descriptive analytics supply chain

In this guest blog series titled: “Memoirs of a Black Belt,” Stephen Boyd a Lean Six Sigma Black Belt and 30-year supply chain veteran, shares his insights on achieving higher levels of performance using data from existing systems. Opinions expressed by guest authors may not reflect Arkieva’s view on the subject. To contribute to the Supply Chain Link Blog email:

How do you modify your planning systems to show what’s necessary and sufficient?

Do you recognize this title from one of Eli Goldratt’s business novels?

“Necessary but Not Sufficient,” A Theory of Constraints Business Novel by Eliyahu M. Goldratt with Eli Schragenheim and Carol Ptak, The North River Press, 2000 ISBN: 0-88427-170-6

My take away from reading this 14 years ago came out crystal clear. Software systems provide the main highways for planning and execution; however, they require constant adjustments to fit the ever-changing business requirements. These ever-changing business requirements come at all levels from dated activity: e.g., shipments (quantity & date) to planning data (e.g., lead time) to descriptive statistics (e.g., the coefficient of variation or COV). It is these descriptive statistics that allow visibility into the world of process variability and finished product velocity. Since each business is unique, there is a unique set of ever-changing descriptive statistics for each business. Quantifying these descriptive statistics and analyzing them to guide a business strategy yields unique opportunities. Instead of becoming overwhelmed by information, the business may now compete on data.

[Read more: How Do You Create a Consistent Data Basis for Planning?]  

Today’s supply chain systems contain a wealth of data for mining. Many software systems provide insight with canned reports or dashboards to alert managers of potential problems. With the advent of desktop computing, modern-day managers can create their reports to keep up with the needs of an ever-changing business environment, using data from the system. However, there needs to be a guiding hand to direct strategy for current conditions for which software Developers can only guess at. Consider ten data elements taken three at a time for a graph or chart. To cover all the combinations, the Developer would have to make 120 charts or graphs. Since this won’t happen, you need insider knowledge such as a Six Sigma Black Belt from the business, working with a Developer to deliver the one-two punch:

  • Develop descriptive statistics for visibility of variability, velocity, volume, and variety
  • Apply these descriptive statistics to key process variables of the business (KPV)

So where do you start? Here’s a brief example of a make to stock business with customer order lead times less than the total lead time to plan-make-deliver.

[Read more: 5 Fundamentals for Building a Collaborative Supply Chain]  

Descriptive Statistics Analysis Examples:  

  • Key customers – Create a list from Sales & Marketing complete with codes from the system, and use a simple ABC segmentation
  • Key Products (segmentation) – Identify the key products by product family using a simple ABC segmentation approach
  • Customer-Product Matrix – An ‘A’ customer buying a ‘C’ product could create a ‘B’ customer product. Create a realistic matrix to identify Customer-Product combinations.
  • COV (coefficient of variation) – Use last 12 months of shipments and calculate the standard deviation divided by the average demand. When the standard deviation (variability) is half the average demand, COV = 0.5. This could be a threshold for variability of the business. You decide. Also decide at what level to measure this: Item-Location (SKU) or Product Family, or both.
  • ADI (average demand interval) – This measures velocity when you divide 12 by the number of months shipped. G., ADI of 4 means it moves every four months or 12/3 in a 12-month period.
  • Customers per SKU – Count the number of customers served by that product out of that warehouse. If a lot of customers buy a difficult to make product, you might want to review your strategy surrounding its sale.
  • Volume-Variability Matrix – Have a business leader tell you what they consider high volume by Product Family (kilos/month). Then have them do the same for high variability (COV). Plot the results at the SKU level and the Product Family level and label the quadrants.
  • Forecast error – Keep this handy at the SKU level as well as the Product Family level
[Read more: Learn How to Improve Your Statistical Forecast Technique]  

Applying Descriptive Analytics to Improve the Strategy

  • Product Pruning – Use the COV, ADI, and quadrants of the Volume-Variability Matrix to create a list of “Good SKU-Bad SKU.” Rid the business of Bad SKUs that do not deliver.
  • Manufacturing Strategy – For High Volume/Low Variability, consider LEAN. For High Volume/High Variability use flexible manufacturing or collaborative planning with customers. Limit losses with Low Volume/High Variability products with sales policies. Use safety stock for Low Volume/Low Variability products. Note also the manufacturing strategy links with Sales/Marketing strategy.
  • Safety Stocks – Adjust levels of safety stock based upon the above descriptive statistics. SKUs serving ‘A’ customer-product combinations with high volume and high velocity should take precedence over low volume/high variability products. Using the visibility of these descriptive statistics makes defending safety stock levels much easier.
  • Alert Reporting – Use conditional logic to create alerts for stock-outs. A Make-To-Stock SKU with an on-hand inventory of zero, in-transit inventory of zero and open customer orders > 0 would be nice to know for a Planner. The more the Planner knows about the condition of the SKU, the easier their job to optimize customer service and focus on the vital few.

Now that we’ve developed a brief understanding of descriptive statistics and their applications to the business, we can see how your planning systems can be modified to show what is Necessary and Sufficient. We’ll continue this discussion next time by looking at “who owns the finished product inventory.”