All cleaning operations

Dotwave can apply a wide range of transformations across the four pipeline steps, and every one of them is recorded in the audit trail under a plain-English name. This article is the complete reference to those operations, grouped by the kind of problem they solve. Each entry uses the same user-facing name you will see in the audit trail, explains what the operation does, and notes which columns it applies to. Use it to understand a step before you run it, or to decode an audit trail after the fact.

Missing values

Fill missing with mean

Replaces every null in a column with the column's average. Applies to numeric columns, and is most appropriate when the values are roughly symmetric, since the mean is sensitive to extreme values.

Fill missing with median

Replaces every null with the column's middle value. Applies to numeric columns and is the safest default for skewed data, because the median is not dragged around by outliers. This is the fill the AI recommends most often.

Fill missing with mode

Replaces every null with the most frequently occurring value. Applies to categorical columns, where an average has no meaning, and to any column with a clear dominant value.

Fill missing with a custom value

Replaces every null with a fixed value you supply, such as Unknown, N/A, or 0. Applies to any column, and is the right choice when a blank has a specific business meaning you want to make explicit.

Drop rows with nulls

Removes every row that is missing a value in the selected column, keeping only complete records. Applies to any column, and is best when a missing value makes the whole row unusable.

Drop a column

Removes an entire column from the dataset. Applies to any column, and is typically used on fields that are mostly empty or irrelevant to the analysis.

Duplicates and text

Remove duplicate rows

Deletes rows that are exact whole-row matches of another row, keeping the first occurrence. Applies to the dataset as a whole rather than a single column.

Fuzzy-deduplicate near matches

Finds and merges rows that are near-matches rather than exact copies — records differing only by capitalization, whitespace, or minor spelling. Applies across rows and is run from the guided cleaning dialog.

Convert text to real dates

Parses a column of date-like text into a proper datetime type so it can be sorted, filtered, and charted as a date. Applies to text columns that hold dates in a recognizable format.

Trim stray whitespace

Removes leading and trailing spaces from text values, so that London and London stop being treated as different entries. Applies to text and categorical columns.

Merge look-alike categories

Combines category labels that clearly refer to the same thing — such as USA, U.S.A., and United States — into a single canonical value. Applies to categorical columns, and is especially useful when the profile flagged high cardinality.

Encoding

Label encode

Converts each distinct category into an integer code, replacing text labels with numbers. Applies to categorical columns and is used to prepare ordinal or low-cardinality fields for analysis that expects numeric input.

One-hot encode

Expands a categorical column into a set of new columns, one per category, each holding a 0 or 1 flag. Applies to categorical columns with a manageable number of distinct values; avoid it on high-cardinality columns, where it produces an unwieldy number of new fields.

Outliers

Cap outliers

Limits extreme values to a sensible boundary rather than removing them, pulling values beyond the interquartile range back to the cap. Applies to numeric columns and preserves the affected rows while reducing the influence of extremes.

Drop outlier rows

Removes rows whose value in a numeric column falls far outside the normal range. Applies to numeric columns, and is the choice when an outlier represents a bad record rather than a value worth keeping.

New columns

Combine two columns (arithmetic)

Creates a new numeric column by applying an arithmetic operation — add, subtract, multiply, or divide — to two existing numeric columns, for example turning unit price and quantity into a line total. Applies to pairs of numeric columns.

Concatenate text columns

Creates a new text column by joining the values of two or more text columns, such as combining first and last name into a full name. Applies to text and categorical columns.

Add a calculated field

Creates a new column from a formula or expression you define over existing columns, giving you derived metrics that were not in the source file. Applies across the columns your expression references.