Solving the Complexity of Demand Management

Supply chain planning is a continuous business process that enables firms from electronics OEMs to hospitals to chemical makers to respond to emerging situations in an intelligent manner in order to maintain a balance between demand and supply.

In the end, intelligent response is achieved by collaboration and alignment throughout the organization supported by software where there are five major activities: demand management, supply planning, inventory management, available to promise, and ongoing analytics. Our focus is a brief review of the core components of best practices in demand management (DM).

Easier said than done

In 1994, the IBM Microelectronics Division, itself a fortune 100 size firm, put in place a major effort to create best in class supply chain planning process for semiconductor based packaged goods (SBPG) used in computers, cell phones, cameras, etc. A critical component of this work the development was a suite of software application by the IBM PROFIT team. I was fortunate to be an original member.

During the first two years, I worked on a complex model with optimization to match assets with demand and then assigned to an effort for a best in class DM. With 20 years of experience in computational methods. I thought DM would be a walk in the park. I quickly learned it was illusively complex. 

Questions to start

To successfully navigate demand management, it's important to start out by asking the right questions:

  1. Which type of demand? Exit demand for finished products from large suppliers either sold to consumers, component parts for another manufacturer, or spare part replenishment. This is in contrast to “exploded demand,” strategic parts, or parts made strictly to order.
  2. What is the purpose? Understand demand for each product to support three business functions: financial commitments; production and capacity decisions, and inventory replenishment.
  3. Understand what? Characterization of demand for each product or logical grouping such as quantity, due date, customer, location, importance, and “type” (firm demand, contractual commitment, build to forecast). Advanced characteristics include what influences demand, how demand between parts is related, segmentation (low/high volume; low/high variability), and demand probability distribution.

Core components of demand management

Next create strong demand management ability by following a handful of critical steps:

1.     Gather data and institutionalize it in a single source, the DM repository.

Start by determining what data is available and establishing a method to securely capture and store it in a repository. Typically, this includes current orders, shipment history, prior forecasts, current estimates, etc. Often the data initially available supports transactions or accounting; not DM. Therefore, the DM application will drive the improvement in quality of data already captured (example exact dates) and capturing new data (example order history). Knowing what data you need requires knowing what methods you will use to create demand estimates, this drives an iterative process.  It is critical to ensure everyone in the organization uses this common source.

2.     Capture information from the sales force.  

Technically this is part of “gather your data,” but is different in nature than most supply chain data, hence its own listing. This component enables the sales force to input and review its estimates of demand in a controlled process at different levels of aggregation. 

3.     Manipulate, mold, and analyze (MMA).  

With the repository in place, it is critical to have software that enables all participants to access the data they are entitled to see and manipulate as well as mold it as desired.Typical MMA activities include filters, dynamic hierarchies, aggregation/disaggregation, tree structures, slices, drill downs, comparisons, tracking changes over time (orders moved in or been pushed out), etc. These must be seamlessly integrated with analytics and visualization that supports department preferences.

 4.     Develop and implement a set of methods to create a demand estimate or forecast. 

For simplicity we will assume the end result is a “point” estimate – for each product or logical grouping a single estimated value is agreed upon for each time period (day, week, monthly).  In a future column, we will address variability and certainty. All quality forecasts require the successful collaboration between individuals with their varying expertise and analytical methods.  Management must nourish collaboration, each member (including the math) has its role, and implements a forecast value add process (assess how each step improves the quality of the estimate). Without the right software to support collaboration, real collaboration is dead on arrival.  The dominant factor becomes the mechanics to get some answer to management by “Tuesday.”

5.     Put in a mechanism for ongoing evaluation and alerts

Typically a complete new forecast is done once a month.  In between, it is critical there is a monitoring and alert process in place.

Lessons learned

In looking at the process of demand management, some general ideas become abundantly clear:

  1. It is critical to consistently measure forecast accuracy and adjust.
  2. Begin with a clear understanding of your demand streams: historical patterns; differences between customers or products; geographic patterns etc. In the early stages, avoid evaluating how much demand can be met through marketing programs.
  3. Understand the difference between precision and accuracy.
  4. Use the forecasting methods appropriate for each product group (or to put it in math talk – segment).  Sometimes to drive consistency, a business will use the same forecasting method for every item – this is a recipe for disaster.
  5. Isolate the cause of variability. For example, the product level forecast may be highly variable; breaking out the data to ship-to level may indicate the variability is limited to a few customers.
  6. Creating a quality demand estimate is hard work! 
  7. Working with a firm experienced in the software and creating the methods can be of immense help, if knowledge is transferred to your people.

Successful interplay

A quality DM application enables the successful interplay between human expertise and analytics in order to generate smarter solutions faster. Let us know your DM successes or horror stories in the comments section below.

0 comments on “Solving the Complexity of Demand Management

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.