Does Group By keep the original rows?

No. Group By replaces the dataset with a grouped summary table. One row per unique group-by combination. Use Window Functions to keep original rows.

Can I group by more than one column?

Yes. Select multiple columns in the group-by picker. The output has one row per unique combination of values across all selected columns, not one row per column.

What is the difference between Count and Count Distinct?

Count counts all non-null rows. Count Distinct counts unique non-null values. 98 orders from 45 customers gives Count 98, Count Distinct 45.

What is the difference between sample and population variance?

Sample (ddof=1) divides by N-1 and is correct when data is a sample from a larger population. Population (ddof=0) divides by N and is correct only for complete populations.

Can I apply two aggregations to the same column?

Yes. Click Add Column, select the same column with a different aggregation, and give it a unique result column name.

Group By | Edilitics Docs

Group rows by one or more columns and apply aggregations like sum, count, mean, median, min, max, mode, first, last, variance, and standard deviation.

Group By collapses your dataset into one row per unique value (or unique combination of values, when grouping by more than one column) and computes aggregations across the remaining columns. The output is a summary table, not the original rows with new columns added. Every operation downstream works on the grouped result.

You can apply multiple aggregations to the same column (for example, both sum and mean on revenue) by adding extra aggregation rows in one operation.

Group By reshapes the entire dataset. All original rows are replaced by the grouped summary. If you need to keep original rows, add a Conditional Column or Column Aggregation instead.

When to Use Group By

Revenue by region. Group by region, sum revenue to get total sales per territory.
Revenue by region and tier. Group by region and product_tier together to get total sales per territory, broken down by subscription tier.
Order count by product tier. Group by product_tier, count order_id to see volume per segment.
Average deal size by sales rep. Group by sales_rep, take the mean of revenue.
Distinct customer count per region. Group by region, count distinct customer_name.
Spread analysis. Group by product_tier, compute std dev of revenue to understand deal size variability per tier.

Sample Dataset

The examples in this doc use the Edilitics sample orders dataset. Download it to follow along in your own workspace.

edilitics_sample_orders.csv

Sample orders dataset for hands-on practice · 500 rows

Download

Relevant columns for Group By examples:

Prop

Type

Aggregations by Column Type

The aggregation options shown depend on the type of the column being aggregated.

Numerical columns:

Aggregation	Label	What it computes
Sum	Sum	Total of all values in the group
Count	Count	Number of non-null rows in the group
Count Distinct	Count Dist	Number of unique non-null values
Minimum	Min	Lowest value in the group
Maximum	Max	Highest value in the group
Mean	Mean	Arithmetic average
Median	Median	Middle value when sorted
Mode	Mode	Most frequently occurring value
First	First	First value in the group's row order
Last	Last	Last value in the group's row order
Variance (sample)	Var (Sample)	Sample variance (ddof=1)
Variance (population)	Var (Pop)	Population variance (ddof=0)
Std Dev (sample)	Std Dev (Sample)	Sample standard deviation (ddof=1)
Std Dev (population)	Std Dev (Pop)	Population standard deviation (ddof=0)

Categorical (string) columns:

Aggregation	Label	What it computes
Count	Count	Number of non-null rows in the group
Count Distinct	Count Dist	Number of unique non-null values
Mode	Mode	Most frequently occurring value
First	First	First value in the group's row order
Last	Last	Last value in the group's row order

Min and Max are not offered for categorical columns: they would compute lexicographic (alphabetical) ordering on the text, which is meaningless or misleading for plain category values (alphabetically, "Blue" comes before "Green" before "Red", but that ordering carries no real-world meaning).

Datetime, date, and time columns:

Aggregation	Label	What it computes
Count	Count	Number of non-null rows in the group
Count Distinct	Count Dist	Number of unique non-null values
Minimum	Min	Earliest timestamp in the group
Maximum	Max	Latest timestamp in the group
First	First	First value in the group's row order
Last	Last	Last value in the group's row order

Mean, median, sum, variance, and standard deviation are not offered for datetime columns.

Use Sample variance/std dev when your dataset is a sample drawn from a larger population (the typical case for order data). Use Population when your dataset is the complete population.

How to Apply Group By

Open the Group By operation

In your Transform pipeline, click Add Operation and select Group By from the operation list.

If your dataset has columns with Object, List, or dictionary types, a warning appears. Flatten those columns first to avoid data loss.

Select the Group By column(s)

Choose one or more columns from the dropdown to group by. All column types are available: string, datetime, and numerical. Object and list columns are excluded.

In the sample dataset, try region alone to group orders by territory, or region and product_tier together to group by territory and subscription segment combined. Selecting more than one column produces one output row per unique combination of values across all of them, not one row per column.

Once at least one column is selected, three column headers appear: Aggregate Column, Aggregation, and Result Column Name.

Review and adjust aggregations

All remaining columns (every column not selected as a group-by key) appear automatically with default aggregations:

Numerical columns default to Sum
Categorical and datetime columns default to Count

The Result Column Name is auto-generated as sum_revenue, count_order_id, etc. Edit any name to something more meaningful.

Change the aggregation for any column using its Aggregation dropdown. Only aggregations valid for that column's type appear (see Aggregations by Column Type).

Remove a column from the output using its remove icon, if you do not need its aggregated value in the result.

Add extra aggregations (optional)

Click Add Column to add a blank aggregation row. Use this to apply a second aggregation to any column, for example mean_revenue alongside sum_revenue.

Each added row needs: an Aggregate Column, an Aggregation, and a unique Result Column Name.

The same aggregation cannot be applied to the same column twice. Already-used aggregations are hidden from the dropdown for that column.

Click Save & Preview

Click Save & Preview. Edilitics applies the group by and aggregations. The success toast confirms: "Group By applied. Preview now reflects grouped and aggregated output."

The preview now shows the grouped summary table, one row per unique group-by value or combination.

Verify in the preview

Check row count. It should equal the number of distinct values (or distinct combinations, for a multi-column group-by) in the group-by selection. Check aggregated values for sense. Nulls in aggregated columns indicate the group had no non-null values for that column.

Before and After

Group By columns: region, product_tier. Aggregations: sum of revenue, count of order_id.

Input (500 rows):

order_id	region	product_tier	revenue
ORD-2024-0001	North America	Enterprise	12450.00
ORD-2024-0047	Europe	Growth	8320.50
ORD-2024-0099	North America	Enterprise	83841.00
ORD-2024-0312	North America	Growth	3750.75
...	...	...	...

Output (one row per region + product_tier combination):

region	product_tier	sum_revenue	count_order_id
North America	Enterprise	421430.50	41
North America	Growth	156210.00	33
Europe	Growth	284210.75	44
Europe	Enterprise	198340.00	22
...	...	...	...

500 rows collapsed to one row per (region, product_tier) pair. All original columns except region and product_tier are replaced by the aggregated columns.

Code Equivalent

-- Compatible with PostgreSQL, BigQuery, Snowflake, Redshift, DuckDB
SELECT
  region,
  product_tier,
  SUM(revenue)          AS sum_revenue,
  COUNT(order_id)       AS count_order_id
FROM orders
GROUP BY region, product_tier;

import polars as pl

df = df.group_by(["region", "product_tier"]).agg([
    pl.col("revenue").sum().alias("sum_revenue"),
    pl.col("order_id").count().alias("count_order_id"),
])

After Save & Preview, the pipeline shows a DQ delta badge on this step - green if the table score improved, red if it dropped. See Data Quality Scoring for how scores are calculated.

Group By

When to Use Group By

Sample Dataset

Aggregations by Column Type

How to Apply Group By

Before and After

Code Equivalent

Operation Reference

Frequently Asked Questions

Next Steps

Filter

Sort / Order By

Window Functions

Pivot / Unpivot

On this page

Group By

Does Group By keep the original rows?

Can I group by more than one column?

What happens to null values in the group-by column?

What is the difference between Count and Count Distinct?

What is the difference between sample and population variance / std dev?

Can I apply two aggregations to the same column?

Why does my dataset have Object or List column warnings?

Why are Min and Max not available for text columns?

What does Mode return if there are ties?

Filter

Sort / Order By

Window Functions

Pivot / Unpivot

On this page