Data Preparation, Cleaning, Wrangling and Transformation
The preparation stage
Preparation is usually the largest part of any analytics project. Cleaning addresses quality problems: removing duplicates, filling or flagging missing values, correcting inconsistent formats (dates, capitalisation, units), removing or investigating outliers, and fixing spelling variants so that "Vic", "Victoria" and "VIC" are treated as one category. Wrangling reshapes data into an analysable structure: merging datasets, splitting or combining columns, pivoting between wide and long formats, and aggregating records. Transformation standardises values - converting units, scaling, encoding categories as numbers, or deriving new fields such as age from a date of birth.
Decisions and justification
Questions reward justified choices. Removing a record with a missing value may bias results if the missingness is systematic; imputing a mean is simple but reduces variability. An outlier may be a data-entry error (remove or correct it) or a genuine extreme case (retain it and note it). Always state the decision and its effect on the analysis.
Traps
Do not confuse cleaning with transformation. Changing "kg" to "g" is transformation; deleting a duplicate row is cleaning. Another trap is assuming all missing data should be deleted - listwise deletion can shrink a dataset and skew results. Also beware of silent errors: a spreadsheet may store a date as text, breaking sorting and calculation. Documenting each step supports reproducibility, so another analyst can repeat the process.
Rule of thumb
Follow the sequence: inspect, then clean, then transform, then verify. Keep an untouched copy of the raw data so any step can be reversed, and log every change so findings can be audited.