OEMs are ready to embrace Lean Manufacturing after the 2001 recession, but traditional approaches were designed for vertically integrated enterprises. The answer to their problem? Extended Lean and Statistical Kanban.
The electronics manufacturing industry has changed dramatically during the past 20 years. Two decades ago, most computer and telecommunications original equipment manufacturers (OEMs) were vertically integrated companies with design, manufacturing, assembly, sales, marketing, service and repair under the same roof. A small community of outsourcers found work making and assembling printed circuit boards and other commodities for these industrial giants, often in small lots for prototype production runs.
Gradually, these tiny contract manufacturers (CMs) expanded their offerings to include more sophisticated multi-layer boards, flexible circuits, surface-mount packaging, and tape-automated bonding, while the OEMs concentrated on building complete systems.
By 1995, contract manufacturers like Solectron, Flextronics, Celestica (then an IBM subsidiary), Sanmina-SCI, and Jabil Circuits whose combined annual revenues were less than $2 billion began spreading their wings, taking on ever more sophisticated systems-level manufacturing, assembly and component sourcing. Meanwhile, the OEMs increasingly focused on their core competencies design, sales and marketing.
By 2000, CMs had given themselves a new name to reflect their elevated status Electronic Manufacturing Services and the transformation of the global electronics manufacturing industry was complete. In fewer than five years, the industry had evolved from an environment dominated by vertically integrated enterprises to an ecosystem of outsourced supply chains with EMS, OEM and tier-one component suppliers making up the three most important links.
Today, the Big Five EMS firms alone employ 270,000 people and have combined annual revenue exceeding $65 billion. And global revenue from all forms of electronics outsourcing will approach $500 billion this year, according to the Outsourcing Institute. But while the EMS and OEM industries have grown to enormous size (Cisco, Dell, Intel, HP and IBM had combined product revenue of about $200 billion last year), most observers forget that supply chain manufacturing is immature and largely untested.
Indeed, since outsourcing became the dominant manufacturing model in the late 1990s it has faced only one economic downturn, and it failed that test miserably. In 2001, electronics manufacturers ran into a perfect storm: a national recession, the dot-com implosion and the 9/11 terrorist attacks. In two years, the semiconductor, computer and telecom OEMs wrote down nearly $13 billion in excess inventory. Since nobody has repeated the business cycle, one wonders how the industry will fare the next time the economy goes boom, then bust.
Anatomy of a Supply Chain Disaster
What caused the industry's $12 billion inventory overhang? Bad forecasting played a part. As did a build-to-forecast manufacturing model based on materials requirements planning (MRP II) principles. While MRP may have worked for vertically integrated manufacturers in slow-moving industries where demand was relatively constant, it never stood a chance in today's fast-paced, outsourced economy where demand fluctuates rapidly and product life-cycles are measured in months instead of years.
The trouble with building to forecast is that the forecast is never accurate for long, states Gary Cortes, co-founder of FlowVision, a Lean Manufacturing consultancy based in Dillon, Colo. The $13 billion write-off happened because there was a complete disconnect between the OEMs, their contract manufacturers (CMs) and the component suppliers, he says. The OEMs gave overly optimistic forecasts to their CMs who then placed orders for components. The distortion rippled through the supply chain, growing larger as each tier added capacity and ordered more and more material. When the downturn came, there was a huge inventory pileup in place.
Larry Leveille witnessed the effects of bad forecasting first-hand. As a business unit director at Jabil Circuits, a $6 billion global EMS firm, Leveille recalls the go-go years prior to the 2001 crash as a euphoric period in which OEMs ignored the dangerous inventory build-up because demand seemed limitless. During the dot-com craze they didn't want to leave any sales on the table, recalls Leveille. OEMs cared only about sales; if you didn't deliver according to their schedule, price and quality, they would go elsewhere. They dropped their forecasts to the CM in [electronic data interchange (EDI)], and you filled the order, end of story.
As long as companies continue to build to forecast using MRP II push-production methods, another inventory disaster is just a matter of time, says Cortes. Most companies are doing a better job keeping inventory in check right now because they were burned last time and senior management is paranoid. But it's not systematic. At some point another bubble will emerge, experienced senior managers will forget or be replaced, and disaster will strike again. It's only a matter of time.
The First Steps...But Are They Working?
Of course, some OEMs learned from the last debacle that build-to-forecast production is dangerous, particularly in an extended supply chain involving hundreds or even thousands of suppliers. After all, the OEMs that insisted that contract manufacturers and suppliers build to their forecast ended up owning the unsold inventory. Now some of these companies are applying Lean Manufacturing practices to internal and external business processes.
While many CMs have Lean initiatives underway, nearly all of them are focused on internal production issues such as line-design, set-up and staffing.
It is almost impossible to do Lean throughout the supply chain without the OEM leading the initiative, comments Leveille. The reason? OEMs have built large planning, forecasting and contract management organizations with which to control the outsourced supply chain. In the 1980s, OEMs would tell their CMs to build a certain number of printed circuit boards. The CM would get quotes from component suppliers and give a bill-of-materials (BOMs) cost to the OEM who would approve it or not, he says.
But that changed once the OEMs began outsourcing the manufacture of complete systems. In order to mitigate risks and keep costs low, OEMs sourced finished products from multiple contract manufacturers. In addition, they insisted on controlling supplier selection as well as negotiating key terms and conditions of the contract with suppliers that govern pricing, volume, liability and so on.
With so little flexibility, the CMs and their suppliers had no choice but to respond to OEM forecasts as if they were gospel. OEMs may have outsourced manufacturing, but they were managing the supply chain with old-fashioned MRP tools designed for vertically integrated enterprises and command-and-control economies.
Of course, OEMs pay a heavy price for insisting on complete control. When the economy goes bust and their forecasts prove wildly inaccurate, they own the results, including all of the material flowing into the supply chain, the work in process, the finished goods inventory, the whole nut. That is why they ended up taking the lion's share of the $13 billion write-down. And it's also why some OEMs are beginning to change their tune and turn to Lean.
Certainly, they will find many willing partners among the forward-looking contract manufacturers and component makers in their supply chains who realize that in the long term Lean is important to their very survival. During the past decade, even well-managed firms that understood and tried to run Lean inside the four walls of their factories had trouble because the most influential members of their supply chains insisted on meeting pre-determined production schedules tied to unrealistic forecasts, irrespective of true demand.
How do you tell suppliers to only ship product based on actual demand (a key element of Lean known as kanban) when the OEM, and other key supply chain partners are responding to an inaccurate forecast? Furthermore, in many cases, sourcing contracts that are set up between the OEM and suppliers permit the supplier to ship components to the contract manufacturer even when they are not needed.
A New Way of Thinking
In today's globally outsourced economy, competition is no longer company against company, but supply chain vs. supply chain. Companies that want to run Lean must learn to do it collaboratively, in groups. In a Lean supply chain, traditional forecasting and MRP production scheduling techniques are used only for planning to run what-if scenarios and communicate trend analysis back to suppliers. But they are not used to order material or produce products. Only kanban signaling is allowed for that.
Kanban is a Japanese word that means signal or billboard. It is the most fundamental concept in Lean manufacturing used to describe a material replenishment process in which every stage of production signals the one preceding it when more material is required or an act of production has been executed and a new one is ready to begin. Unlike MRP systems, which push material through production according to a pre-determined schedule that is based on a potentially inaccurate forecast, kanban is a pull-production process that is based on true customer demand. Kanban eliminates all but a small reserve of inventory or safety stock that functions as a re-order point while enabling the manufacturer to respond to unexpected surges in demand or interruptions in supply. When material drops below this re-order point, a kanban signal is sent to replenish it.
OEM executives love Extended Lean in a downturn because the only inventory in the pipeline (besides a small safety stock) represents actual customer demand. But how does a Lean supply chain respond during an upturn? Can Lean contract manufacturers and Lean component suppliers respond quickly to an unexpected upsurge in demand? Or will Lean OEMs find themselves stuck in first gear while competitors race ahead? In other words, how responsive is Extended Lean since, by definition, it must orchestrate complex activities across multiple tiers in a supply chain in which the OEM has relinquished centralized control?
It is a fair question, and until recently it was the Achilles Heel of Lean as it applied to outsourcing and supply chains. That is why most proponents of Traditional Lean sequester themselves within the four walls of the factory. Traditional Lean makes sense in a vertically-integrated enterprise, or within the confined space of a factory where managers can remove wasteful steps and redesign the lines. But Traditional Lean does not work in a modern supply chain.
Extended Lean makes the supply chain responsive by using a new technique known as Statistical Kanban, which was developed by Gary Cortes and his colleagues at FlowVision. How does it work?
The Lowdown on Statistical Kanban
Statistical Kanban enables manufacturers to anticipate fluctuations in demand and meet virtually any guaranteed service level that a customer requires, says Cortes. Let's say we have to carry some level of finished goods inventory to guarantee a specific level of service required by an OEM customer. This is inventory that the OEM has ordered and agreed to purchase. We can statistically determine the ideal level of inventory that is needed at each stage of the supply chain based on historical usage patterns and trends, says Cortes.
Statistical Kanban is used to guarantee a service level for which a customer is willing to pay. For example, FlowVision is working with a company in San Jose, Calif., that makes complex electronics-based systems with a long supply lead time and for which there is high demand. The company wants to guarantee that it can meet 99.7 percent of orders within a short delivery schedule. Using Statistical Kanban, FlowVision calculates the amount of finished goods inventory required to meet 99.7 percent of orders within that company's guaranteed time-frame. Using Statistical Kanban, the company guarantees that at most it will miss only 3/10ths of one percent of orders. Its customers are delighted with that level of service.
In order to achieve that level of service, we need to have a certain amount of material in the pipeline, explains Cortes. At a minimum, all of our components must achieve a 99.7 percent service level. When we statistically size the inventory, whether for finished goods or components, we mathematically calculate the variation in historical usage of those components and that finished product. We also look forward to account for any future anticipated changes in demand. Then we size the inventory to achieve a specific confidence level for which the OEM customer is willing to pay. Obviously, the higher the confidence level, the more inventory we will require, and the higher the cost to the customer. But the customer knows that up-front and can choose according to different scenarios. What is the financial impact of moving from a 99.5 percent service level to a 99.7 percent service level? With Statistical Kanban you can tell exactly what those additional two-tenths percent of certainty are going to cost.
Properly applied, Statistical Kanban is the key to implementing Extended Lean throughout a supply chain. Once the OEM, contract manufacturer and supplier agree on a guaranteed service level, they can mathematically determine the level of finished goods inventory. At that point, the only time they produce more of that particular product is when there is a signal from the distribution center or wherever the finished goods reside that the OEM has reached a re-order point. Then and only then does the supply chain produce more.
Says Cortes, We still use forecasts to tell us where we think we're going, and we use MRP to establish our lead-time offset, so we know how long it will take to order, receive, build and ship a product, but we are not going to order or build anything unless the customer actually buys a product. It is a completely different mindset from how most manufacturers operate today, particularly in a complex supply chain where the OEMs talk to their sales guys, estimate demand, add a little extra, then buy material and fill up their finished goods inventory. And if the forecast is wrong? Hey, it's just another multibillion dollar write-off.
About the Author:John Rumasuglia is the president and CEO of Extended Lean Solutions, an Extended Lean education and consulting firm in San Jose, Calif. Write to him at [email protected].