SKU optimization: How to improve your supply chains in COVID-19
- COVID-19 has wreaked havoc on supply chains
- SKU proliferation is a key problem
- Companies can leverage machine learning to focus their SKU portfolio and reduce complexity
- Microsoft’s Project Bonsai is especially useful for unprecedented events where historical data isn’t an option
A perfect storm: COVID-19 and SKU proliferation
COVID-19 has exposed a supply chain challenge that has existed for many years: SKU proliferation.
Even under normal circumstances, large SKU portfolios made it more difficult to forecast accurately, manage warehouse inventory levels, and control operational costs. As consumer goods companies have grown accustomed to providing consumers with more choices to increase market share, they have also created a perfect storm for an unprecedented and unpredictable event like a global pandemic.
Bottom line, the supply chain complexities associated with an overabundance of SKU’s are no match for the unpredictable demand spikes experienced with COVID-related pantry loading and shopping frenzies.
How can businesses strengthen their supply chains in a pandemic?
For many CPG companies, the answer is to keep it simple.
Maintaining a large SKU portfolio with hundreds or thousands of SKU’s is no longer economically viable. Think about it: When millions of Americans rushed to the stores to stock up on toilet paper in early March, would it have made any difference if there were only one type and package size on the shelf? If the manufacturers were focused only on delivering 20-packs of single-ply toilet paper in the standard role, perhaps there would have been fewer occurrences of the empty store shelves that were so often present in the news.
Keeping it simple means reducing the number of overall SKU’s and focusing on delivering core SKU’s. This results in a more efficient supply chain by reducing the variety of materials needed while also streamlining operations for retail customers.
While the ROI on the “long tail” of SKU’s has always been questionable, very few CPG companies have been successful at implementing SKU rationalization tactics that provide an in-depth understanding of SKU performance (including cannibalization effects). This is especially true when factors such as margin, market share, and product attributes are considered.
Fewer SKU’s leads to less complexity, lower costs, and more accurate forecasts. But how do you optimize your SKU portfolio for an unprecedented and unpredictable environment?
Leveraging machine learning to optimize your SKU assortments
Neal Analytics has worked with several well-known CPG companies to help them rationalize and optimize their SKU portfolios with the advanced machine-learning models found in our SKU Assortment Optimization solution.
In addition to helping businesses understand which SKU’s should remain in their portfolio (and why), SKU Assortment Optimization can target specific retail points of sale. By accounting for localized demand patterns and demand drivers, CPG companies can dynamically manage their inventory to realize specific business objectives such as volume, margin, and market share.
We are particularly excited about some of the new innovations enabled by Microsoft R&D and their Project Bonsai Deep Reinforcement Learning toolset from the Autonomous Systems team. With reinforcement learning, it’s now possible to overcome one of the biggest SKU optimization challenges: the shortage of historical data needed to train effective machine learning models.
Training the Project Bonsai “brain”
Traditional machine learning approaches to SKU optimization require significant amounts of SKU-level transaction data (typically a minimum of 2-3 years) in order to develop a model that can effectively explain and predict SKU performance.
This is especially challenging when rationalizing SKUs in the “long tail” where transaction volumes are much smaller. Reinforcement learning overcomes this challenge by continuously learning how SKU performance and assortment decisions are impacting business objectives. You can think of it as A/B testing on steroids.
Leveraging Microsoft Project Bonsai, we’re able to teach a “brain”, the AI engine Bonsai generates, to rationalize SKU assortments for the desired outcome. When the objective is achieved, the “brain” is rewarded for its decision, thus reinforcing the action and teaching it how it should behave in the future to best meet those goals.
The real beauty of this approach is that you can teach a “brain” without mountains of historical data by using a simulator. This not only solves the problem of data sparsity but also accelerates the teaching process since you can feed millions or billions of simulated transactions into the “brain” via the simulator.
Microsoft’s Project Bonsai and deep reinforcement learning are exactly what retail and CPG companies need right now. Historical sales data and demand patterns are no longer accurate predictors of future behavior.
What’s really exciting about Microsoft Project Bonsai and deep reinforcement learning, is that this is exactly what’s needed in times like these. Historical sales data and demand patterns are no longer accurate predictors of future behavior. Businesses will need to streamline their supply chains and optimize their inventory to dynamically respond to unexpected demand and disruption through this pandemic.
Another arrow in the SKU assortment optimization quiver
While traditional machine learning approaches are still valuable and important to solving SKU Rationalization/Assortment Optimization scenarios, these unpredictable times require an approach that combines these approaches with one the continuously learns and optimizes as behaviors and environments change.
This article was also published on LinkedIn.