Markdown Copy

Ask AI

Check types

Every kind of check DataQ can author, generated from the check editor's catalog and the backend's vetted allowlist — so this page cannot drift from what the product actually offers. Every GX type on this page is executed in CI on a dataframe batch, and on a SQL batch too unless its row says it is dataframe-only.

	Count
Check types in the editor	49
GX expectation types vetted by the backend	25

How to read a row: Parameters are the editor's fields (mostly is GX's optional row tolerance, a fraction). Thresholds are the severity bands read from the result. Dimension is the default data-quality dimension the check is classified under; you can change it on any check.

Each section ends with examples: the body you would send to create the check (POST /api/v1/suites/{suite_id}/checks; the same fields work over MCP and in a suite import) and, for every type that runs on a plain table, the result DataQ reports on the sample below. Those results are produced by DataQ's own check engine and re-checked in CI.

The sample table every example below runs on

Ten orders, with one deliberate defect in most columns.

order_id	customer_email	status	amount	discount	subtotal	tax	total	country	sku	ordered_at	shipped_at	payload	legacy_code	invoice_total
1	ana@example.com	shipped	120.0	10.0	90.0	10.0	100.0	US	SKU-0001	2026-09-20 09:15:00	2026-09-21 10:00:00	{"gift": false}	NULL	100.0
2	bo@example.com	pending	75.5	0.0	80.0	20.0	100.0	GB	SKU-0002	2026-09-21 14:02:00	2026-09-22 08:30:00	{"gift": true}	NULL	100.0
3	chen@example.com	shipped	310.0	25.0	70.0	30.0	100.0	IN	SKU-0003	2026-09-22 18:45:00	2026-09-23 12:00:00	{}	NULL	100.0
4	NULL	cancelled	42.0	0.0	95.0	5.0	100.0	DE	SKU-0004	2026-09-23 07:30:00	2026-09-24 09:00:00	{"gift": false}	NULL	105.0
5	dev@example.com	refunded	18.25	2.0	60.0	40.0	100.0	US	SKU-0005	2026-09-24 11:11:00	2026-09-25 16:20:00	{"gift": false}	NULL	100.0
6	eli@example.com	shipped	-5.0	0.0	85.0	15.0	100.0	USA	SKU-0006	2026-09-25 22:05:00	2026-09-26 07:45:00	{"gift": true}	NULL	100.0
7	not-an-email	pending	99.99	150.0	50.0	50.0	100.0	FR	SKU-0007	2026-09-26 13:40:00	2026-09-27 10:10:00	{"gift": false}	NULL	100.0
8	fay@example.com	NULL	250.0	20.0	99.0	1.0	100.0	GB	sku_8	2026-09-27 08:00:00	2026-09-27 06:00:00	{bad json	NULL	100.0
9	gus@example.com	shipped	60.0	5.0	75.0	25.0	100.0	IN	SKU-0009	2026-09-27 19:25:00	2026-09-28 09:30:00	{"gift": false}	NULL	100.0
7	hal@example.com	shipped	88.0	8.0	40.0	50.0	90.0	US	SKU-0010	2026/09/28 06:00	2026-09-28 18:00:00	{"gift": true}	NULL	90.0

Column values

Great Expectations built-ins that look at the values in one or more columns. Each returns an unexpected-% that the warn / fail / critical severity bands read.

Check	Type	What it checks	Dimension	Parameters	Thresholds	Runs on
Column values not null	expect_column_values_to_not_be_null	Every value in the column is non-null.	Completeness	column, mostly (optional)	warn / fail / critical	All datasources · SQL pushdown on Unity Catalog
Column values unique	expect_column_values_to_be_unique	Values in the column are distinct (no duplicates).	Uniqueness	column, mostly (optional)	warn / fail / critical	All datasources · SQL pushdown on Unity Catalog
Column values in range	expect_column_values_to_be_between	Numeric values fall within [min, max].	Validity	column, min_value (optional), max_value (optional), mostly (optional)	warn / fail / critical	All datasources · SQL pushdown on Unity Catalog
Column values in set	expect_column_values_to_be_in_set	Every value is one of an allowed set.	Validity	column, value_set, mostly (optional)	warn / fail / critical	All datasources · SQL pushdown on Unity Catalog
Column values null	expect_column_values_to_be_null	Every value in the column is null — for a deprecated or not-yet-populated column that should stay empty. The inverse of “Column values not null”.	Validity	column, mostly (optional)	warn / fail / critical	All datasources · SQL pushdown on Unity Catalog
Column values not in set	expect_column_values_to_not_be_in_set	No value is one of a forbidden set — e.g. a status that should never reach this table, or placeholder values like “N/A” and “UNKNOWN”.	Validity	column, value_set, mostly (optional)	warn / fail / critical	All datasources · SQL pushdown on Unity Catalog
Column value lengths in range	expect_column_value_lengths_to_be_between	String lengths fall within [min, max].	Validity	column, min_value (optional), max_value (optional), mostly (optional)	warn / fail / critical	All datasources · SQL pushdown on Unity Catalog
Column value lengths equal	expect_column_value_lengths_to_equal	Every value is exactly the given number of characters — for a fixed-width code (ISO country, SKU, account number).	Validity	column, value, mostly (optional)	warn / fail / critical	All datasources · SQL pushdown on Unity Catalog
Column values match regex	expect_column_values_to_match_regex	Every value matches the given regular expression.	Validity	column, regex, mostly (optional)	warn / fail / critical	Snowflake, ADLS Gen2, AWS S3, Google Cloud Storage, Unity Catalog, Apache Iceberg, PostgreSQL, MySQL / MariaDB, Trino, Amazon Athena, Amazon Redshift, Google BigQuery — not SQL Server (no translation for that SQL dialect; refused at author time) · SQL pushdown on Unity Catalog
Column values do not match regex	expect_column_values_to_not_match_regex	No value matches the given regular expression — for catching a pattern that should never appear (a stray delimiter, an unredacted identifier).	Validity	column, regex, mostly (optional)	warn / fail / critical	Snowflake, ADLS Gen2, AWS S3, Google Cloud Storage, Unity Catalog, Apache Iceberg, PostgreSQL, MySQL / MariaDB, Trino, Amazon Athena, Amazon Redshift, Google BigQuery — not SQL Server (no translation for that SQL dialect; refused at author time) · SQL pushdown on Unity Catalog
Column values match a list of regexes	expect_column_values_to_match_regex_list	Every value matches the regexes in the list — by default ANY one of them is enough, for a column carrying several legitimate formats (e.g. two phone-number conventions).	Validity	column, regex_list, match_on (optional), mostly (optional)	warn / fail / critical	Snowflake, ADLS Gen2, AWS S3, Google Cloud Storage, Unity Catalog, Apache Iceberg, PostgreSQL, MySQL / MariaDB, Trino, Amazon Athena, Amazon Redshift, Google BigQuery — not SQL Server (no translation for that SQL dialect; refused at author time) · SQL pushdown on Unity Catalog
Column values match none of a list of regexes	expect_column_values_to_not_match_regex_list	No value matches ANY regex in the list — a deny-list of forbidden formats.	Validity	column, regex_list, mostly (optional)	warn / fail / critical	Snowflake, ADLS Gen2, AWS S3, Google Cloud Storage, Unity Catalog, Apache Iceberg, PostgreSQL, MySQL / MariaDB, Trino, Amazon Athena, Amazon Redshift, Google BigQuery — not SQL Server (no translation for that SQL dialect; refused at author time) · SQL pushdown on Unity Catalog
Column values are valid JSON	expect_column_values_to_be_json_parseable	Every value parses as JSON — for a payload/metadata column stored as text. Not offered on Snowflake, PostgreSQL, MySQL, Trino, SQL Server, Athena, Redshift or BigQuery: Great Expectations implements this one only for dataframe batches, so a SQL batch would error on every run. Use a custom-SQL check there (or, on Snowflake, a VARIANT column).	Validity	column, mostly (optional)	warn / fail / critical	ADLS Gen2, AWS S3, Google Cloud Storage, Unity Catalog, Apache Iceberg — not Snowflake, PostgreSQL, MySQL / MariaDB, Trino, SQL Server, Amazon Athena, Amazon Redshift, Google BigQuery (no SQL implementation; refused at author time)
Column values are of type	expect_column_values_to_be_of_type	Every value in the column matches the given data type.	Validity	column, type_, mostly (optional)	warn / fail / critical	All datasources
Column values are of one of several types	expect_column_values_to_be_in_type_list	Every value in the column matches at least one of the given data types — the tolerant sibling of “Column values are of type”, for a column whose type legitimately varies by datasource or load.	Validity	column, type_list, mostly (optional)	warn / fail / critical	All datasources
Compound columns unique	expect_compound_columns_to_be_unique	The COMBINATION of values across the listed columns is distinct on every row — a multi-column primary or business key. Each column on its own may repeat freely.	Uniqueness	column_list, mostly (optional)	warn / fail / critical	All datasources · SQL pushdown on Unity Catalog
Column A greater than column B	expect_column_pair_values_a_to_be_greater_than_b	Row by row, column A is greater than column B — e.g. ended_at > started_at, or total >= discount.	Validity	column_A, column_B, or_equal (optional), mostly (optional)	warn / fail / critical	All datasources · SQL pushdown on Unity Catalog
Column A equals column B	expect_column_pair_values_to_be_equal	Row by row, the two columns hold the same value — e.g. a denormalised copy that must agree with its source, or a total that must match a recomputed one.	Validity	column_A, column_B, mostly (optional)	warn / fail / critical	All datasources · SQL pushdown on Unity Catalog
Values unique within each row	expect_select_column_values_to_be_unique_within_record	Within a single row, the listed columns all hold different values — e.g. a transfer whose source and destination account must not be the same. This is per-row; use “Compound columns unique” for uniqueness ACROSS rows.	Uniqueness	column_list, mostly (optional)	warn / fail / critical	All datasources · SQL pushdown on Unity Catalog
Columns sum to a total	expect_multicolumn_sum_to_equal	Row by row, the listed columns add up to the given total — e.g. subtotal + tax + shipping = total.	Validity	column_list, sum_total, mostly (optional)	warn / fail / critical	All datasources · SQL pushdown on Unity Catalog
Column distinct values in set	expect_column_distinct_values_to_be_in_set	Every DISTINCT value present in the column is one of an allowed set — reports WHICH unexpected values exist rather than how many rows carry them. Use “Column values in set” when you care about the row count.	Validity	column, value_set	None — pass/fail only	All datasources · SQL pushdown on Unity Catalog
Column distinct values contain set	expect_column_distinct_values_to_contain_set	Every value in the given set appears at least once in the column — catches a category that stopped arriving. The column may also contain other values.	Completeness	column, value_set	None — pass/fail only	All datasources · SQL pushdown on Unity Catalog
Column values match a date format	expect_column_values_to_match_strftime_format	Every value parses under the given strftime format — for a date or timestamp stored as text. Not offered on Snowflake, PostgreSQL, MySQL, Trino, SQL Server, Athena, Redshift or BigQuery: Great Expectations implements this one only for dataframe batches, so a SQL batch would error on every run. Use a custom-SQL check there.	Validity	column, strftime_format, mostly (optional)	warn / fail / critical	ADLS Gen2, AWS S3, Google Cloud Storage, Unity Catalog, Apache Iceberg — not Snowflake, PostgreSQL, MySQL / MariaDB, Trino, SQL Server, Amazon Athena, Amazon Redshift, Google BigQuery (no SQL implementation; refused at author time)

Examples

Column values not null

{
  "name": "Column values not null",
  "expectation_type": "expect_column_values_to_not_be_null",
  "config": {
    "column": "customer_email"
  },
  "warn_threshold": 0,
  "fail_threshold": 5
}

On the sample: FAIL · 10% unexpected (1 row) · unexpected values [null]

Order 4 has no email: 1 of 10 rows (10%).

Column values unique

{
  "name": "Column values unique",
  "expectation_type": "expect_column_values_to_be_unique",
  "config": {
    "column": "order_id"
  }
}

On the sample: FAIL · 20% unexpected (2 rows) · unexpected values [7, 7]

Order id 7 appears twice, so both rows are unexpected.

Column values in range

{
  "name": "Column values in range",
  "expectation_type": "expect_column_values_to_be_between",
  "config": {
    "column": "amount",
    "min_value": 0,
    "max_value": 1000
  }
}

On the sample: FAIL · 10% unexpected (1 row) · unexpected values [-5.0]

Order 6 has a negative amount.

Column values in set

{
  "name": "Column values in set",
  "expectation_type": "expect_column_values_to_be_in_set",
  "config": {
    "column": "status",
    "value_set": [
      "pending",
      "shipped",
      "cancelled"
    ]
  }
}

On the sample: FAIL · 11.11% unexpected (1 row) · unexpected values ["refunded"]

refunded is not in the allowed set: 1 of the 9 non-NULL values (GX skips a NULL).

Column values null

{
  "name": "Column values null",
  "expectation_type": "expect_column_values_to_be_null",
  "config": {
    "column": "legacy_code"
  }
}

On the sample: PASS · 0% unexpected (0 rows)

A retired column that must stay empty: every row is NULL.

Column values not in set

{
  "name": "Column values not in set",
  "expectation_type": "expect_column_values_to_not_be_in_set",
  "config": {
    "column": "status",
    "value_set": [
      "refunded",
      "unknown"
    ]
  }
}

On the sample: FAIL · 11.11% unexpected (1 row) · unexpected values ["refunded"]

One order is refunded, a status that should never reach this table.

Column value lengths in range

{
  "name": "Column value lengths in range",
  "expectation_type": "expect_column_value_lengths_to_be_between",
  "config": {
    "column": "sku",
    "min_value": 8,
    "max_value": 8
  }
}

On the sample: FAIL · 10% unexpected (1 row) · unexpected values ["sku_8"]

sku_8 is five characters.

Column value lengths equal

{
  "name": "Column value lengths equal",
  "expectation_type": "expect_column_value_lengths_to_equal",
  "config": {
    "column": "country",
    "value": 2
  }
}

On the sample: FAIL · 10% unexpected (1 row) · unexpected values ["USA"]

USA is not an ISO 3166 alpha-2 code.

Column values match regex

{
  "name": "Column values match regex",
  "expectation_type": "expect_column_values_to_match_regex",
  "config": {
    "column": "sku",
    "regex": "^SKU-\\d{4}$"
  }
}

On the sample: FAIL · 10% unexpected (1 row) · unexpected values ["sku_8"]

sku_8 doesn't match the SKU pattern.

Column values do not match regex

{
  "name": "Column values do not match regex",
  "expectation_type": "expect_column_values_to_not_match_regex",
  "config": {
    "column": "customer_email",
    "regex": "^[^@]+$"
  }
}

On the sample: FAIL · 11.11% unexpected (1 row) · unexpected values ["not-an-email"]

A value with no @ is not an email: not-an-email.

Column values match a list of regexes

{
  "name": "Column values match a list of regexes",
  "expectation_type": "expect_column_values_to_match_regex_list",
  "config": {
    "column": "sku",
    "regex_list": [
      "^SKU-\\d{4}$",
      "^LEGACY-\\d+$"
    ],
    "match_on": "any"
  }
}

On the sample: FAIL · 10% unexpected (1 row) · unexpected values ["sku_8"]

Two accepted formats; sku_8 matches neither.

Column values match none of a list of regexes

{
  "name": "Column values match none of a list of regexes",
  "expectation_type": "expect_column_values_to_not_match_regex_list",
  "config": {
    "column": "customer_email",
    "regex_list": [
      "@test\\.",
      "^not-"
    ]
  }
}

On the sample: FAIL · 11.11% unexpected (1 row) · unexpected values ["not-an-email"]

A deny-list of placeholder addresses; not-an-email is caught.

Column values are valid JSON

{
  "name": "Column values are valid JSON",
  "expectation_type": "expect_column_values_to_be_json_parseable",
  "config": {
    "column": "payload"
  }
}

On the sample: FAIL · 10% unexpected (1 row) · unexpected values ["{bad json"]

Order 8's payload is truncated JSON.

Column values are of type

{
  "name": "Column values are of type",
  "expectation_type": "expect_column_values_to_be_of_type",
  "config": {
    "column": "amount",
    "type_": "float64"
  }
}

On the sample: PASS · observed "float64"

The sample is a pandas frame, so the dtype name is float64 (see the dtype table in the datasources guide).

Column values are of one of several types

{
  "name": "Column values are of one of several types",
  "expectation_type": "expect_column_values_to_be_in_type_list",
  "config": {
    "column": "order_id",
    "type_list": [
      "int64",
      "int32"
    ]
  }
}

On the sample: PASS · observed "int64"

Either integer width is accepted.

Compound columns unique

{
  "name": "Compound columns unique",
  "expectation_type": "expect_compound_columns_to_be_unique",
  "config": {
    "column_list": [
      "order_id",
      "customer_email"
    ]
  }
}

On the sample: PASS · 0% unexpected (0 rows)

The duplicated order id carries two different emails, so the pair is still unique.

Column A greater than column B

{
  "name": "Column A greater than column B",
  "expectation_type": "expect_column_pair_values_a_to_be_greater_than_b",
  "config": {
    "column_A": "shipped_at",
    "column_B": "ordered_at",
    "or_equal": true
  }
}

On the sample: FAIL · 20% unexpected (2 rows) · unexpected values [["2026-09-27 06:00:00", "2026-09-27 08:00:00"], ["2026-09-28 18:00:00", "2026/09/28 06:00"]]

Order 8 shipped before it was ordered. These are text columns, so the last row, whose date is written 2026/09/28 06:00, also compares out of order: parse dates upstream.

Column A equals column B

{
  "name": "Column A equals column B",
  "expectation_type": "expect_column_pair_values_to_be_equal",
  "config": {
    "column_A": "total",
    "column_B": "invoice_total"
  }
}

On the sample: FAIL · 10% unexpected (1 row) · unexpected values [[100.0, 105.0]]

Order 4 was invoiced 5 more than its total, so 1 row of 10 fails.

Values unique within each row

{
  "name": "Values unique within each row",
  "expectation_type": "expect_select_column_values_to_be_unique_within_record",
  "config": {
    "column_list": [
      "subtotal",
      "tax"
    ]
  }
}

On the sample: FAIL · 10% unexpected (1 row) · unexpected values [{"subtotal": 50.0, "tax": 50.0}]

Order 7 has subtotal and tax both 50.

Columns sum to a total

{
  "name": "Columns sum to a total",
  "expectation_type": "expect_multicolumn_sum_to_equal",
  "config": {
    "column_list": [
      "subtotal",
      "tax"
    ],
    "sum_total": 100
  }
}

On the sample: FAIL · 10% unexpected (1 row) · unexpected values [{"subtotal": 40.0, "tax": 50.0}]

The last row's parts sum to 90.

Column distinct values in set

{
  "name": "Column distinct values in set",
  "expectation_type": "expect_column_distinct_values_to_be_in_set",
  "config": {
    "column": "country",
    "value_set": [
      "US",
      "GB",
      "IN",
      "DE",
      "FR"
    ]
  }
}

On the sample: FAIL · unexpected values ["USA"]

USA is outside the set. There is no unexpected-%: the check passes or fails on the set of distinct values, and the failing sample lists the ones outside it.

Column distinct values contain set

{
  "name": "Column distinct values contain set",
  "expectation_type": "expect_column_distinct_values_to_contain_set",
  "config": {
    "column": "country",
    "value_set": [
      "US",
      "GB",
      "JP"
    ]
  }
}

On the sample: FAIL

No order comes from Japan.

Column values match a date format

{
  "name": "Column values match a date format",
  "expectation_type": "expect_column_values_to_match_strftime_format",
  "config": {
    "column": "ordered_at",
    "strftime_format": "%Y-%m-%d %H:%M:%S"
  }
}

On the sample: FAIL · 10% unexpected (1 row) · unexpected values ["2026/09/28 06:00"]

The last row was written as 2026/09/28 06:00.

Table shape

Whole-table expectations.

Check	Type	What it checks	Dimension	Parameters	Thresholds	Runs on
Table row count in range	expect_table_row_count_to_be_between	The table’s row count falls within [min, max].	Completeness	min_value (optional), max_value (optional)	warn / fail / critical	All datasources · SQL pushdown on Unity Catalog

Examples

Table row count in range

{
  "name": "Table row count in range",
  "expectation_type": "expect_table_row_count_to_be_between",
  "config": {
    "min_value": 5,
    "max_value": 1000
  }
}

On the sample: PASS · observed 10

Ten rows is inside the band.

Freshness

How stale is the target? Measured from a timestamp column (or file arrival time on flat files), reported in hours, banded by age. Requires a fail or critical threshold.

Check	Type	What it checks	Dimension	Parameters	Thresholds	Runs on
Freshness	monitor:freshness	How stale is the target? Measures hours since the latest timestamp in the data — or, on a flat file with no column set, since the file last landed.	Timeliness	column	warn / fail / critical (fail or critical required)	All datasources

Examples

Freshness

{
  "name": "Freshness",
  "expectation_type": "monitor:freshness",
  "kind": "freshness",
  "config": {
    "column": "shipped_at"
  },
  "warn_threshold": 24,
  "fail_threshold": 48
}

Reports the newest shipped_at and its age in hours as the metric; with these thresholds, 24 h old or more warns and 48 h fails. Leave column out on a flat file to measure the file's arrival time instead.

Volume

Did the load deliver the expected row count? Banded by count. Requires a fail or critical threshold.

Check	Type	What it checks	Dimension	Parameters	Thresholds	Runs on
Volume	monitor:volume	Did the load deliver the expected row count? Flags a count outside an allowed range.	Completeness	min_rows, max_rows	warn / fail / critical	All datasources

Examples

Volume

{
  "name": "Volume",
  "expectation_type": "monitor:volume",
  "kind": "volume",
  "config": {
    "min_rows": 5,
    "max_rows": 1000
  }
}

Reports the row count as the metric; outside the band fails. On the sample, 10 rows passes.

Aggregate

Does a column statistic — mean, median, sum, standard deviation, min or max — stay inside two-sided bands? The value is recorded every run, so it trends; an empty table or all-NULL column reports error, never a made-up 0.

Check	Type	What it checks	Dimension	Parameters	Thresholds	Runs on
Aggregate statistic	monitor:aggregate	A column statistic — mean, median, sum, standard deviation, min or max — stays inside two-sided bands. The value itself is recorded every run, so it trends. An empty table or an all-NULL column reports error, never a made-up 0.	— (set it yourself)	aggregate, column, min_value (optional), max_value (optional), warn_min (optional), warn_max (optional), critical_min (optional), critical_max (optional)	Two-sided warn / fail / critical bounds, set as parameters	All datasources — median not on MySQL / MariaDB, Trino, Amazon Athena, Google BigQuery (no exact median; refused at author time)

Examples

Aggregate statistic

{
  "name": "Aggregate statistic",
  "expectation_type": "monitor:aggregate",
  "kind": "aggregate",
  "config": {
    "aggregate": "mean",
    "column": "amount",
    "min_value": 50,
    "warn_max": 100
  }
}

Reports the mean of amount as the metric and records it every run, so it trends. On the sample the mean is 105.87: above warn_max, inside the fail band, so WARN.

Schema

Did the table's columns change against a captured baseline?

Check	Type	What it checks	Dimension	Parameters	Thresholds	Runs on
Schema drift	monitor:schema_drift	Did the table’s column shape change? Diffs the live columns (names + types) against a baseline captured on the first run. Works on every datasource — warehouses via information_schema, flat files via the file header/footer, Iceberg from table metadata.	Consistency	ignore_columns (optional)	warn / fail / critical	All datasources

Examples

Schema drift

{
  "name": "Schema drift",
  "expectation_type": "monitor:schema_drift",
  "kind": "schema_drift",
  "config": {
    "ignore_columns": [
      "legacy_code"
    ]
  }
}

The first run captures the columns and their types as the baseline; later runs report every added, dropped or retyped column. legacy_code is never compared.

Anomaly

Is today's value unusual against a rolling baseline of this check's own history? Skips until enough history exists.

Check	Type	What it checks	Dimension	Parameters	Thresholds	Runs on
Anomaly	monitor:anomaly	Learns a rolling baseline (mean/stddev) from this check’s own metric history and flags how far this run deviates (a z-score). Reports skip, never a fake pass/fail, until enough history accrues.	— (set it yourself)	target_metric, date_column, value_column, period (optional), timezone (optional), settle_hours (optional), backfill (optional), column, window (optional), min_points (optional), seasonality (optional), yearly (optional), seasonal_periods (optional)	warn / fail / critical (fail or critical required)	Snowflake, Unity Catalog, PostgreSQL, MySQL / MariaDB, Trino, SQL Server, Amazon Athena, Amazon Redshift, Google BigQuery

Examples

Anomaly

{
  "name": "Anomaly",
  "expectation_type": "monitor:anomaly",
  "kind": "anomaly",
  "config": {
    "target_metric": "row_count",
    "window": 14,
    "min_points": 7,
    "seasonality": true
  },
  "warn_threshold": 2,
  "fail_threshold": 3
}

Reports the z-score of today's row count against the last 14 runs on the same weekday (seasonality keeps only those; off, it is the last 14 runs of any day). It skips until 7 such runs exist, so on a daily schedule about seven weeks.

Privacy

Did a column start carrying personal data that nobody classified? Profiles the table every run; the first run learns the current set and reports skip.

Check	Type	What it checks	Dimension	Parameters	Thresholds	Runs on
PII drift	monitor:pii_drift	Did a column start carrying personal data that nobody classified? Every run profiles every column, keeps what looks like email addresses or encoded blobs as a classification of the asset, and counts the columns that are new against the baseline and carry no warehouse tag, no inherited classification and no entry in this suite’s pii_columns (the identifier column is not exempt). The first run only learns the current set and reports skip; the baseline never moves on its own, so the check keeps failing until the column is classified, dismissed or ignored, or the check is re-baselined.	— (set it yourself)	ignore_columns (optional)	warn / fail / critical	All datasources

Examples

PII drift

{
  "name": "PII drift",
  "expectation_type": "monitor:pii_drift",
  "kind": "pii_drift",
  "config": {
    "ignore_columns": [
      "session_token"
    ]
  },
  "warn_threshold": 1,
  "fail_threshold": 1
}

The first run profiles every column, keeps what looks like email addresses or encoded blobs as a classification of the asset, and reports skip. Each later run counts the columns that are new against that baseline and that nobody has classified (no warehouse tag, no inherited classification, not in this suite's pii_columns, not dismissed) and fails on one — and keeps failing until someone classifies, dismisses or ignores the column. session_token is never counted.

Comparison

Reconcile the suite's target against a second dataset, possibly on another connection.

Check	Type	What it checks	Dimension	Parameters	Thresholds	Runs on
Records reconciliation	comparison:records	Diff this suite’s dataset (the target under test) against a baseline on another connection, joined on key columns — matched / mismatched / additional-per-side ROW buckets.	Consistency	—	warn / fail / critical	All datasources
Column-level reconciliation	comparison:columns	Same key-joined diff, counted per VALUE: each column reports its own matched / mismatched / additional-per-side counts. Pick this when you need to know WHICH columns drift, not just which rows.	Consistency	—	warn / fail / critical	All datasources

Examples

Records reconciliation

{
  "name": "Records reconciliation",
  "expectation_type": "comparison:records",
  "kind": "comparison",
  "config": {
    "source": {
      "table": "orders_staging",
      "schema": "public"
    },
    "keys": [
      "order_id"
    ]
  },
  "source_connection_id": "<connection id>"
}

Joins the suite's target to orders_staging on order_id and reports matched, mismatched and missing rows on each side; set source_connection_id to the connection holding the source.

Column-level reconciliation

{
  "name": "Column-level reconciliation",
  "expectation_type": "comparison:columns",
  "kind": "comparison",
  "config": {
    "source": {
      "table": "orders_staging",
      "schema": "public"
    },
    "keys": [
      "order_id"
    ],
    "columns": [
      "amount",
      "status"
    ]
  },
  "source_connection_id": "<connection id>"
}

Like the record comparison, but compares only the listed columns and reports the mismatch rate for each.

Custom SQL

Any predicate you can write in SQL, validated before it runs.

Check	Type	What it checks	Dimension	Parameters	Thresholds	Runs on
Custom SQL	unexpected_rows_expectation	A SQL query that should return no rows — any rows it returns are failures.	— (set it yourself)	unexpected_rows_query	warn / fail / critical	Snowflake, Unity Catalog, PostgreSQL, MySQL / MariaDB, Trino, SQL Server, Amazon Athena, Amazon Redshift, Google BigQuery

Examples

Custom SQL

{
  "name": "Custom SQL",
  "expectation_type": "unexpected_rows_expectation",
  "config": {
    "unexpected_rows_query": "SELECT * FROM {batch} WHERE discount > amount"
  }
}

Every row the query returns is unexpected. On the sample, order 7's discount (150) exceeds its amount, so 1 row fails. {batch} is the suite's target.

Snowflake DMF

Snowflake's native Data Metric Functions, evaluated inside Snowflake.

Check	Type	What it checks	Dimension	Parameters	Thresholds	Runs on
Null count (DMF)	dmf:null_count	Snowflake’s system NULL_COUNT metric function, computed natively in the warehouse.	Completeness	column	warn / fail / critical (fail or critical required)	Snowflake
Null percent (DMF)	dmf:null_percent	Snowflake’s system NULL_PERCENT metric function (0–100), computed natively in the warehouse.	Completeness	column	warn / fail / critical (fail or critical required)	Snowflake
Duplicate count (DMF)	dmf:duplicate_count	Snowflake’s system DUPLICATE_COUNT metric function, computed natively in the warehouse.	Uniqueness	column	warn / fail / critical (fail or critical required)	Snowflake
Blank count (DMF)	dmf:blank_count	Snowflake’s system BLANK_COUNT metric function, computed natively in the warehouse — counts empty or space-only strings (not NULLs; tabs/newlines aren’t treated as blank). VARCHAR columns only.	Completeness	column	warn / fail / critical (fail or critical required)	Snowflake
Future timestamp percent (DMF)	dmf:future_timestamp_percent	Snowflake’s system FUTURE_TIMESTAMP_PERCENT metric function (0–100): the share of rows dated after the evaluation time. DATE, TIMESTAMP_LTZ and TIMESTAMP_TZ columns only.	Validity	column	warn / fail / critical (fail or critical required)	Snowflake
Accepted values (DMF)	dmf:accepted_values	Snowflake’s system ACCEPTED_VALUES metric function, evaluated in the warehouse: counts rows whose value is not in the list. NULLs are not counted as violations.	Validity	column, value_set	warn / fail / critical (fail or critical required)	Snowflake
Unique count (DMF)	dmf:unique_count	Snowflake’s system UNIQUE_COUNT metric function, computed natively in the warehouse. Degrades downward, so this type carries no thresholds — read the observed value directly.	Uniqueness	column	None — pass/fail only	Snowflake
Custom DMF	dmf:custom	A data metric function your team created in Snowflake (CREATE DATA METRIC FUNCTION), called in the warehouse over columns of this suite’s table. Its return value is the metric, shown as-is — write DMFs that return a count or a percentage, not a data value.	— (set it yourself)	function, columns	warn / fail / critical (fail or critical required)	Snowflake

Examples

Null count (DMF)

{
  "name": "Null count (DMF)",
  "expectation_type": "dmf:null_count",
  "engine": "dmf",
  "config": {
    "column": "CUSTOMER_EMAIL"
  },
  "fail_threshold": 1
}

Snowflake counts the NULLs itself (SNOWFLAKE.CORE.NULL_COUNT); the count is the metric. On the sample it would be 1.

Null percent (DMF)

{
  "name": "Null percent (DMF)",
  "expectation_type": "dmf:null_percent",
  "engine": "dmf",
  "config": {
    "column": "CUSTOMER_EMAIL"
  },
  "warn_threshold": 1,
  "fail_threshold": 5
}

The NULL percentage (10 on the sample).

Duplicate count (DMF)

{
  "name": "Duplicate count (DMF)",
  "expectation_type": "dmf:duplicate_count",
  "engine": "dmf",
  "config": {
    "column": "ORDER_ID"
  },
  "fail_threshold": 1
}

Rows whose value repeats another row's (2 on the sample: both order 7s).

Blank count (DMF)

{
  "name": "Blank count (DMF)",
  "expectation_type": "dmf:blank_count",
  "engine": "dmf",
  "config": {
    "column": "STATUS"
  },
  "fail_threshold": 1
}

Empty or whitespace-only strings; a NULL is not blank.

Future timestamp percent (DMF)

{
  "name": "Future timestamp percent (DMF)",
  "expectation_type": "dmf:future_timestamp_percent",
  "engine": "dmf",
  "config": {
    "column": "SHIPPED_AT"
  },
  "fail_threshold": 0.01
}

The percentage of timestamps later than now; takes DATE, TIMESTAMP_LTZ and TIMESTAMP_TZ columns only.

Accepted values (DMF)

{
  "name": "Accepted values (DMF)",
  "expectation_type": "dmf:accepted_values",
  "engine": "dmf",
  "config": {
    "column": "STATUS",
    "value_set": [
      "pending",
      "shipped",
      "cancelled"
    ]
  },
  "fail_threshold": 1
}

Values outside the set.

Unique count (DMF)

{
  "name": "Unique count (DMF)",
  "expectation_type": "dmf:unique_count",
  "engine": "dmf",
  "config": {
    "column": "ORDER_ID"
  }
}

The number of distinct non-NULL values (9 on the sample).

Custom DMF

{
  "name": "Custom DMF",
  "expectation_type": "dmf:custom",
  "engine": "dmf",
  "config": {
    "function": "ANALYTICS.DQ.NEGATIVE_AMOUNTS",
    "columns": [
      "AMOUNT"
    ]
  },
  "fail_threshold": 1
}

Calls your own data metric function over the listed columns; whatever it returns is the metric, banded by the thresholds.

Databricks DQX

Databricks Labs DQX row rules, evaluated by a serverless job in your own workspace.

Check	Type	What it checks	Dimension	Parameters	Thresholds	Runs on
Not null (DQX)	dqx:is_not_null	Rows where the column is NULL.	Completeness	column, mode (optional)	warn / fail / critical	All datasources
Not empty (DQX)	dqx:is_not_empty	Rows where the column is an empty string.	Completeness	column, mode (optional)	warn / fail / critical	All datasources
Not null or empty (DQX)	dqx:is_not_null_and_not_empty	Rows where the column is NULL or an empty string.	Completeness	column, mode (optional)	warn / fail / critical	All datasources
In list (DQX)	dqx:is_in_list	Rows whose value is not one of the allowed values.	Validity	column, allowed, mode (optional)	warn / fail / critical	All datasources
In range (DQX)	dqx:is_in_range	Rows whose value falls outside the inclusive range.	Validity	column, min_limit, max_limit, mode (optional)	warn / fail / critical	All datasources
Matches regex (DQX)	dqx:regex_match	Rows whose value does not match the regular expression.	Validity	column, regex, mode (optional)	warn / fail / critical	All datasources
Not less than (DQX)	dqx:is_not_less_than	Rows whose value is below the limit.	Validity	column, limit, mode (optional)	warn / fail / critical	All datasources
Not greater than (DQX)	dqx:is_not_greater_than	Rows whose value is above the limit.	Validity	column, limit, mode (optional)	warn / fail / critical	All datasources

Examples

Not null (DQX)

{
  "name": "Not null (DQX)",
  "expectation_type": "dqx:is_not_null",
  "engine": "dqx",
  "config": {
    "column": "customer_email"
  },
  "fail_threshold": 1
}

Databricks DQX counts failing rows in your workspace; the failing count is the metric (1 on the sample).

Not empty (DQX)

{
  "name": "Not empty (DQX)",
  "expectation_type": "dqx:is_not_empty",
  "engine": "dqx",
  "config": {
    "column": "status"
  },
  "fail_threshold": 1
}

Rows whose value is an empty string.

Not null or empty (DQX)

{
  "name": "Not null or empty (DQX)",
  "expectation_type": "dqx:is_not_null_and_not_empty",
  "engine": "dqx",
  "config": {
    "column": "status",
    "mode": "stream"
  },
  "fail_threshold": 1
}

NULL or empty. mode: stream evaluates only rows appended since the last run.

In list (DQX)

{
  "name": "In list (DQX)",
  "expectation_type": "dqx:is_in_list",
  "engine": "dqx",
  "config": {
    "column": "status",
    "allowed": [
      "pending",
      "shipped",
      "cancelled"
    ]
  },
  "fail_threshold": 1
}

Rows outside the list (refunded on the sample).

In range (DQX)

{
  "name": "In range (DQX)",
  "expectation_type": "dqx:is_in_range",
  "engine": "dqx",
  "config": {
    "column": "amount",
    "min_limit": 0,
    "max_limit": 1000
  },
  "fail_threshold": 1
}

Rows outside [0, 1000] (order 6).

Matches regex (DQX)

{
  "name": "Matches regex (DQX)",
  "expectation_type": "dqx:regex_match",
  "engine": "dqx",
  "config": {
    "column": "sku",
    "regex": "^SKU-[0-9]{4}$"
  },
  "fail_threshold": 1
}

Rows not matching (sku_8).

Not less than (DQX)

{
  "name": "Not less than (DQX)",
  "expectation_type": "dqx:is_not_less_than",
  "engine": "dqx",
  "config": {
    "column": "amount",
    "limit": 0
  },
  "fail_threshold": 1
}

Rows below the limit.

Not greater than (DQX)

{
  "name": "Not greater than (DQX)",
  "expectation_type": "dqx:is_not_greater_than",
  "engine": "dqx",
  "config": {
    "column": "discount",
    "limit": 100
  },
  "fail_threshold": 1
}

Rows above the limit (order 7's discount of 150).

Authorable outside the editor

Vetted by the backend but with no editor widget: usable over the REST API, MCP and suite import, which hand the backend raw JSON.

• expect_column_pair_values_to_be_in_set

Not offered, and why

GX's scalar aggregates (expect_column_mean_to_be_between and its siblings) report one number and no unexpected-%, so severity bands have nothing to band — the Aggregate monitor measures that shape instead, with two-sided bands and a trend. Whole-table column-set comparisons are what the Schema-drift monitor does against a captured baseline. For anything else, write a custom-SQL check.

---

Generated by scripts/docs/gen-check-catalog.py — edit the catalog or the allowlist, not this page.