Kristen Daihes
Aug 15th, 2016

iStock_73626421_XLARGE Frustrated Planner Web

When I was a planner more than 15 years ago for a consumer goods manufacturer, I approached each Monday morning with dread.  Not because I did not enjoy the company I was working for, rather I knew that I would be entering into a period of painstaking 5-Why research looking back a week to categorize the root cause for stock-outs in my portfolio.  It was dreadful because the critical hours being tied up looking back at what had happened could have been spent by preventing future problems.

One time, I pre-built inventory for hair styling aid products knowing a strategic retail customer was about to drop some large orders to support execution of an in-store promotion.  So why in the world did a portion of these orders go unfulfilled?  The customer must have dropped the orders across our DCs differently than the planned forecast.  And then the inventory being subsequently transshipped was late in arriving to the proper warehouse and not properly processed before the orders dropped.  Yes, that must have been what happened.

While this weekly exercise was critical to ensure adequate visibility into the key drivers of stock-outs and devise action plans forward, I would not bet money on the accuracy of all my root cause assignments.  Today I can tell you that not much has changed inside many companies.  It is time to stop the madness!

Over the last few months I have seen countless examples of value created in the retail industry by using advanced analytics to tackle stock-outs in stores.  You can leverage machine learning to create algorithms to identify 5-Why trends and use these algorithms to then assign the root cause for stock-outs.  Think about all those countless hours doing research when I could have had an algorithm do this work for me… and more accurately!  And in many instances these algorithms can be used in an ongoing cycle to recognize new patterns – leading to the discovery of chronic root causes that may have been overlooked or are new losses.

Remember those unfulfilled orders for the in-store promotion of my hair styling aid products?  It turns out the orders dropped exactly as planned, only I was not aware that we were gaining increased distribution for these products with other customers on the same timing.   Today, it can be possible for an algorithm to do a better job of determining root cause.

You do not have to be a data scientist to be playing in this space.  As a supply chain practitioner you play a critical role in determining the key questions that should be asked:

  • What drivers contribute to stock-outs?
  • Which drivers have the biggest impact on OOS, and therefore, improved in-stock?
  • Are there obvious or hidden patterns to the stock-outs? How can we leverage these patterns to make strategic changes to the plan?
  • Can we predict stock-outs and build operational tools like exception reports?

What excites me the most about the possibility in this work is the value creation potential this can drive in all nodes of a supply chain.

  • Accuracy. My 5-Why exercise was heavily influenced by known business rules that an algorithm may challenge, creating the context for new discovery.
  • Speed to Insight. We can leverage new technologies today to dramatically cut down the speed to insight, not to mention enabling more automation of the root cause assignment so planners can shift their focus to eradicating these losses.
  • Predictive Capability. Think about the possibility of these algorithms alerting a planner that a pattern has been detected (late shipment possibility) in time for preventative action to be taken.
  • Prescriptive Capability. Even better, think about an algorithm alerting the planner of the detection and recommending the best course of action to prevent the stock-out at minimum cost.   (Is there a benefit in shipping product from a different store or expediting replenishment from a DC out of territory?  Should I bump up the safety stock for the item in this store going forward?)

Advanced analytics used to be a scary term.  It is becoming comfortable for early adopters of experimentation.  “Why did X happen.  When will X happen again?  What caused X to happen?”  If you are not investing in advanced analytics to help you answer these questions, you are leaving yourself vulnerable to competition because machine learning algorithms will play a much larger role in the world of supply chain operations going forward.