ML Works standardized a CPG firm’s ML pipeline, reducing production downtime

Jack joined a Fortune 500 CPG firm to lead its revenue growth management (RGM) initiatives. The business covers 13 countries/regions and has 20 retail stores with more than 1,000 SKUs. When reviewing the existing process, he was alarmed that it would take more than a week for service engineers to fix bugs as over 100,000 ML models were used in the RGM planning.

Jack found out that with no coding standards:

  • The team has to go through millions of lines of code to identify bugs.
  • The team was deprived of using pre-existing assets for developing new models.
  • Each KPI had multiple definitions, impacting revenue visibility.
  • The company did not have access to the model output for ~25% of the time.

ML Works laid the foundation for ML pipeline/coding

ML Works’ visual provenance graph helped the service engineers accelerate error detection, perform root cause analysis, thus, putting a method to the madness.

Here’s how:

  • ML Works ingested existing code, standardizing it in the process.
  • ML Works provided templatized notebooks for future pipelines/codes, bringing uniformity across the organization.

Measurable benefits

  • Production downtime reduced by >95%.
  • Availability of insights increased by 25% per month.