By Team Hypomnema · Data
Flag unusual numeric values using IQR fences, z-scores or median deviation.
Use the guide alongside the working tool.
Open Outlier Detector →Explicit working sample
id,region,product,amount,units,date
001,North,Notebook,24.5,2,2026-10-01
002,South,Pen,9,3,2026-10-02
003,North,Pen,15,5,2026-10-03
004,West,Notebook,,1,2026-10-04
005,South,Notebook,49,4,2026-10-05
006,North,Pen,1200,2,2026-10-06Flag unusual numeric values using IQR fences, z-scores or median deviation. Flags are review candidates, not proof of bad data. Numeric columns must contain numeric or missing cells. Z-scores use sample standard deviation. Modified z-scores use 0.67448975 × deviation / MAD; zero dispersion produces a warning and no score. No rows are deleted automatically.
Prepare the input
Flag unusual numeric values using IQR fences, z-scores or median deviation.
CSV, TSV and arrays of scalar JSON objects are supported. Identifiers stay text. Empty cells are missing by default; add special missing tokens only if their meaning is certain. Use the sample to inspect the exact workflow.
Review the method
Flags are review candidates, not proof of bad data. Numeric columns must contain numeric or missing cells. Z-scores use sample standard deviation. Modified z-scores use 0.67448975 × deviation / MAD; zero dispersion produces a warning and no score. No rows are deleted automatically.
Keep the useful output
Review the summary and result records. Transformation tools retain separate change, rejected-record or diagnostics datasets where relevant. Choose the dataset before exporting; a report and a cleaned dataset serve different purposes.
CSV exports protect formula-like text. JSON keeps data portable without making an Excel workbook. Reports and editable backups may contain private source values; download only what you intend to share. No automatic account storage or server processing is used.
Ready to apply this workflow?
Try Outlier Detector →Explore a related workflow
- Data Quality Profiler: methods and data limits
Inspect missingness, uniqueness, inferred types and numeric summaries.
- Missing Value Cleaner: methods and data limits
Fill or remove missing values with a complete, reviewable change log.
- Data Validation Rules: methods and data limits
Check records against clear rules and separate passing and failing rows.
- Fuzzy Record Matcher: methods and data limits
Compare approximate record names and review ranked match candidates.
Keep your original files and review the result. Found unclear instructions? Contact Team Hypomnema.