I’ve just completed a global market study on supply chain planning. That gave me a chance to talk to all sorts of interesting people. For example, I talked to a member of the Executive Board of an ecommerce retailer in an emerging market. The executive did not want his company named because he did not want his direct competitors to copy what they were doing. But he did agree to talk because he wanted to support Solvoyo, the supplier of the supply chain planning solution his company implemented. As intriguing as the benefits this ecommerce retailer achieved from the planning solution, is the ongoing journey they are on to leverage artificial intelligence to deliver even greater competitive differentiation.
The company itself has a business model much like Amazon’s. They sell goods they buy and stock in their warehouses. They also sell items from third parties stocked at their partners’ warehouses. They sell across many product categories and have millions of stock keeping units (SKUs). Finally, they are the leading ecommerce retailer in their region.
This executive explained to me that emerging markets have their own nuances. Culture influences how things get done. Further, supply chains in emerging markets can be unreliable. Lead times are longer and more variable, and the difference between the delivery times for inventory they stock versus what can be delivered by third party partners can vary greatly.
The goal of working with Solvoyo was to achieve better service through better replenishment. But they did not want to achieve improved service at the expense of carrying too much inventory. “Cash flow in ecommerce is extremely important, the more free cash we generate, the better we operate. It is extremely important for this type of business.”
The Solvoyo platform is integrated with their internal enterprise system, their point of sale data, and has access to the inventory committed by their partners to support their sales.
Solvoyo’s solution provides multi-echelon inventory optimization (MEIO). In other words, it provides prescriptive optimization techniques for adjusting inventory levels at different points in the supply chain. Using MEIO, the company has achieved 95% in stock for the SKUs being managed by the platform. This includes the great majority of the items they stock, as we as some of the high velocity SKUs offered by their partners.
But improvement in their in-stock position has not come at the expense of high inventory levels. They improved their inventory turns by roughly 35%. In cases where Solvoyo is used to manage replenishment and procurement, 95% of the time it has led to improved cash flow based on having the right inventory at the right location at the right time. This has allowed the company to improve cash flow by 35-40%. “The savings run into the tens of millions of dollars.” In contrast, “when we buy outside the system, our free cash flow gets worse.”
When I mentioned that many companies offer MEIO solutions and asked whey they chose to work with Solvoyo, the executive mentioned a few different things. “The solution offers deeper analysis.” Not just traditional calculations based on lead times, but it also identifies lost sales by identifying those sales that were lost because the company did not have the right inventory in stock. They also use the system to look at SKUs offered by third party vendors which it would make financial sense for the retailer to stock. “Items with sales spikes are not worth holding, but if there is a long-term sales trend, we should hold it.” Based on the sales analysis, the platform drives procurement decisions.
This retailer got great results from the Solvoyo supply chain planning solution, but that is not the whole story. About 9 months ago they started a new project with Solvoyo to leverage a wider variety of data sources and test whether these data sets improve demand forecasting and replenishment. Those new sources of data include searches on their web site, searches on third party sites like Google, syndicated data from vendors like Nielsen, and social media data related to the comments made about products they carry. This data will be analyzed using a form of artificial intelligence – neural networks. Then the forecasts come back into the platform and their accuracy measured. This feedback loop supports improved forecasting through machine learning.
Much of the social and search data is analyzed by humans before it is put in the platform. They are not using algorithmic sentiment analysis for this. And they are using the 20/80 rule in determining what gets analyzed. Not all categories will be analyzed initially. And comments that are hard to interpret will be excluded. Finally, the human analysis is initially focused on whether the comments are positive or negative, with no finer nuances being coded.
The goal of this project is to support their procurement teams by doing forecasts on what inventory to hold 6 to 12 months out. These forecasts will also help them do better budgeting on their inventory spend. Further, the company does buy many products from China. The longer forecast horizons also help to plan around the longer lead times associated with global trade. And this helps to improve customer service.
While they have been at it for 9 months, it is still a “little early to talk results.” There were some external economic factors that skewed the demand trend lines. They believe they need another 3 to 6 months of data and active use of the platform to discover how much the solution will improve their category planning.
Even though it is preliminary, I asked if any of the new data sources were particularly promising. The executive answered, “I cannot say that any one source is more important than another but putting them all together adds predictive value.”
I asked the Board member if he had any final thoughts. “It is very simple, I’m doing this and you (my competitor) are not.” They have already used the existing supply chain planning system to improve predictions. “But we still have a lot to learn. We still have lots of changes that we must make.” The AI project is teaching the company how to use new data sets, and how add more data over time. “We believe we will have better optimization with AI than without it.”
Finally, “data is king. We aim to understand the data that underlies customer behavior better than our competition does. We can start small, but over time that can lead to a big competitive difference.”