The Supply Chain as Growth Driver

The mandate from SanDisk's leadership was clear: the company's supply chain operations would need to support a planned five-fold increase in business. And the only way that would happen would be through a dramatic improvement in supply chain performance. Here's how they responded to the challenge—through a transformation of the demand planning, S&OP, and excess inventory-disposition process.

By Bill Paganini and John Kenny -- Supply Chain Management Review, 5/1/2007

• Boosting the "Core Strength"
• Role Reversal for Product Marketing
• A Return to Quantitative Forecasting
• Input from Customers
• Reinforcing S&OP Activities
• Strength in Numbers
• Managing Excess Inventory
• Soft Benefits Count, Too

There's a good chance you already own some of SanDisk's products—a memory card for your digital camera or a Sansa MP3 player, perhaps, or a flash drive for sharing your data or taking it with you. The company has been growing at a phenomenal clip, hitting 2006 revenues of $3.3 billion—nearly three times its top line in 2003. SanDisk Corp., headquartered in Milpitas, Calif., has one of the highest revenue per employee ratios in the technology industry.

But SanDisk's top management team saw even more performance potential. So partway through 2005, they started setting the bar much higher. They began mapping out what it would take to get to annual revenues of $10 billion—five times the company's size at that time. The challenge for their operations managers: figure out, evaluate and implement the business processes that would sustain a company of that size.

The authors of this article will describe how SanDisk's managers quickly overhauled their supply chain operations to meet the tough new mandate. Specifically, the program focused on tearing down silos and building a high-performance demand management process supported by integrated sales and supply chain teams. It called for formalizing the demand management function and its leadership. It also meant defining the sales and operations planning (S&OP) process in ways that would meet the objectives of both sales and supply chain groups, while working to reduce the amount of time spent preparing for and conducting S&OP meetings.

The article will also drill down into inventory asset management—a process that is often overlooked at many companies—to show how SanDisk uses this low-overhead method of identifying and disposing of at-risk inventory to close the loop on operational excellence.

A little more than a year later, SanDisk's revenues are up nearly 50 percent and its critical fourth-quarter business – during which the company can pull in more than 40 percent of its revenue and profits – has run like clockwork for the first time. Underlying operational metrics have improved substantially: Retail inventory turns are 20 percent better than a year earlier and on-time delivery is up by 30 percent. And the company continues to hold its number one market share position worldwide in flash memory products.

To an athlete, “core strength” refers to good condition in the back, abdominal, and side muscles. In the case of a manufacturing company, the term applies to the robustness of its demand management and S&OP processes. Essentially, SanDisk's conditioning exercises have helped the company reach new heights in performance (revenue) while increasing stamina (productivity) and avoiding injury (margin loss and customer dissatisfaction).

The conditioning exercises were crucial. Reliable, repeatable, universally understood planning processes provide the fundamental basis for the financial plan, the shipment plan, and the build/inventory plan. Getting it right at the core allows managers to focus on the relatively straightforward processes of execution. But getting it wrong throws the organization into the chaos of expediting, second-guessing, and overspending in a frantic bid to hit its quarterly financial goals.

In the third quarter of 2005 came the agreement to have a formal S&OP team and a demand management team; SanDisk's chief operating officer asked co-author Bill Paganini to head up the teams. By the first quarter of 2006, Paganini and his lieutenants had named strong teams, in some cases hiring from the outside to get the experienced demand planners and business analysts they needed. The team comprised business and data analysts and i2 demand management system specialists—11 highly skilled individuals in all.

At about that time, the company had gone live with i2's Demand Manager module. The tool would help the groups' planning activities, eliminating functional ad hoc databases and combining all relevant numbers in a single database to provide one system of record and enhance coordination among different functions.

The foremost objective for the new team was to reduce forecast errors. Unfortunately, the high-tech sector has come to accept large margins of error in forecasting—20 percent and more is not uncommon—with the blame pinned on the vagaries of consumer demand, the velocity of new product introduction, and unknowns from competitors. As they do in any business, the flawed demand projections were preventing key supply chain managers from understanding what the operating plan really called for. Without structured sales and operations planning activities, imperfect forecasts effectively drove the plan.

So SanDisk management set a goal of halving the forecast error inside a year, recognizing the substantial downstream benefit of getting both the mix forecast and aggregate quantities right the first time. The tightening of demand management would reduce supply chain noise and minimize the bull-whip effect for suppliers and component suppliers. More importantly, the critical measures of customer on-time delivery and retail shelf stock-out rates would be dramatically improved at the same or reduced levels of operational expense and inventory. The plan called for improvements of around 10 percent on retailers' in-stock levels, roughly 20 percent advances in on-time delivery and reductions of about 20 percent in retail inventory—finished goods and inventory in the channel.

The demand management team recognized that in order to get the kinds of results being requested, it was critical to step back and look at where the business was headed. It quickly became clear that if forecast error was to be cut by half, each of SanDisk's two fundamental businesses, retail and OEM, warranted its own distinct demand management process. The retail side is what you see when you visit a major electronics retailer: blister packs of memory cards, flash drives, and so on. The process improvement efforts were focused primarily on the retail business for two reasons. First, it comprises about 70 percent of SanDisk's business, so any improvements there would drive the greatest pick-up in demand management for the company as a whole. Second, there was no clearly defined process for forecasting the retail business, and the continuous challenges of new product introduction, channel inventory targets and promotional cycles can quickly drive the retail forecasting process out of control.

Attention then turned to SanDisk's forecasting approach. It was the typical bottom-up route familiar to most manufacturers, with the sales organization having the primary responsibility for determining demand across the whole one-year forecast horizon. The territory forecasts were rolled up into the regional forecasts and those, in turn, were rolled up to the worldwide forecast. At the end of the process, the product marketing group was responsible for contributing information on the projected impact of new product phase-in over a given quarter. At that time, the S&OP process didn't do much more than aggregate the two numbers (sales' bottom-up forecast and marketing's top-down projections) and hand over the result to the supply chain organization, which translated it, more or less verbatim, into the production and inventory plans.

By the second quarter of 2006, the new demand management team members knew their roles well and understood the processes. They soon determined that the inherent “wrongness” of the legacy forecasting process stemmed from the poor assignment of responsibility and accountability across the sales and product marketing departments, and to a lesser degree across the finance and supply chain operations. Sales was being asked to take ownership of developing a demand forecast for 12 months even though they typically lacked market visibility beyond four months. And the product marketing unit's insight into longer term market and technology trends was vastly underutilized.

With such fundamental flaws at work, it was not surprising that adversarial relationships flourished among the four operations as each sought to place blame elsewhere for missed targets and production cost overruns. The single point on which there was unanimity was the need for a better process. In something of an epiphany, SanDisk's management team recognized the need for two fundamental process changes: role responsibilities and forecasting methodology.

The first and arguably the most profound process change in demand management came with a near-complete role reversal for the sales and product marketing groups. It was not an easy sell. The product marketers had traditionally been incented to quickly devise winning products, with the sales group handling the inherently erroneous process of developing a 12-month demand projection based on four months of product line visibility.

But since the product marketers were the ones with visibility of the overall size and growth of the flash memory market and of the dynamics affecting SanDisk's worldwide market share, they were asked to dramatically expand their roles to handle product-line forecasting over the lifecycle of a product—a move that effectively aligned accountability with visibility. In essence, the product marketers now “own” the forecast. They are the primary source of worldwide demand numbers for the entire forecast horizon, incorporating both product phase-in and phase-out cycles. Sales is hardly off the hook, however; they still are accountable for attaining quarterly numbers.

The process has become a collaborative one, with a much savvier product marketing unit “setting the table” with the baseline 12-month forecast that incorporates trends, product lifecycles, competitive issues and other market factors. The sales teams corroborate with their specific knowledge of channel demand for the next one to two quarters. The result is shared accountability, and a demand forecast upon which both organizations fundamentally agree.

However, although the process had been improved by aligning accountability with the appropriate knowledge source, forecast accuracy still fell short of the target. Where could the team go for further improvement? The answer was to address the forecasting methodology—– specifically to revisit quantitative techniques.

In the heyday of MRP II and ERP deployment of the 1980's and 1990's, quantitative forecasting lost favor. It was considered to be backward-looking and thus inherently wrong—especially in high tech where entire product lines can change to new product within one year. So SanDisk's demand management team worked somewhat under the radar for about three to four months to incorporate, aggregate and analyze the company's demand data. Gradually and steadily, the team was able to develop and prove mathematical models that would characterize repeated patterns—or pattern silos—in the company's demand data. Factors such as seasonality, new product introduction cycles and channel inventory modulation came to light in the mathematical models that the team developed.

The quantitative models provide another set of guide points to add to the inputs from sales and product marketing; they helped support the consensus process, which we will discuss later in this article. Although forecasting using backward-looking data can be misleading in a fast-paced environment where products change constantly, several valuable patterns emerged from SanDisk's data. These patterns have become regular features in the development of the demand plan today.

First and most apparent was the emergence of seasonality trends, for shipments to the channel as well as for sell-through out of retail. The team found that these trends were marked and repetitive, based on both quarter-to-quarter relationships and on demand trends within a quarter. They vary by channel type and region and could be codified fairly well given the five years of forecast data that were available.

SanDisk found that product changeover could be an ally or a foe in the development of an accurate forecast. The same wealth of data that helped with seasonality also contributed here. Most SanDisk products are sold at a specific memory capacity point—one-, two-, four- and eight-gigabyte flash drives, for instance—and those points increase over time as technology marches on and lower prices bring higher capacity products into the consumer's sweet spot. At each point there's a clear lifecycle pattern that can be used as a statistical starting point for planning the next point using the proper time offsets.

A third factor played a key role. The final contribution to the improvement in SanDisk's forecast accuracy came from combining the statistical forecast efforts with regular input from customers. Formally known as Collaborative Planning, Forecasting, and Replenishment (CPFR), this methodology essentially consists of the customer (the retailer) and supplier meeting regularly (monthly in SanDisk's case) in a formal setting to discuss demand planning over a relatively short time horizon—three to six months, for instance. The CPFR process enables the sharing and testing of the assumptions upon which the forecast was developed, resulting in improved accuracy and shared accountability. For SanDisk, it represented the last mile in forecast accuracy.

SanDisk had used CPFR techniques and processes informally, but management's mandate for operational excellence opened the gate for investment in tools and talent. Several members of the demand management team are part the regular CPFR meetings. The topics covered in the meetings are out of the range of mathematical models, marketing forecasts and sales forecasts. They include: promotions planning; merchandizing plans; changes in store inventory policy; advertising (rotations and timing); target-setting for joint channel inventory weeks-of-supply to modulate the flow of inventory; consumer electronics market activity and the associated boost that SanDisk might experience from launching products such as cameras and cell phones; and the distribution channel's perspective on broad market trends relating to market size and share. Gone are the hair-pulling days of “being dinged on customer delivery performance for product I cannot ship and you do not need!” Today there is a strong partnership and level of trust with key retailers, where the retailers look forward to conducting the planning process as much as does SanDisk.

Collectively the three factors described—role reversal for sales and product marketing, quantitative modeling and CPFR—contributed to the performance breakthrough that SanDisk sought in forecast accuracy. The year-over-year comparison is eye-opening: In 2005 SanDisk was only 40 percent successful in consistently attaining its forecast accuracy levels. With an improved demand management process in place, forecast accuracy has surpassed the goal by more than 20 percent. More significant is the fact that the accuracy goal incorporates both the aggregate unit forecast and the mix forecast. In fact, all of SanDisk's “balanced scorecard” metrics have improved significantly.

The most profound and long-lasting impact, however, may surface in the behavioral changes evident in the demand management process. In effect, a resilient, adaptive, cross-functional high-performance team has been formed out of a disparate group of functional adversaries. The monthly demand management meeting, run by the 11-member team, was intentionally renamed the “Consensus Meeting” in order to accurately characterize the open collaboration between the groups.

Sales and marketing managers do have their disagreements, of course, particularly in the challenging period one quarter out from the current quarter. But all stakeholders have moved to manage the discussion in a collaborative way, proactively seeking input and direction from both quantitative models and direct input from channel partners. And the demand management team drives a continuous improvement process of Input / Manage / Revise / Check until the Consensus Meeting participants agree on and support the final numbers.

By the fourth quarter of 2006, SanDisk's demand management team was running very well, and was confident in its ability to “get it right, from the top” in the form of improved forecast accuracy. The next step in the operational improvement effort was to upgrade S&OP.

The primary function of SanDisk's S&OP process is to integrate the sales plan, operations plan, and the financial plan. The details, of course, are where the rubber meets the road. The demand management process has yielded believable, consensus-driven, 12-month forecasts of units and product mix. The outputs of the S&OP process include longer-range capacity planning for SanDisk's memory-chip fabrication plants, financial plans for revenue and margin, demand generation actions required, assembly capacity planning, and the driving of production and inventory plans.

The process has become another nexus for consensus, this time incorporating the head of each functional area: sales, marketing, supply chain, and finance. In fact, since the third quarter of 2006, the S&OP meeting is now embedded in an expanded executive staff meeting, taking place over three hours on a Friday afternoon during the first fiscal week of every month. S&OP has become “the way to run the business.” SanDisk's COO has mandated a “No Fly Week,” during which no executive can travel far when the S&OP meetings are taking place.

Supported by the demand management team's tools and models, the executive S&OP meeting is the forum for agreeing on the supply plan needed to execute the demand plan, and for identifying risks and determining the actions required to ameliorate the causes of those risks. The participants review the finished goods inventory strategy, consisting of the distribution center and channel stocking levels by configuration by geography. The S&OP process also reviews near-term manufacturing capacity plans and capacity levels. In support of the monthly S&OP process at the executive level there is also a weekly S&OP meeting where managers “trim the sails” —assessing supply and demand performance against plan, eyeing any nascent disconnects, and making quick course corrections.

With multiple new product introductions (NPI) each quarter, a critical function of the S&OP process is to shine a bright light on the NPI process, particularly when new products constitute a major part of the quarter's projected revenue. Here again each functional area shares accountability: Product marketing and supply chain together ensure that the release to production handoff is executed flawlessly. For their part, the sales teams must ensure that their quarterly quotas incorporate the new product targets and they are equipped with the tools and training necessary to attain those quotas. And the finance department must make sure that the finance and order processing systems are ready and that all data set-up issues relative to the new products have been addressed.

S&OP is a way of seeing the big picture laid out as quantitatively as possible. Each meeting is used both to “tell the story” and reach agreement on the numbers and on the actions required. The meetings also serve to provide the longer-term strategic guidance for future quarters from the executive committee. This guidance is incorporated into the “assumptions” package that provides a starting point for the next monthly planning cycle. The overall result is powerful and all-encompassing: a shipment plan that becomes the financial plan, which in turn drives budgets and even Wall Street briefings. Shared accountability leads to “One company, one plan.”

Together, the strengthened core processes of demand management and S&OP have earned widespread executive buy-in at SanDisk. The results confirm their endorsement. Against an overall target for on-time delivery, SanDisk has seen more than a 30 percent improvement, moving to within 10 percent of the goal. Retail excess inventory has been reduced by nearly 40 percent—far ahead of goal. The other half of SanDisk's inventory metric is the measure of channel inventory, or channel weeks of supply (WOS). Here, the CPFR process has helped the partners to reduce channel inventory by more than 20 percent while avoiding the downside exposure of inventory stock-outs. And there has been a notable pick-up in in-store stocking levels—a critical performance indicator for a retailer, and in turn for the manufacturer that is competing for the retailer's mindshare. At the inception of SanDisk's CPFR process, performance against this metric was below 80 percent of target; the new process brought performance to target within one year.

SanDisk's crucial fourth-quarter performance is better than it has ever been—a vital touchstone for any company so dependent on retail business. (It's typical for more than 40 percent of yearly revenues and earnings to come through in Q4;) The company even managed to open up its market-share lead over competitors in Q4 2006. The company's executives did acknowledge the improvements throughout. And retailers have not been slow to show their gratitude for SanDisk's help in setting them up for success during a busy shopping season.

Collectively, the significant steps forward in demand management process and S&OP have led to better supply chain execution and predictability for SanDisk. Combining those gains with the externally facing CPFR programs with SanDisk's channel partners has helped accelerate inventory velocity and effectiveness and drive big gains in in-stock levels while maintaining channel inventory targets. But there has been one more initiative that since 2005 has been tweaking operational excellence that bit more.

For all of its product successes, revenue growth, and operational performance improvements, SanDisk is not immune to excruciating margin pressure. So in 2004, the company had turned its attention to identifying at-risk or excess inventory and moving it through secondary markets, converting idle assets to cash.

Launching more than 20 new products each quarter, SanDisk's supply chain managers had not been able to focus effectively on the constant stream of products that are being made obsolete. The best efforts yielded on average 20 cents on the dollar in recovery—this in a market with insatiable aftermarket demand for flash memory-based products. And the disposition efforts, typically happened only twice a year, when the excess inventory had built up to a point where it began to be a physical obstacle.

At the same time, the company simply did not have the resources to manage the process of routine identification of excess, finding market opportunities and managing the incremental transaction flow. Adding the two or three personnel who could ensure that the job was done properly was out of the question: any additional headcount would be applied where it mattered much more: in demand management. So SanDisk chose to outsource to a specialist in end-to-end asset recovery—meaning everything from at-risk inventory identification to market research and pricing recommendations right through to market-making and identification of buyers—and to final settlement, collection, and other operational issues.

Early conversations began in the fourth quarter of 2004 with FreeFlow, a specialist in at-risk inventory identification and disposition. By March 2005, FreeFlow and SanDisk together agreed on the definition of at-risk inventory and then began mapping out the flow of product returns across all three geographic regions: North America, Europe and Asia Pacific. Over the next two months, the inventory asset recovery team — including SanDisk's inventory managers, product specification engineers, product marketers, and representatives from the finance department—thrashed out inventory management policies as well as target pricing and reserve pricing for inventory auctions.

The SanDisk participants sensed a significant opportunity to generate new revenues. For most technology products, price erosion is steep, beginning just months after new product introduction. For flash memory-based products, it can be as high as 15 percent per quarter. Late-stage inventory is already well down the price curve, making it imperative to have crisp decision-making thresholds, pre-defined markets, and low-overhead processes. If its disposition process could take place every month instead of twice a year, SanDisk could essentially claim back two quarters' worth of product value.

FreeFlow set up a monthly series of remote auctions—meaning the inventory is sold on site, at SanDisk's distribution center, with the winning bidder paying the cost of freight. (Previously, SanDisk had usually incurred the shipping costs.) With the ubiquitous presence of online auctions (thanks to eBay), SanDisk took FreeFlow's recommendation to establish a private, branded auction platform for the liquidation of its excess inventory. The platform, SanDiskExcess.com, allows SanDisk to approve the members, the minimum price points and all bids. Competitive bidding alone — helped by attention to the metric of bid spreads — has shown consistent performance improvement of more than 30 percent for SanDisk. On top of that, SanDisk no longer has to spend on the marketing promotions, rebates, and other programs usually needed to dispose of surplus inventory.

Throughout, SanDisk used its Progressive Disposition Process (PDP) matrix, developed with FreeFlow to facilitate review and decision-making with respect to excess inventory. The matrix integrates sales and supply chain operations, inventory policy and product lifecycle into a concise, threshold-driven decision support system. It prescribes a distinct course of action for three categories of inventory:

1. active and obsolete new inventory still in the company's warehouses;
2. active and obsolete new inventory in the channel; and
3. returned inventory, itself broken out into three categories, with “defective” product to be refurbished for resale.

For each category the complete decision-making process, accountability, and action is specified: inventory threshold, product lifecycle status, frequency of review, process owner, and the disposition action.

Over the past 15 months, SanDisk has made huge strides in improving its operational performance. It has realigned responsibilities and accountabilities for the sales forecast, with its product marketing group emerging as key driver of the global forecast over the 12-month planning horizon. The recently formed demand management team has helped boost forecast accuracy by driving the demand management process and providing the technical expertise to incorporate relevant mathematical models into the forecast process.

SanDisk has also made disciplined use of a single system of record for demand management—i2's Demand Manager tools—to ensure that no time is wasted debating over disparate data sources. The company has also applied CPFR processes, getting its retail customers' input in order to close the gaps in its channel inventory and retail in-stock targets. It has elevated its S&OP processes so that they have become embedded in the monthly duties of the company's most senior executives. Finally, SanDisk has paid close attention to the details, collaborating with an asset recovery expert to accelerate the disposal of excess inventory and expand a previously neglected source of revenues.

But just as impressive as the hard numbers are the soft benefits of more effective collaboration within the company. That all by itself is a lesson well learned.