Skip to content

Results

Each entry in the results array is a single observation — either a check (a pass/fail assertion) or a metric (a measured value). A check reports a pass/fail outcome; a metric reports a measured value (its status still indicates whether the value met expectations).

Back to TOC

Example

Minimal

A result requires only id, type, name, and status.

- id: no_dup_rows
  type: check
  name: No duplicate rows
  status: pass
{
  "id": "no_dup_rows",
  "type": "check",
  "name": "No duplicate rows",
  "status": "pass"
}

Full

A result using every field — a failing check linked to a contract rule via relatesTo, with a target, a measured value against a threshold, a message, and custom properties.

- id: dup_rows_myTable
  type: check
  name: No duplicate rows
  status: fail
  relatesTo: schema/orders_tbl/quality/dq_no_duplicates
  severity: critical
  dimension: uniqueness
  target:
    resourceType: table
    resourceIdentifier: myTable
    filter: order_date >= '2026-01-01'
  measure:
    metric: duplicateValues
    value: 3
    threshold:
      mustBe: 0
    unit: rows
    rowCount: 1200
    failedRowCount: 3
  message: Multiple duplicate rows found in myTable!
  tags: ['critical-path', 'uniqueness']
  customProperties:
    - property: checkEngine
      value: soda-core
{
  "id": "dup_rows_myTable",
  "type": "check",
  "name": "No duplicate rows",
  "status": "fail",
  "relatesTo": "schema/orders_tbl/quality/dq_no_duplicates",
  "severity": "critical",
  "dimension": "uniqueness",
  "target": {
    "resourceType": "table",
    "resourceIdentifier": "myTable",
    "filter": "order_date >= '2026-01-01'"
  },
  "measure": {
    "metric": "duplicateValues",
    "value": 3,
    "threshold": {
      "mustBe": 0
    },
    "unit": "rows",
    "rowCount": 1200,
    "failedRowCount": 3
  },
  "message": "Multiple duplicate rows found in myTable!",
  "tags": ["critical-path", "uniqueness"],
  "customProperties": [
    {
      "property": "checkEngine",
      "value": "soda-core"
    }
  ]
}

Field Descriptions

Key Type UX label Required Description Example
id string Result ID Yes A unique identifier for the specific check/metric. dup_rows_myTable
type string Type Yes The type of observation: check (pass/fail) or metric (measured value). check
name string Name Yes A human-readable name for the check/metric. No duplicate rows
status string Status Yes The outcome of the observation: pass, fail, warn, error, skip, or unknown. See Status values. fail
relatesTo string Relates To No A reference to the element this result relates to in the linked contract (e.g., a path to a data quality rule or property). schema/orders_tbl/quality/dq_no_duplicates
severity string Severity No The severity level, e.g., critical, high, medium, low, warning. Orthogonal to status: status is the runtime outcome, severity is the definition-time importance. critical
dimension string Dimension No The data quality dimension (following ODCS data quality dimensions), e.g., accuracy, completeness, timeliness, uniqueness. uniqueness
target object Target No Object describing the resource being observed.
target.resourceType string Resource Type No The type of resource, e.g., table, column, dataset. table
target.resourceIdentifier string / array Resource Identifier No The identifier(s) of the resource — a single string or an array of strings. myTable
target.filter string Filter No A condition that restricted the observed data to a subset of the resource, e.g., a partition or a where clause. When present, the result applies only to that subset, not to the whole resource. order_date >= '2026-10-01'
measure object Measure No Object describing the measurement details.
measure.metric string Metric No The name of the metric being measured, e.g., rowCount, age, latency. For data quality metrics, use the ODCS library metric names where one applies: rowCount, nullValues, missingValues, invalidValues, duplicateValues. duplicateValues
measure.value number / string Value No The observed value of the metric. 3
measure.threshold object Threshold No The threshold criteria, e.g., mustBe, mustBeLessThan (following ODCS operator definitions). { "mustBe": 0 }
measure.unit string Unit No The unit of measurement, e.g., rows, seconds, percent. rows
measure.rowCount integer Row Count No The number of rows the observation evaluated. 153214
measure.failedRowCount integer Failed Row Count No The number of evaluated rows that did not meet the expectation. Together with rowCount, it gives the share of failing rows, independent of the metric in value. 3
message string Message No A human-readable message providing additional context about the observation. Multiple duplicate rows found!
tags array Tags No A list of tags for categorizing the result. See Tags. ['critical-path']
customProperties array Custom Properties No A list of key/value pairs for custom properties. See Custom & Other Properties.

Status values

status separates a known-bad verdict from the absence of a verdict:

Value Meaning
pass Executed, met expectation.
fail Executed, breached a hard threshold.
warn Executed, breached a soft/warning threshold — alert, but do not fail a gate.
error Could not be evaluated — verdict unknown (e.g. the query failed, or the source was unreachable).
skip Intentionally not run (filtered out, disabled, or an unmet precondition).
unknown The producer cannot tell the outcome, e.g. it relays a result from another tool whose status it does not recognize.

error is deliberately distinct from fail: a failure is a known-bad verdict, an error is the absence of a verdict and must not be counted as a failure. Counting an infrastructure error as a data failure poisons dashboards and triggers the wrong remediation. Likewise, a skip must not read as either a pass or a fail.

unknown is distinct from error: an error means the evaluation broke and someone has to fix it, while unknown means the producer has no outcome to report. Like skip, it must be counted neither as a pass nor as a failure.

severity is orthogonal to status and composes with it — a critical-severity check with status: error is the loudest possible alert, because the check you most care about could not even run.

Example

- id: dup_rows_myTable
  type: check
  name: No duplicate rows
  status: fail
- id: freshness_myDataProduct
  type: metric
  name: Data should not be stale
  status: warn
- id: schema_conformance_orders
  type: check
  name: Orders conform to the contract schema
  status: error
  severity: critical
- id: pii_scan_orders
  type: check
  name: No PII in the orders extract
  status: skip
- id: positive_amounts_orders
  type: check
  name: Order amounts are positive
  status: unknown