e.g., age vs date of birth, pregnancy vs sex)  
6. **Duplicates** and near-duplicates  
7. **Date/time sanity** (ordering, impossible dates)  

Automate these in:

- R (with tidyverse/janitor)
- Python (with pandas)

---

## How to Turn This Into a Portfolio Project

- Take any public health or development dataset
- Build:
  - A script that runs all 7 checks
  - A short report or dashboard summarizing issues
- Include:
  - A “recommended data cleaning plan”
  - Examples of how findings explain weird results

Great analysts are **paranoid about data quality**—and employers love that.