30% Efficiency Improvement: Overhauling the Model Point Grouping Process

The client requested an investigation into their current model point
(MP) grouping process, with a focus on:

• Efficiency and accuracy of the current grouping algorithm, with a
  view to reducing Prophet model run times through increasing MP
  compression whilst maintaining accuracy.
• Suitability of their current tools and applications used i.e. DCS and
  evaluation of alternatives.

Analyse

• Analysed distribution of key risk drivers and used statistical metrics
  to assess goodness-of-fit.
• Benchmarked against industry standards and SME knowledge used
  to identify opportunities for additional MP compression.
• Evaluated machine-learning approach – recommended to keep
  algorithmic approach for reliability & repeatability.

Build

• Developed a custom Python solution to produce current and further compressed MPs with an audit trail.
• Enhanced Prophet models using intelligent grouping to maintain
  accuracy.
• Performed extensive testing and built custom extraction files to
  confirm the accuracy of results.

Handover

• Comprehensive documentation on new grouping and MP methodology.
• Delivered extensive training and handover of new tool to end-users.
• Presented MP compression and accuracy improvements to senior
  stakeholders.

• Delivered a revised MP grouping process which achieved 30%
  efficiency improvement, whilst maintaining over 99% accuracy of
  Prophet results.
• Custom Python solution offers increased functionality and
  auditability, whilst also being aligned to the client’s long-term
  technology aims.
• Runtime of MP grouping also reduced in Python.
• Comprehensive documentation & handover of new tools to end-users.