Dotwave Creating calculated columns
Cleaning gets your data honest; feature engineering makes it useful. Step 4 of the Dotwave pipeline, Feature Engineering, is where you derive new columns from the ones you already have — turning raw fields into the metrics your analysis actually needs. A calculated column is any field computed from existing columns rather than imported: a profit figure, a month extracted from a date, a customer's spending tier. This article covers what you can build, how to build it by describing it in plain English, and where the result lands in your dataset.
What you can create
Calculated columns in Dotwave cover the transformations analysts reach for most often:
- Arithmetic — combine numeric columns, for example
profit = revenue − costor a ratio between two fields. - Date parts — pull the year, month, day, or quarter out of a date column so you can group and trend by period.
- Bins — group a continuous numeric column into labelled ranges, such as turning ages into "18–25", "26–40", and so on.
- Rolling averages and sums — compute a moving window over an ordered column to smooth noise or track running totals.
- Concatenating text — join text columns into one, for example building a full name or a composite key.
Each of these produces a genuine new column that sits alongside your existing data and travels with it through the rest of the pipeline and on into export.
How to create a calculated column
You do not need to write a formula language. Dotwave takes a plain-English description and generates the column definition for you, then shows it for confirmation before anything is added:
Move to the fourth stage of the pipeline once your data is cleaned and preprocessed.
This opens the input where you describe the column you want.
Write what you want in ordinary language — for example, "profit margin as profit divided by revenue". Name the columns as they appear in your dataset.
It interprets your description against your actual columns and produces the definition that will compute the new field.
Check that the definition matches your intent, then confirm to add it. Nothing is written to your dataset until you approve.
Division operations are checked for zero denominators. If your denominator column contains zeros, Dotwave will warn you.
The column appears in your dataset
Once confirmed, the calculated column is added to your working dataset and is immediately available to every subsequent step, chart, and export. The action is recorded in the audit trail as "Added calculated field 'X'", so anyone reviewing the dataset can see exactly which fields were derived rather than imported, and the column is written out with your data when you export. Because the definition lives in your recipe, the same calculated column is reproduced consistently if you re-run the pipeline on refreshed data.
Give calculated columns clear, self-explanatory names. A field called profit_margin tells the next person reading your export exactly what it holds, while a name like calc1 forces them back to the audit trail to find out.
Feature engineering is iterative — build a base metric, then build on top of it. You might derive profit first, then a profit_margin that references it, then a bin that groups margins into tiers. Each column you add is one more confirmed, audited step in your recipe, so the logic behind your final dataset stays transparent and repeatable.