Skip to main content

Identify Repeated Issues

This action analyzes a given set of tickets to identify patterns of repeated issues. It helps in proactively recognizing and addressing common problems that affect specific companies or occur frequently overall.

Why is it useful?

Automating the identification of repeated issues allows MSPs to spot trends, address root causes more effectively, and potentially reduce recurring ticket volume. This leads to improved service quality and operational efficiency.

info

This action is available only as a Scheduled action in a workflow. Let us know if you'd like to use it as a Triggered one.

What It Does​

  • Groups incoming tickets by the company they belong to.
  • For each company, it uses an AI-driven clustering mechanism (HDBSCAN) to find groups of similar tickets based on their content (specifically, AI-generated issue summaries and embeddings).
  • Analyzes these clusters (considering only groups with a minimum of 3 tickets) using another AI model to determine if the cluster represents a genuine recurring issue.
  • Generates a detailed report summarizing the findings. If repeated issues are found, the report details them per company, including the reasoning for identifying an issue as repeated, a summary of the common issue, and a list of the affected tickets.

How It Works​

1. Input Data​

The action takes a list of PsaTicket objects as input (default_data).

2. Company-Level Grouping​

The input tickets are first grouped by their associated company name. Tickets without a company are grouped under "Unknown".

3. Per-Company Analysis​

For each company's set of tickets:

  • AI Clustering:
    • Neo retrieves AI-generated issue embeddings for each ticket.
    • It then applies the HDBSCAN (Hierarchical Density-Based Spatial Clustering of Applications with Noise) algorithm to these embeddings. This algorithm groups tickets with similar issue embeddings together. The process involves:
      • Preparing embeddings by calculating pairwise cosine distances.
      • Auto-tuning HDBSCAN parameters (min_cluster_size, min_samples, cluster_selection_epsilon) to find an optimal clustering configuration. The system aims for a balance between meaningful clusters and acceptable noise levels (tickets that don't fit any cluster).
    • Tickets without embeddings are initially grouped as "No issue identified".
  • Recurring Issue Identification:
    • Only clusters containing at least 3 tickets are considered for further analysis.
    • For each such cluster, an AI model analyzes the ticket group to understand the root cause and determine if it represents a genuine repeated issue. This involves generating a group name and an issues summary for the cluster.
  • Report Generation (per company):
    • If recurring issues are identified for a company, a human-friendly report section is generated. This includes:
      • The group name and number of tickets in the group.
      • AI-generated reasoning for why the group is considered a recurring issue.
      • An AI-generated summary of the common issues within the group.
      • A list of tickets belonging to the group, with links to each ticket and their titles.

4. Final Report Aggregation​

The reports from each company (if any repeated issues were found) are aggregated into a single output message. If no repeated issues are found across any companies, a message indicating this is produced.

Configuration Fields​

When you add the "Identify Repeated Issues" action to a workflow, you'll configure the following:

report_format​

Specifies the format of the generated report. (default: V2_SINGLE_MESSAGE)

  • V2_SINGLE_MESSAGE: Generates a single consolidated message containing the report for all companies.

Output Fields​

This action makes the following information available to subsequent actions in the workflow (via SimpleNotificationOutput model):

default_messages_for_dashboard​

A list of strings. Contains the detailed report of identified repeated issues, formatted for display in the Neo Dashboard Event History. If no issues are found, it will contain a message like "No reoccuring issues found.".

default_messages_for_msp​

A list of WorkflowMessages objects. Contains the same report as default_messages_for_dashboard, intended for internal MSP use (e.g., to be sent via notification or added to a summary ticket).

credits_consumed​

An integer indicating the number of credits consumed by the action, which is equivalent to the number of input tickets processed.

Use Cases​

Proactive Problem Management​

  • Scenario: An MSP wants to identify and address underlying issues that cause multiple tickets for their clients.
  • Workflow:
    1. Action: Find Entities (Scheduled daily/weekly to gather recently closed or updated tickets).
    2. Action: Identify Repeated Issues.
    3. Action: Notify Internal Team (e.g., problem management team or service desk manager) with the report from default_messages_for_msp.

Monthly Client Health Check​

  • Scenario: As part of a monthly review, an MSP wants to understand recurring problems for specific key clients.
  • Workflow:
    1. Action: Find Entities (Scheduled monthly, filtered for tickets from specific client companies).
    2. Action: Identify Repeated Issues.
    3. Action: Build Message (to format the output for a client-facing summary, perhaps focusing on the issues_summary and number of tickets).
    4. Action: Notify Users (e.g., Account Manager) or Add Ticket Note to a summary/reporting ticket.

Best Practices​

  • Schedule Appropriately: Running this action too frequently on a large volume of tickets might be resource-intensive. Consider weekly or monthly schedules, or run it on tickets from specific timeframes (e.g., "last 7 days").
  • Filter Input Tickets: Use the Find Entities action effectively to narrow down the tickets being analyzed. For example, focus on tickets closed within a certain period, or tickets of a specific type/priority that are more likely to indicate underlying problems.
  • Review Report Actionability: The quality of "AI Issue" summaries on tickets (if that's what the embeddings are based on) will influence clustering. Ensure tickets have good, concise summaries or problem descriptions.
  • Minimum Group Size: The action internally uses a minimum group size of 3 tickets for a cluster to be considered for reoccurring issue analysis. This helps filter out very small, potentially coincidental groupings.
  • Monitor Event History: Regularly check the Event History to see the reports generated. This will help you understand the types of issues being flagged and refine your input ticket criteria if needed.
  • Combine with Other Actions: Use the output of this action to drive further automation or investigation. For example, create tasks in a project management tool for each identified repeated issue, or generate more detailed analyses on the specific ticket groups.