What’s in a data quality score?

The Cleaner agent’s score starts at 100. Every issue it finds costs points by severity — high −10, medium −5, low −2 — plus a small deduction for the overall share of empty cells. Issues it looks for:

• Missing values — severity depends on how much of the column is empty.
• Duplicate rows — exact copies inflate every total and count.
• Extreme outliers — values beyond both the 3×IQR fence and 1.5× the 99th percentile, so naturally skewed metrics aren’t punished.
• Inconsistent labels — “Ahmedabad”, “AHMEDABAD” and “ahmedabad ” counted as three cities.

On the Quick-Commerce sample the raw file scores 79; one click of Auto-clean takes it to 100.

Scores in practice

• 90+ — analysis-ready.
• 75–89 — good, fix the flagged columns before reporting.
• Below 75 — clean first; conclusions may be wrong.

Hit Auto-clean and DataOps creates a cleaned copy (never overwriting your original) with a log of every change it made.