How Time Distribution Size Checks Work

Definition

Asserts that the number of records in each time segment of a date field falls within a range.

Overview

The Time Distribution Size rule buckets the rows of a container by a date or timestamp field, one bucket per Interval (yearly, monthly, weekly, daily, or hourly), and counts the rows in each. Every bucket must hold between Min Count and Max Count records, inclusive.

Where a row-count rule watches the container as a whole, this one watches its shape over time. A table can hold exactly the number of rows you expect while a whole day is missing and the next one is doubled, and only a per-segment view makes that visible.

Typical use cases:

• Detect a day or hour where the load produced far fewer records than usual.
• Catch a segment that was ingested twice and now holds double the rows.
• Confirm a backfill produced a plausible volume for every period it covered.

Field Scope

Single: The rule evaluates one field per check: the date or timestamp used to place each row into a time segment. It accepts Timestamp and Date fields.

Accepted Types

Anomaly Types

Evaluation Flow

Every Time Distribution Size check follows the same four-step evaluation flow:

1. Apply the filter clause. If the check has a filter set, only rows matching the filter expression are considered. Filtered-out rows are not counted into any segment.
2. Group the rows into time segments. Each row is placed into a segment according to its value in the evaluated field and the chosen Interval. Rows with no date go into no segment.
3. Count the records in each segment. The platform counts how many rows landed in every segment the data produced.
4. Compare each count against the range. A segment passes when its count is between Min Count and Max Count, inclusive. Any segment outside that range produces a Shape Anomaly, which reports the failing segments together.

Only Segments the Data Produces Are Evaluated

Segments come from the values in the field. If no row carries a date in a given day, that day produces no segment, so it is not evaluated and cannot fail the count range.

This is the rule's most important limitation: it catches a day that arrived short, not a day that did not arrive at all. To detect an entirely absent period, pair it with a Freshness check, which watches how recent the newest record is.

Both Bounds Are Inclusive

A segment holding exactly Min Count records passes, and so does one holding exactly Max Count. The range is closed at both ends, so set the bounds to the counts you are willing to accept rather than to the first count you would reject.

The Interval Decides What the Check Can See

A daily interval on a table with a few rows per day produces noisy segments where normal variation looks like a failure. A yearly interval on the same table hides a missing week inside a total that still looks right.

Pick the interval at which the data has a rhythm you can state: the level where you could confidently say "a normal one of these holds between X and Y records".

The Filter Reshapes the Segments

The filter runs before the grouping, so filtered-out rows never reach a segment and the counts reflect only the rows that remain. Scoping a check to one region turns it into a per-region distribution check, which is often useful and occasionally surprising if the range was sized for the whole table.

Coverage Does Not Change the Outcome

The form shows a Coverage slider because it is part of the shared check form, but Time Distribution Size evaluates segment counts rather than individual rows. There is no per-row pass rate for a threshold to relax, so the setting does not affect whether the check passes.

The Filter Clause

The filter clause is a SQL WHERE expression applied before the evaluation. Filtered-out rows are ignored entirely.

When a filter is set, the Shape Anomaly message ends with [filter: <expression>] so the evaluated scope is visible in the alert.

See Also