Managing Incidents

Overview

Upriver enables organizations to detect, group, and manage data incidents across their entire data ecosystem.

It intelligently groups related incidents together, allowing you to comment, update statuses, and assign appropriate ownership. This ensures that every incident is addressed properly and efficiently, promoting better collaboration and maintaining data integrity.

Incident Grouping

Upriver reduces alert volume by grouping alerts based on data source, timeframe, and incident type, providing contextual insight into the broader data incident. This allows users to understand the scope and root cause of the issue more quickly.

In addition, incidents for data sources with Pivot Fields are also grouped so that only the top level incident is shown unless the user chooses to do a deep dive into to the incident.

Grouped and derived incidents

Incident Status

The status of an incident can be changed by the user and serves the following purposes:

• Improve communication - Make sure all relevant stakeholder to know which incidents have been acknowledged, reviewed or dismissed. This allows you to ensure that no incident goes unhandled by your team.

• Improve operational efficiency - Track the number of incidents and the time taken to triage and resolve issues. This will help you see progress over time, set goals and communicate the effects of data quality to the rest of the organization.

• Upriver's Model precision - Upriver’s AI system takes in to consideration the status given to each incident by removing or setting weights according to the status in order to help set more accurate thresholds for each metric.

Status	Meaning	Effect on Upriver's Model
Unresolved	The incident was created and not yet triaged by a user.	No effect.
Under Review	Someone is aware of this and is currently assessing whether it’s a significant issue.	No effect.
Fixed	The issue was fixed and should behave like it did before.	Incident is ignored when retraining the model and the profile before the incident is expected.
Expected Change	The incident is valid but is the result of an expected change.	Retrain the model taking into consideration the new values.
Misdetection	The detection is not valid.	Retrain the model with taking into consideration the incident as a valid value.

Severity

Upriver can help with your incident management process by setting clear severity for each incident. In addition, Upriver can set a default severity for each data source to help users manage data quality incidents in a more automated way.

Comments

Upriver allows users to add and edit comment on each incident in order to increase the collaboration around every incident.

Owners

Assign owners to incidents to make clear who is responsible for investigating the alert, and a severity to classify the incident.