Incident Types

Upriver classifies the data incidents into different types so that users can immediately understand what is happening in their data.

Type	Meaning	Examples
Schema Incident	The schema of the data source has changed.	Type change from string to int for a specific field. A new field has been added/removed in the data source.
Metata Incident	An anomaly was detected in either the freshness or the volume of data.	A sharp drop in the number of items being ingested. An anomaly in the age of data in the data source.
Format Incident	The semantic format for a given field has changed or the ratio between different formats seen for the field has changed.	Change in the distribution between URLs and IPs in a domain field. A field receiving emails begins seeing malformed emails values.
Distribution Drift	A change has been detected in the distribution of one of the measured metrics for the datasource. This can include change in the expected ranges, percentiles and so on. These incidents can be caused either by anomalous values for a given metric or by a drift in the distribution, i.e. the expectations from the datasources have changed.	An out of range value in a numerical column. A shift in the expected mean value of a numerical column after multiple runs.
Field Metadata Incident	An anomaly was detected in the completeness or uniqueness of a given field.	The completeness (ratio of non-null values to null values) has changed for a specific field/column in the data.
Unmet User Field Expectation	The behavior of a field does not match an expectation explicitly set by the user.	A vlue outside of a given range defined by the user. The completeness of a field is below the threshold the user defined.
Unmet User Metadata Expectation	The volume or freshness do not match the expectation set by the user.	The newest available data is older than a certain threshold defined. The number items being ingested is higher than a given threshold.

Incidents groups might include more than one type of incident. For example, a drop in the number of rows can appear as both an "Unmet User Metadata Expectation" and as a "Metadata Expectation".

Incidents can be filtered by there type in the incident list page.