Where to apply machine learning for supply chain optimization

Supply Chain
/
Where to apply machine learning for supply chain optimization

Artificial intelligence, specifically machine learning (ML), is quickly becoming essential for running smarter business operations. One of the greatest features of Dynamics 365 is its ability to incorporate ML capabilities within business applications, which provides predictive insights and helps businesses execute operations in more effective manner.

According to a recent study by Mckinsey Global Institute, advanced AI technologies have the potential to unlock a global economic impact of $10-15T across all industry segments.

Sales, marketing, supply chain management, and manufacturing are major segments that could significantly benefit from machine and deep learning technologies in retail and CPG. 

Where-to-apply-machine-learning-for-supply-chain-optimization-1


Below are a few candidate scenarios for AI-enabled optimization for the retail and CPG veticals in particular. Later in the article, one use case is explained in detail using Microsoft business applications.

  1. ML based demand and sales forecasting
  2. Personalized product recommendations
  3. Price and promotion recommendations to optimize markups and margins
  4. Inventory optimization with correct stock levels
  5. Logistics planning workbench and warehouse throughput optimization
  6. Build a 360° view of consumers
  7. Consumer insights (sentiment analysis/preferences/social listening) using cognitive services
  8. Shop-floor yield optimization
  9. Predictive equipment maintenance in factories
  10. Predictive lead scoring to improve lead qualification, prioritization, and acquisition

“61% of organizations picked machine learning as their company’s most significant data initiative for next year.” Source: Forbes.com

Dynamics 365 Operations and Azure Machine Learning Studio

Demand Forecasting Use Case

Dynamics 365 for Finance and Operations allows you to integrate Azure Machine Learning into your Dynamics environment to predict demand more accurately by infusing more demand planning parameters and considering new statistical models.

Where-to-apply-machine-learning-for-supply-chain-optimization-2

1) Historical data: The first and most important step of the process is gathering and preparing the transactional data from Dynamics 365 and providing it to Azure Machine Learning Studio for training the mode.

Where-to-apply-machine-learning-for-supply-chain-optimization-3

Item allocation keys are used to bundle similar products for which the demand-forecasting algorithm should run on historical sales.

Navigation to set up similar products in the historical data:

Master Planning -> Setup -> Demand Forecasting -> Item Allocation Keys

2) Train your model: Once historical data is loaded, a model needs to be trained for accurate forecasting. In this example, we have an R script, however, we can build forecasting models in Python as well. In this scenario, we are using a predefined model available in the Experiment Lab.

Where-to-apply-machine-learning-for-supply-chain-optimization-4

Is integration real-time between ML and ERP?

Since Dynamics 365 and Azure are both in the Microsoft family, they are easily integrated to allow for real-time results.

Where-to-apply-machine-learning-for-supply-chain-optimization-5

Navigation to enable this integration:

Master Planning -> Setup -> Demand Forecasting -> Demand Forecasting Parameters

On the Azure Machine Learning FastTab, provide the web service key and endpoint received from Machine Learning Studio.

3) Generate a statistical baseline forecast: After completing the setup and configuring the demand forecasting parameters, we generate a statistical baseline forecast.

Navigation to generate the statistical baseline forecast:

Master Planning -> Forecasting -> Demand Forecasting -> Statistical Baseline Forecast

Once you click OK, it looks for the best-fit model as per the forecasting parameters and generates the forecast from the ML engine.

4) Adjustment and approval: We can now adjust and approve the forecast as per business needs and including other factors like market volatility, etc.

Where-to-apply-machine-learning-for-supply-chain-optimization-6

Master Planning -> Forecasting -> Demand Forecasting -> Adjusted Demand Forecast

Once we authorize the demand forecast, we can run the master planning and create planned orders. Each planned order generated includes master planning parameters e.g. production process, minimum lead time, lowest unit price, and safety stock calculation based on the coverage group assigned to the item.

Hence, with a few steps, we were able to set up a predefined, AI-infused demand forecasting model in Dynamics 365 to generate forecasts.

With the Microsoft Dynamics 365 platform, all of this is can be customized to fulfill your business needs. Microsoft has provided numerous ML models under cognitive services that can be used as a foundation and further trained for customer-specific scenarios.

Visionet is a trusted Microsoft Gold partner and has 20+ years of experience delivering projects in retail and CPG. If you are looking to enable these AI initiatives for your organization, please contact Visionet today.

Adeel Ehsan
Adeel is a supply chain & distribution consultant based out of New Jersey. He has been the business and technical leader on many ERP, PLM, WMS, Retail and POS Global Supply Chain implementations over his 20+ year’s career. He has successfully delivered ERP projects for varying organization sizes across multiple industry verticals.