Identify Missing Data, Duplicates, and Outliers

What You Will Learn

You will learn how to use Gemini to inspect a dataset for missing values, duplicate records, outliers, and suspicious data patterns, then convert findings into a data-quality issue log.

Required Dataset

Use the Atlas_Procurement_Dataset.

Use the tab:

Purchase_Orders

Create three new tabs:

Data_Quality_Practice

Data_Quality_Issue_Log

Data_Quality_Check

Before You Start

Copy 50–100 rows from Purchase_Orders into Data_Quality_Practice. Include fields such as PO Number, PO Date, Supplier Name, Spend Category, Quantity, Unit Price, Extended Spend, Approval Status, and Emergency Purchase Flag.

Step-by-Step Instructions

Step 1. Ask Gemini to inspect the dataset

Open Gemini and enter:

Act as a data-quality analyst.

Inspect the Data_Quality_Practice tab.

Identify:

1. Missing values
2. Duplicate PO numbers
3. Outliers in Quantity, Unit Price, and Extended Spend
4. Suspicious or inconsistent category values
5. Rows that need manual review

Return a data-quality issue log with issue type, field affected, row or record, why it matters, recommended fix, and priority.

Paste Gemini’s output into Data_Quality_Issue_Log.

Step 2. Add simple formula checks

In Data_Quality_Practice, add columns called:

Missing Data Flag

Duplicate PO Flag

Outlier Flag

Review Needed

For missing data, use a formula like:

=IF(OR(A2=””,B2=””,C2=””,F2=””,H2=””),”Missing Data”,””)

For duplicate PO number, use:

=IF(COUNTIF($A:$A,A2)>1,”Duplicate PO”,””)

For a simple spend outlier flag, use:

=IF(H2>AVERAGE($H:$H)+2*STDEV($H:$H),”High Spend Outlier”,””)

Adjust column references to match your sheet.

Step 3. Ask Gemini to review the flags

Use this prompt:

Review the Missing Data Flag, Duplicate PO Flag, Outlier Flag, and Review Needed columns. Identify which rows should be reviewed first and explain why.

Expected Result

You should have a data-quality issue log that identifies missing fields, duplicate records, outliers, inconsistent categories, and high-priority rows for review.

Check Your Work

Confirm that missing values are flagged, duplicate PO numbers are identified, outliers are treated as review items rather than automatic errors, and every issue has a recommended fix.

Optional Extension

Ask Gemini:

Summarize the top data-quality risks in Data_Quality_Issue_Log and recommend the first five cleanup actions before analysis begins.