"No one is harder on a talented person than the person themselves" - Linda Wilkinson ; "Trust your guts and don't follow the herd" ; "Validate direction not destination" ;

April 21, 2021

Inventory Optimization

Key Notes
  • o9 AI Platform
  • Integrated Business Planning, In memory codebase platform on graph model
  • Supply Chain planning 
  • Enterprise Knowledge Graph
  • Forecast demand, lead time, demand and supply uncertainty, right inventory levels
  • Right size + right mix of inventory to be positioned across right locations
  • Inventory Optimization terms





  • Initial Forecast + Safety Stock - Weekly / Daily 
  • Recommended Safety Stock + Adjusted Levels
  • Factors affecting Optimization




  • Supply Chain Network
  • Multiechelon optimization 
  • Recommendation with holistic view of network

  • Inventory target across supply chain
  • Dependent demand calculation
  • Measure historical actuals and deviations
  • Propagates forecast upstream / Forecast waterfall

  • Lagged forecasts reconciled with actuals
  • Lag correspond to lead times
  • 3 week / 2 week lag


  • ML to arrive at actual lead time

  • Turn around time computation
  • Lead time variability 


Service Levels
  • Classify portfolio
  • Segmentation
  • Service Levels per customer
  • Margin Revenue / Volume
  • Lead time variability







Product Features
CIO view


Planner view


Sensitivity Analysis
  • 95% of baseline forecast
  • 105% of the baseline forecast
  • Scenario Analysis


Multi echelon optimization
  • Cascade at the entire network level
  • Check baseline / nostock scenarios



Feature Variables
  • Understand the influencing factors
  • Promotions
  • Events
  • Historical Data
  • Weekend
  • Temperature / Weather impact
  • Consumer confidence
  • Market Size / Market Share
Long Term / Short Term Forecasting
  • 52-week horizon
  • Short horizon (3 weeks)
  • Holiday seasons
Forecasting 7 pointers
  • Add a safety stock value. Total forecast = Forecast + Safestock value
  • Lag time is identified, Forecast updated between 2-3 weeks, This is to get more accurate numbers as the lag time closes in. 3 months back value vs current forecast value
  • Cluster into different products based on different lead time and then forecast
  • Provide a dashboard for user to pick low range, medium, high range estimate as a graph
  • Forecast should be from bottom up, DC to enterprise level. Each dc will forecast based on its location, weather, historical data
  • The amount of money required to store items is also critical, Adjustments in stock need to be made to accomodate current safety stock quantity
  • Feature variables for forecasting can be provided as configurable / Additional parameters Understand the influencing factors
    • Promotions
    • Events
    • Historical Data
    • Weekend
    • Temperature / Weather impact
    • Consumer confidence
    • Market Size / Market Share
Keep Thinking!!!!

No comments: