Scoring Model Implementation Modes

  • Mode one
    Scoring model on our server

    Support

    • Retraining once every 6 months
    • Provide API and integration assistance if needed Payment

    Payment

    • Monthly payment for use – from 990 to 1 600 $ depending on the final functionality
    • Prior to the start of the work in this scenario we take a prepayment for the first 3 months
  • Mode two
    Scoring model on your server

    Support

    • Retraining once every 6 months (2 000 $)

    Price and payment terms

    • Price of the model from 18 000 to 40 000 $
    • Partial payment

In both variants, if necessary, we also make manual maps for underwriting (the final AUC cards due to simplification will be less due to simplification compared to ML).

Stages of work
  • 1
    NDA signing
  • 2
    Customer’s description of current data sources used for decision making, process descriptions
  • 3
    Harmonization of the format, quality, quantity of data that will enter the model
  • 4
    Customer’s description of the current data sources used for decision making, process descriptions
  • 5
    Sample design for model training
  • 6
    Data enrichment in FscoreLab (geo-transferable, etc.)
  • 7
    Sample generation for model testing (without target variable 20% of total sample. Loan repayment result is filled in Fskorlab and passed to the Customer for calculation of quality metrics)
  • 8
    Model training (1-2 months)
  • 9
    Test model with test sample transmission with filled target variable and multiple thresholds
  • 10
    Coordination of the results, revision of the model if necessary (1-2 weeks)
  • 11
    Model preparation for deployment and documentation (1 week)
  • 12
    Model integration, functional testing
  • 13
    Test operation
  • 14
    Product operation