Shape / rows · All operations

Split rows into equal buckets

A variant of Number or rank rows

Assign each row to one of N buckets of equal size, in order.

Status

In arcform

Runs as a SQL step

arcform carries it as a sql: step today.

The DuckDB step

Split rows into N buckets of equal size, in order

It reads the orders table the proofs share. Create it first:

SQL
CREATE TABLE orders AS SELECT * FROM (VALUES
  (1,  'Alice Ng',      'north', DATE '2026-01-03', 120.00, 'widget', 'a|b',   '{"channel":"web","items":[1,2]}'),
  (2,  'alice ng',      'north', DATE '2026-01-05',  80.50, 'gadget', 'b',     '{"channel":"shop","items":[3]}'),
  (3,  'Bob  Okafor',   'south', DATE '2026-01-05', 300.00, 'widget', 'a|c',   '{"channel":"web","items":[4,5,6]}'),
  (4,  'Carla Díaz',    'east',  DATE '2026-01-09',  NULL,  'gizmo',  NULL,    '{"channel":"web","items":[]}'),
  (5,  'Bob Okafor',    'south', DATE '2026-01-12',  45.25, 'gadget', 'c',     '{"channel":"phone","items":[7]}'),
  (6,  'Dev Patel',     'west',  DATE '2026-01-12', 510.00, 'widget', 'a',     '{"channel":"web","items":[8,9]}'),
  (7,  'Carla Diaz',    'east',  DATE '2026-01-20',  99.99, 'gizmo',  'b|c',   '{"channel":"shop","items":[10]}'),
  (8,  'Dev Patel',     'west',  DATE '2026-02-02', 510.00, 'widget', 'a',     '{"channel":"web","items":[8,9]}'),
  (9,  'Eve Larsen',    NULL,    DATE '2026-02-03',  15.00, 'gadget', NULL,    '{"channel":"web","items":[11]}'),
  (10, 'Alice Ng',      'north', DATE '2026-02-10', 220.00, 'gizmo',  'a|b|c', '{"channel":"phone","items":[12,13]}')
) t(order_id, customer, region, order_date, amount, product, tags, meta);

Then run the statement:

SQL
CREATE TEMP TABLE bucketed AS
SELECT order_id, amount,
       ntile(5) OVER (ORDER BY amount NULLS LAST, order_id) AS fifth,
       ntile(4) OVER (ORDER BY amount NULLS LAST, order_id) AS quarter
FROM orders;
SELECT * FROM bucketed ORDER BY amount NULLS LAST, order_id;
SELECT CASE WHEN ((SELECT list(ids ORDER BY fifth) FROM (SELECT fifth, list(order_id ORDER BY amount NULLS LAST, order_id) AS ids FROM bucketed GROUP BY fifth)) = [[9, 5], [2, 7], [1, 10], [3, 6], [8, 4]] AND (SELECT list(n ORDER BY quarter) FROM (SELECT quarter, count(*) AS n FROM bucketed GROUP BY quarter)) = [3, 3, 2, 2]) THEN 'as expected' ELSE error('ten rows in five buckets were not five pairs in amount order, or in four buckets were not sized 3, 3, 2 and 2') END AS verdict;
  • It needs nothing installed beyond DuckDB itself.
  • The statement checks its own result and raises an error when it is wrong: “ten rows in five buckets were not five pairs in amount order, or in four buckets were not sized 3, 3, 2 and 2”.
  • It passed on DuckDB v2.0.0 (Cyanoptera) d54cbbe3df.

What other tools call it

Meridian's classification; an independent blind rating agreed on 95.7% of 324 rows (variant, version 3).

Apache DataFusion (1)
  • ntileread 2026-09-27

    Window: Bucket number from 1 to n dividing the partition as equally as possible

DuckDB (1)
  • ntileread 2026-09-27

    ntile aggregate

Duckle (1)
  • NTileread 2026-09-25

    NTILE bucket over a window

Ibis (1)
  • NTileread 2026-09-27

    Assign each row of an ordered window to one of N equal-sized buckets

RapidMiner Studio (1)
  • Generate Batchread 2026-09-27

    Add a batch-number column dividing rows into equally sized batches

Background

In SQL Server, NTILE distributes the rows of an ordered partition into a given number of numbered groups and returns the group each row belongs to.[1] Common quantiles have special names, such as quartiles (four groups), deciles (ten groups) and percentiles (100 groups).[2] In SQL Server, when the row count does not divide by the number of groups, group sizes differ by one and the larger groups come first.[1]

When to use it

In SQL Server, 53 rows split into five groups give three groups of 11 rows and two groups of 10 rows.[1] In the dplyr package for R, ntile() ignores ties: it creates evenly sized buckets even if the same value ends up in different buckets.[3]

In SQL Server, NTILE is nondeterministic.[1]

The method

In PostgreSQL and DuckDB, ntile returns an integer ranging from 1 to its argument, dividing the partition as equally as possible.[4][5] In SQL Server, NTILE with PARTITION BY first partitions the rows and then divides each partition into the given number of groups.[1] In statistics, q-quantiles are values that partition a finite set of values into q subsets of nearly equal sizes.[2] Quantiles are cut points, though the terms for them, such as quartiles, are more often used for the groups created than for the cut points.[2] Read an ntile bucket number as the group a row fell into, not as a boundary value, and compute the cut points separately if you need them. (Meridian's view)

Compared with the alternatives

In dplyr, ntile() is a sort of very rough rank, which breaks the input vector into n buckets.[3] Use a bucket when you want a fixed number of groups of about the same size, and a percentile rank when you want each row's own position as a fraction. (Meridian's view)

Sources

  1. [1]NTILE (Transact-SQL) - SQL Server | Microsoft Learndocumentation
  2. [2]Quantile - Wikipediasecondary source
  3. [3]R: Bucket a numeric vector into 'n' groupsdocumentation
  4. [4]PostgreSQL: Documentation: 18: 9.22. Window Functionsdocumentation
  5. [5]Window Functions – DuckDBdocumentation