Optimize logistics & supply chain for productivity planning using ML

5 Minutes Read

BACKGROUND

A leading fashion establishment wanted to maintain cost control on the supply chain and logistics of their retail merchandise. They needed a solution that optimized the monetary cost associated with trucks arriving at the store with less than the actual packaged units that were forecasted.

The goal was to achieve a productivity planning dashboard that presented an overall 360 degree view into their supply chain and logistics management

CHALLENGE

  1. Cost Control: Rising energy/fuel and freight costs, large global customer base, new regulations & technologies, increasing labor rates, and rising commodity prices mean that operating costs are under extreme pressure. As a result money is lost when truck arrives to the stores with actual units less than the forecast units.
  2. Central Buying: Keep track of the Purchase Orders (PO) or product group level quantities placed to vendor (as an aggregate of identified stores). Data will be used to compare the demand upon which POs are based v/s actual delivery at Distribution Center (DC).
  3. Supply Chain: Store level distribution (store level demand and inventory on and) of the quantity by Product group level received from the vendor.
  4. Distribution Center: Utilize the volume shipped from the distribution center to the stores to align with allocated and received numbers.
  5. Shipping (Trucks): Measure the actual shipped or delivered quantity details from the DC to the stores.
  6. Cost forecasting methods: Measure the impact of cost of labor via current forecasting methods compared to derived labor from the PoC.

SOLUTION

Develop a Machine learning based model that utilized historical data to more accurately forecast/predict units in order to improve associate planning and save money. The model also used data based on projected demand or marketing event for a given day, week or month.

Technology used: Machine Learning Algorithms

  1. Timeseries forecasting using
  2. Mean of historical data
  3. Naïve method,
  4. ARIMA
  5. ARIMAX
  6. Regression:
  7. Multivariate regression
  8. Random forest
  9. Extreme Gradient Boosting
  • Obtain data from EDW and multiple data sources that are stored as cloud SQL data
  • Use cloud ML algorithms on the processed data to provide insights that can be used for further analysis

Order and shipment analysis:

Based on historical data & root cause analysis to develop an ML-based model that accurately predicts the number of packages per purchase order.

Analytics & reporting:

Data analytics applied on Big Data to produce reports or data sheets for sharing.

OUTCOME

The client was able to optimize their shipping and logistics costs by accurately predicting the number of packages per order that arrived at the stores.

SUCCESS METRICS USED:

  1. Raw model accuracy measured % unit variance to actuals
  2. Keep associate hours on target (±12 hours)
  3. Labor hours to labor dollars (Prevent underfund or overfund)

ABOUT KLOUD9

Kloud9 helps retailers build an ecosystem on Data Science, Machine Learning and Cloud. We provide a full suite of cloud services along with Data Science & AI capabilities. Enterprise clients depend on us for customized cloud application migration and managed services to accelerate their performance with retail science.

As your trusted advisors in the transformation we are determined to build a deep partnership along the way. Our cloud and retail experts will ease your transition to the cloud.

Kloud9 gives you the insight into what your business needs and requires

Kloud9 gives you the insight into what your business needs and requires