Kehat Shahar, the Vice President of Supply Chain Planning at SanDisk, spoke on the journey that SanDisk has taken to improve their supply chain operations at the eft conference on June 18th in Chicago. While most of the content for this article came from the speech at eft, some information came from the company’s 10-K, and some from an interview Mr. Shahar gave at a JDA conference that is posted on LinkedIn.
SanDisk is a leader in flash storage solutions. This $6.6 billion company garners about two thirds of their sales from commercial clients, all the smart phone companies buy their flash drives. The other third of their products are sold through retail chains.
SanDisk has invested heavily in a vertically integrated business model, which includes a high volume, manufacturing facility in Japan based upon a joint venture with Toshiba. This supply is “captive” – it is fully controlled by SanDisk. They also purchase non-captive flash memory from other suppliers.
The wafers are sorted and tested at captive and third-party facilities in China, Japan and Taiwan. Their products are then assembled and tested at both their in-house facilities in Shanghai and through a network of contract manufacturers. Finally, the millions of units they produce are shipped out of regional distribution centers run by 3PL partners.
Mr. Shahar said that SanDisk’s journey to substantially improve their supply chain operations began in 2008. At that time, they primarily served the retail channel, including big retailers like Walmart and Best Buy. In 2008, they were a build to forecast company with a frozen forecast period of six weeks. But their forecast accuracy was poor, only 50-60 percent accurate. Consequently, they had too much inventory and poor service levels. They knew they had to change.
They did not believe they could sufficiently improve forecasting to solve their problems, so they focused on responding to “certain demand” – committed orders. This required manufacturing and replenishment agility. They moved from a build to forecast to a build to target process. This allowed them to reduce their frozen production fence for their captive WIP manufacturing to two times per week. To support these quicker planning cycles, SanDisk became more rigorous in measuring variability in supply, manufacturing quality, and demand, and then working to reduce that variability. Less variability allows for quicker lead times and the ability to hit service levels with lower inventory levels.
SanDisk also worked to delay product differentiation through postponement. They build generic inventory and delay creating the unique SKU until an order arrives. The company also continues to work to drive in more component commonality across their product lines. Agility is supported by air transport; their raw materials, work-in-process, and finished products are primarily shipped via air.
While their agility was greatly improved, they realized that different customer segments need different service levels. Big collaborative planning, forecasting and replenishment (CPFR) retail customers in the US require that they make inventory commitments at the store level. In emerging markets, customers want a low price and quick delivery; for some SKUs they need be able to deliver within 24 hours.
Further, in 2009-10, their focus on the commercial business greatly increased. These are primarily vendor managed inventory relationships; SanDisk’s goal is to be the number one or two supplier on these customer’s scorecards.
All of this planning was, and continues to be, facilitated by supply chain planning tools that recommends what to build on a daily basis. They use an advanced planning system (APS) and a multi-echelon inventory optimization (MEIO) solution from JDA to support this. The APS starts by planning for 3 weeks of inventory to be held by their DCs. Then the IO solution looks at the supply and demand inventory variability and the lead times they have experienced over the past 13 weeks, and adjusts the inventory targets accordingly.
But optimization is not enough. They have set up an analytics center of excellence staffed by two PhDs in operations research. This team takes a data driven approach to looking at their network flows and supply chain policies. As one example, this team looked at how many factories should be enabled to manufacture particular products. Previously, they had assumed the more products that could be made at more factories, the better their service levels would be. But they discovered that enabling just two factories to produce most products was sufficient from a service level perspective and also much less costly. Optimization is great, but optimizing an inefficient network is itself not optimal.
To support customer segmentation, in the last one to two years SanDisk began pricing service levels and value added services into their customer contracts. Their goal is to maximize the dollars per gigabyte. Higher service levels typically means more inventory and thus higher costs.
In emerging markets, some customers buy high volumes of only a few key SKUs. Despite fast turnaround requirements, this ordering pattern reduces forecast variability and the inventory they must hold. For some customers they operate virtual hubs – manufacturing capacity that is allocated to the customer with their resulting inventory positioned in one central location; for some customers where they receive sufficient lead time, they will engage in build to order; for many customers it is a configure to order process. Each of these processes has a different cost structure that must be priced into the customer contract.
Mr. Shahar had a few key pieces of advice for companies that might want to develop a segmented supply chain. First, their SCP system is a key enabler of the segmented supply chain. The software manages the eight different supply chains they run, but all of this is invisible to the factory. The factories are just told what they need to produce and when.
Surprisingly, getting buy in for a cost to serve segmented supply chain model was not difficult. Mr. Shahar did not explain why this was so, but in a different part of the speech he mentioned that SanDisk has also improved their sales and operations planning process by tightly integrating it with the financial planning process. This could be part of the explanation for why what has been a very difficult cultural transformation at many other companies, was not so difficult for SanDisk.
Finally, SanDisk did not take a “big bang” approach. As this narrative makes clear, these capabilities were built step by step.
Nevertheless, it is clear that since 2008 when SanDisk began this journey, they have made substantial progress. They have improved on-time deliveries from 50-60 percent to best in class. And these service levels are achieved with less inventory. Their key measure on inventory is inventory turns above die bank. Based on this metric, they have reduced their inventory by almost half.