How ERP Semantic Mining determines candidate score and potential

  • Release version: Yokohama
  • Updated January 30, 2025
  • 2 minutes to read
    • Summarize
    Summarized using AI
    This content was generated using new OpenAI-powered functionality. Results are provided on an as is basis and are not guaranteed to be accurate or complete.

    Summary of How ERP Semantic Mining Determines Candidate Score and Potential

    ERP Semantic Mining (ERP-CM) ranks legacy ERP system candidates for replatforming onto the ServiceNow AI Platform by generating a potential score. This score helps ServiceNow customers assess how well an ERP candidate can be migrated using existing ERP models and data extraction methods configured within the platform.

    Show full answer Show less

    Candidate Evaluation Process

    Each ERP candidate is linked to a specific ERP module, which is central to evaluating its replatforming potential. The evaluation considers the candidate’s usage of remote tables—populated via scripts from external data sources—and extraction tables—populated via scheduled queries and processed with transform tables within the ServiceNow AI Platform.

    Administrators must first configure ERP system connections using the Zero Copy Connector for ERP to enable this evaluation.

    Understanding Candidate Potential Scores

    The potential score reflects how compatible a candidate’s tables and fields are with the ServiceNow AI Platform:

    • High potential score: Indicates immediate usability of remote and extraction tables matching the ERP model without needing additional changes.
    • Low potential score: Indicates limited match with the available remote and extraction tables in ERP models.

    Score Calculation Metrics

    The potential score is derived from several key metrics:

    • ERP models: Counts how many ERP models the candidate utilizes through remote and extraction tables.
    • Similar candidates: Measures the number of candidates exceeding a similarity threshold based on table and model similarity; this threshold and logical conditions can be customized in the System Properties.
    • Supported table score: Ratio of custom application tables supported by any ERP model to the total custom app tables, excluding technical and ServiceNow cluster tables.
    • Supported table usage: Ratio of supported tables actively used by the custom app.
    • Unsupported model penalty: Penalizes candidates for unsupported operations on ERP model tables, scaled between 0 and 1 using a sigmoid function.
    • Unsupported table extensions: Counts custom app tables suggested as model extensions but not currently supported.
    • Model inaccuracy: Accounts for supported tables in ERP models not used by the custom app, adjusted via a sigmoid function.

    Practical Implications for ServiceNow Customers

    By understanding these scoring factors, ServiceNow customers can:

    • Identify high-potential ERP candidates that align well with the AI Platform’s existing models, facilitating a smoother replatforming process.
    • Recognize candidates requiring further customization due to lower compatibility scores.
    • Adjust similarity thresholds and scoring logic to fine-tune candidate evaluations based on organizational needs.
    • Ensure prerequisite setup of ERP system connections in Zero Copy Connector for ERP before leveraging ERP Semantic Mining.

    ERP Semantic Mining (ERP-CM) generates a score to rank the potential for replatforming legacy ERP (Enterprise Resource Planning) candidates onto the ServiceNow AI Platform.

    Every candidate has an ERP module specified in the candidate details in ERP-CM. The ERP module is used to evaluate the potential score of the candidate for replatforming, as well as the remote tables and extraction tables the model contains.
    • Remote tables get their records from running an associated script against an external data source.
    • Extraction tables retrieve large amounts of data using a scheduled query, and use transform tables to process data for use on the ServiceNow AI Platform.
    Note:

    Admins must first configure the connection to the ERP system in Zero Copy Connector for ERP. For more information, see Working with ERP systems in Zero Copy Connector for ERP.

    High and low scores for candidate potential

    ERP-CM evaluates candidates based on how well their tables and fields are supported by the ServiceNow AI Platform.
    • A high potential indicates that ERP-CM can immediately use remote tables and extraction tables that match the ERP model for the application candidate without making additional changes.
    • A low potential indicates that the application candidate matches few of the remote tables and extraction tables in the ERP models in Zero Copy Connector for ERP.

    How scores are calculated

    The candidate potential score is calculated using the following metrics:
    • ERP models: The number of ERP models that the candidate uses remote tables and extraction tables from.
    • Similar candidates: The number of candidates with a similarity score above the threshold, which accounts for both table-based similarity and model-based similarity.

      The threshold can be adjusted in the System Properties [sys_properties] table, and the default OR condition can be changed to AND.

    • Supported table score: The ratio of the number of custom app tables that are supported by any ERP model relative to the number of custom app tables.
      Note:
      Tables from either the Technical or ServiceNow table clusters are ignored from these computations.
    • Supported table usage: The ratio of tables supported by the relevant ERP models that are used by the custom app.
    • Unsupported model penalty: A penalty for the number of unsupported operations on tables in ERP models. The number of unsupported operations is passed through a sigmoid function, so it ranges from 0.0 and 1.0.
    • Unsupported table extensions: The number of custom app tables that are also suggested as model extensions.
    • Model inaccuracy: The number of tables supported by relevant ERP models that aren’t used by custom apps, and are passed through a sigmoid function.