dfaudit 0.1.0

Audit your pandas DataFrame before you trust it — missing values, cardinality and top categories in one call.

dfaudit

Know if you can trust your data before you use it.

A fast, modern data audit tool for pandas DataFrames -- built for analysts who need to assess data reliability, not just explore it.

---

What is dfaudit?

dfaudit is a lightweight Python package for auditing pandas DataFrames. Before you clean, model, or report -- you need to know whether your data is reliable. dfaudit gives you that answer fast.

Think of it as an auditor's first pass: one call to understand how complete, consistent, and trustworthy your data actually is. Missing values, cardinality, dominant categories -- the things that tell you if the data holds up.

Install

pip install dfaudit

Quickstart

import pandas as pd
import dfaudit as dfa

df = pd.read_csv("https://raw.githubusercontent.com/pandas-dev/pandas/master/doc/data/titanic.csv")

dfa.overview(df)
dfa.missing_matrix(df)

What you get

overview(df) - Column-level data quality table

A styled summary table covering every column at once: dtypes, missing counts and percentages, unique value counts, mode, and top-3 categories - all color coded so problems jump out instantly.

missing — standard NaN / None / NaT values pandas considers null.

soft_missing — values that are technically non-null but meaningless in practice: empty strings, placeholder text, and numeric sentinel codes commonly used in legacy systems or exports. dfaudit flags these automatically:

• Strings: "", "n/a", "na", "null", "none", "nan", "nil", "-", "--", "?", "unknown", "missing", "not available", "not applicable"
• Numbers: -999, -9999, 9999, 99999

Missing percentage cells are highlighted red for high missingness, blue for high cardinality.

missing_matrix(df) - Row-level missingness visualization

A matrix plot that shows where your data is missing across every row, with a bar chart of missing percentages per column on the right. Spot correlated missingness, systematic gaps, and data loading issues that column-level stats miss entirely.

dfa.missing_matrix(df, style="vivid")

Why I built this

I work with a lot of different data on a daily basis. For years I carried the same audit snippets from notebook to notebook -- copy, paste, tweak, repeat. At some point I decided to stop doing that and package the most common data audit tasks into something reusable, with visuals I actually want to look at.

dfaudit is what I reach for now at the start of every new dataset. I hope it saves you the same time it saves me.

If you work with modern data and want to contribute -- please do, pull requests are very welcome.

---

Made by Ivan Kumeyko  ·  MIT License